UCNJU N I O N C O L L E G E O F UNION COUNTY, NJUndergraduate Research JournalVolume 8 | No. 1 | Fall 2025
UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8 | No. 1 | Fall 2025 UCNJ does not discriminate and prohibits discrimination, as required by state and/or federal law, in all programs and activities, including employment and access to its career and technical programs.Editorial TeamEditor-in-ChiefMohamed Mohamed, Ph.D.Director of Student Research and Science LaboratoriesAssociate EditorMelissa Sande, Ph.D.Associate VP of Academic Affairs & Dean of HumanitiesShuchi Agrawal, Ph.D.Assistant Dean of STEMOlubisi Ashiru, Ph.D.Academic Specialist, BiologyMirza Baig, Ph.D.Academic Specialist, BiologySunjin Jo, Ph.D.Academic Specialist, ChemistryYohan Kim, Ph.D.Academic Specialist, BiologySanaz OghlidosAcademic Specialist, BiologySusana Sequeira, AIAAcademic Specialist, ArchitectureFaraz Siddique, Ed.D.Dean of STEM
1 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025TABLE OFCONTENTS03Enhancing Phishing Detection using Machine Learning: Comparison of Text-Based and Image-Based Approaches10Unmasking Reality: Distinguishing Genuine Content from AI-Generated Facades16Efficacy of Herbal and Spice Extracts at Different Concentrations Against Grampositive and Gram-negative Bacteria22Discrepancy Between Cost and Effectiveness of Cleaning Agents: Role of Active Ingredients 26Heat-Induced Gene Expression in Tomato Plants: A Molecular Response to Climate Change32Natural Food Preservation Using HouseholdIngredients: A Chemistry-Based Investigation
2 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025MESSAGE FROM THE EDITOR We are excited to announce the release of the Fall 2025 edition of the UCNJ Undergraduate Research Journal, underscoring its vital role as a platform for showcasing our students' innovative work. The journal has become a symbol of academic excellence at UCNJ, offering students a unique opportunity to share their discoveries with a broader audience. The journal's continued success reflects UCNJ’s dedication to fostering a culture of inquiry and scholarly engagement. We believe that allowing students to experience the process of publishing an article is transformative, empowering them to become future leaders in their fields. The impactful research featured in these pages demonstrates our students ' accomplishments. This volume highlights the diverse academic interests thriving on our campus. Readers will find a range of original research, including articles on cybersecurity, enhancing phishing detection, and unmasking reality. An interesting study examines the molecular physiology and gene expression in stressed plants, providing new insights into their adaptive mechanisms. We also feature two articles on microbiology—one examining the efficacy of modern cleaning agents and another exploring the antimicrobial properties of various spices. Additionally, another contribution examines innovative methods for natural food preservation. We want to express our appreciation to UCNJ President Dr. Margaret M. McMenamin, Executive Vice President and Provost Dr. Maris Lown, and Vice President of Academic Affairs Dr. Sara Lacagnino, whose support has been instrumental in the growth and development of this publication. My sincere thanks also extend to the dedicated editorial team, the faculty mentors who guided our students, and the lab staff for their tireless efforts. Looking ahead, the UCNJ Undergraduate Research Journal remains committed to promoting excellence in undergraduate research. We invite UCNJ students from all disciplines to prepare and submit their work for our upcoming issues. Thank you for your continued support of UCNJ's dedication to academic innovation.Sincerely,Mohamed Mohamed, Ph.D. Director of Student Research and Science Laboratories UCNJ Union College of Union County, NJ
3 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Enhancing Phishing Detection using Machine Learning: Comparison of Text-Based and Image-Based ApproachesCedric Desinor, Jaswinder Singh, Olivia Chow, Rudolph A. FenelonMentor: Academic Specialist Madhusudhana Reddy STEM Division, UCNJ Union College of Union County, NJAbstract: Phishing attacks, fraudulent emails, fake websites, and deceptive social media messages are employed to steal sensitive information from individuals and organizations. This study explored two machine learning-driven phishing detection techniques: (1) textbased analysis utilizing natural language processing (NLP) and (2) visual-based detection with convolutional neural networks (CNNs). For textual phishing attempts, our system uses Bidirectional Encoder Representations from Transformers (BERT), a leading NLP model, to analyze language patterns in messages. The model achieved outstanding performance, with 99.14% accuracy in spam detection, along with high precision (0.99) and recall (0.95), ensuring dependable identification of malicious content. For phishing websites, we trained a CNN on the Phish-Iris dataset to detect visual cues, such as altered logos, irregular layouts, or color mismatches, achieving a classification accuracy of 86.4%. IntroductionPhishing attacks remain a major cybersecurity threat, using deceptive emails, fake websites, and fraudulent messages to steal sensitive data. This study examines two effective methods: text-based phishing detection using BERT (Devlin et al., 2019), a state-of-the-art NLP model, and visualbased detection with Convolutional Neural Networks (CNNs) to identify website inconsistencies (Dalgic et al., 2018). Machine Learning (ML) based phishing detection methods generally fall into two categories: text-based detection using NLP and image-based detection using Computer Vision (CV). Text-based phishing detection involves analyzing textual elements within phishing attempts. Natural Language Processing (NLP) is a branch of artificial intelligence that enables computers to understand, interpret, and respond to human language. Image-based phishing detection, on the other hand, utilizes Computer Vision Techniques to assess the visual authenticity of websites and graphical elements involved in phishing attempts. This paper compares text-based and imagebased approaches for phishing detection using machine learning.MethodologyA. BERT Text-based methodFigure 1: Flowchart of the process of a Natural Language Processing system
4 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Bidirectional Encoder Representations from Transformers (Patat, 2023) is an innovative machine learning model that learns from text and helps us understand the meaning of words in a sentence. What makes BERT special is that it looks at the whole sentence, not just one word at a time. For example, in the sentence “He went to the bank,” the word “bank” could mean the side of a river or a place where you keep money. BERT looks at the other words in the sentence to figure out which meaning is correct. It reads the sentence both forward and backward, so it can better understand the whole meaning than older systems. To test an effective spam detection system, begin with a dataset of 5,574 text messages (Smith, 2022) as teaching examples for the model. This dataset includes both spam (unwanted/junk messages) and ham (legitimate messages), with a noticeable class imbalance: only 13% are spam, while 87% are ham. This imbalance presents a challenge because, if left unaddressed, the model might simply default to always predicting \"ham\" and still achieve high accuracy without effectively learning to detect spam. The preprocessing pipeline starts with text normalization, which cleans and standardizes the messages. First, apply case folding by converting all text to lowercase (e.g., converting \"HELLO\" to \"hello\") so that the model treats word variations consistently. Next, remove special characters, emojis, and punctuation using regular expressions (Regex), eliminating anything that is not alphanumeric to focus on meaningful content. Figure 2. Distribution of LabelsThen tokenize the messages using NLTK's word tokenize function, breaking sentences into individual words or subword units, such as outlining key points from a paragraph. Finally, normalize whitespace by removing extra spaces between words to create clean, uniform text inputs.Since computers process numbers more efficiently than text, our categorical labels were converted to numerical values using label encoding, assigning \"1\" to spam and \"0\" to ham messages, as shown in Figure 2. Additionally, we create Boolean flags (True/False values) that make the data compatible with PyTorch and other machine learning frameworks, ensuring seamless integration with thetraining pipeline. This thorough preprocessing stage converts raw, unstructured messages into a format that the BERT model can effectively learn from, while maintaining all the linguistic patterns necessary for accurate spam detection. Figure 3. Splitting the dataTo train the model effectively, the dataset was split into two parts: a training set (75%) and a test set (25%), as shown in Figure 3. The training set, comprising of 4,180 messages, was used to train the model, while the testing set consisted of 1,394 messages. Stratified sampling ensured that the proportions of spam and ham messages remained consistent across both subsets, enabling a balanced and fair evaluation. Analyzing message length revealed that the average message contained 18 tokens, and95% of the messages had 38 tokens or fewer. To ensure consistency and compatibility with the BERT model, all messages were limited to a maximum
5 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025of 128 tokens, a constraint that BERT can handle efficiently.Started by setting up the BERT model as described by Smith, 2022, with two key parameters: MAXLEN=128 limits the model to analyzing only the first 128 words of a message (like skimming the most important parts of a long text), while batch size=32 means the model learns from 32 messages at once (similar to studying flashcards in groups for better efficiency). Before training, preprocess the text through tokenization—breaking sentences into words or subworlds (e.g., \"You won 100!\" becomes [\"You,” “won\" \" 100!\" becomes [\"YOU,” “won,\" \"100,\" \"!\"])—and numerical encoding, where each word gets a unique number (like \"won\" = 2048) so the model can process it.For training, we use fit one cycle, a method that adjusts the learning rate in three phases (Leslie, 2018): starting slow (like training wheels), accelerating (riding freely), and then fine-tuning (mastering complex parts). After each complete training pass (epoch), we evaluate performance using the training loss (mistakes on practice data), the validation loss (mistakes on new data), and accuracy (correct predictions, including spam/not-spam).This approach works because clean data acts as well-organized study material, structured training promotes efficient learning, and regular testing confirms genuine understanding. The model behaves like an ideal student—focusing on key details (MAXLEN), learning in batches (batch size), pacing itself (fit one cycle), and verifying progress (evaluation metrics). Together, these technical components create an accurate spam detector that minimizes errors, functioning like a perfectly tuned learning system. B. CNN – Image-based methodA Convolutional Neural Network(CNN)functions like a detective solving a picture puzzle through four smart steps (Dalgic et al., 2018). First, it scans images with \"magnifying glasses\" (filters) to identify small details, such as edges or colors, similar to noticing a cat's pointy ears. Second, it simplifies clues by zooming out (pooling), emphasizing important features while ignoring minor differences, such as fur patterns. Third, it analyzes which clues matter most (activation function), remaining strong in matches (\"pointy ears mean cat!\") and discarding weak guesses. Lastly, a \"boss layer\" makes the final decision (\"85% cat, 10% dog\"). When spotting fake websites, the CNN detective looks for suspicious clues, such as blurry logos, oddly placed buttons, or mismatched colors, just as one would recognize a poorly copied dollar bill. What makes CNNs remarkable is that they learn these skills by studying thousands of examples and are used in cutting-edge technologies such as facial recognition on phones and self-driving cars.Figure 4. Flowchart for the classification of websitesFor logo checks, CNN compares every pixel against stored genuine versions, flagging imperfections such as blurry edges (photocopied-quality logos), distorted proportions (a stretched Amazon smile logo), or color mismatches (Netflix's red appearing slightly orange)—like spotting a fake sneaker’s misaligned Nike swoosh. Next, it analyzes button layouts by scanning for rectangular shapes and cross-referencing their positions with known \"maps\" of legitimate sites (e.g., login buttons should be positioned in the top-right corner,
6 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025and search bars should be centered at the top). Suspicious placements, such as login fields floating oddly or duplicate submit buttons, trigger alerts, much like a face would look \"off\" with misplaced eyes or two noses. The CNN also conducts a color forensic audit, creating histograms to quantify pixel shades and comparing them to reference \"fingerprints\" (e.g., valid Facebook’s 45% #1877F2 blue). Deviations, such as oversaturated blues or missing gradients, raise red flags, mirroring how art experts uncover forged paintings by analyzing the distribution of pigments.The CNN then weighs all the evidence, much like a jury deliberating a case. Each clue is assigned a level of importance: logo matches (35%), button layouts (25%), color schemes (15%), and other factors, such as missing copyright text (25%). Using these weighted inputs, it calculates a phishing probability score—for instance, a site with a 72% logo match, 20px off buttons, and a +12% blue saturation might receive a 87% phishing risk verdict. This systematic approach mirrors how humans intuitively combine multiple subtle cues (e.g., poor print quality, odd stitching, and incorrect label font) to identify counterfeit products, but the CNN does it mathematically, at scale, and without ever becoming tired.Table I. Phish-Iris DATASET used for Training.The model was trained on the Phish-Iris dataset. This dataset serves as a critical resource for training and evaluating convolutional neural networks in phishing detection, providing a carefully curated collection of website screenshots that encapsulate real-world cyber threats. Comprising 1,313 training images and 1,539 testing images, as seen in Figure 2. This dataset provides a solid foundation for machine learning models to learn and test detection capabilities. The dataset comprises 15 distinct categories, including 14 carefully crafted phishing examples targeting high-profile entities such as Amazon and Adobe, as well as other commonly impersonated services. Additionally, there is another category that contains legitimate websites for comparison. Using Python libraries such as TensorFlow, NumPy, and Pandas, the dataset was cleaned and compiled into a model. Results and DiscussionThis system performed like a straight-A student, achieving an impressive 99.14% accuracy on new messages, as shown in Figure 4 and Table I. This means that out of 100 texts, it would correctly identify 99 as either spam or legitimate messages, making only one mistake. The system demonstrated exceptional precision with a 0.99 score, functioning like an ultra-careful email filter that rarely sends legitimate messages to your spam folder. For every 100 messages it flags as spam, only one might be a standard message you would want to see. Its 0.95 recall score shows it captures 95% of all spam, acting like a highly vigilant security guard who spots 95 out of 100 shoplifters, with only the most cleverly disguised spam messages (about 5%) slipping through its defenses. Figure 5. Classification result flowchart
7 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Table II. Training and Validation PerformanceEpoch Steps Time/StepsTrain. LossTrain. AccVal. LossVal. Acc.1/10 131 15 0.1697 0.9526 0.0435 0.98642/10 131 12 0.0295 0.9933 0.0366 0.98783/10 131 12 0.0219 0.9940 0.0416 0.99144/10 131 12 0.0096 0.9974 0.0511 0.99215/10 131 12 0.0029 0.9998 0.0537 0.99216/10 131 12 0.0163 0.9957 0.0623 0.99077/10 131 12 0.0040 0.9995 0.0486 0.98788/10 131 12 0.0200 0.9964 0.0354 0.98719/10 131 12 0.0018 0.9993 0.0554 0.991410/10 131 12 0.0003 1.0000 0.0570 0.9914The CNN functions through three main components that work together to process and classify visual data. First, the convolutional layers serve as the network's feature detectors, using small 3×3 pixel filters that systematically scan across the input image. These filters identify both basic visual elements, such as edges, color variations, and textures (low-level features), as well as more complex components, including buttons, logos, and text fields (high-level features). For example, when analyzing website screenshots from the Phish-Iris training set, these layers effectively detected subtle imperfections in cloned logos, such as the slightly distorted 'smile' arrow in fake Amazon logos. Pages (class 1) or the pixelated 'A' in counterfeit Adobe login screens (class 3). Second, the pooling layers streamline this information by reducing the image dimensions while preserving the most significant features through max-pooling. This process proved particularly effective in handling variations within the dataset, such as different placements of login forms across phishing samples impersonating PayPal (class 7) while consistently recognizing their suspicious nature. Finally, the fully connected layers serve as the network's decision-making center, leveraging all extracted features to classify content. During training, these layers learned to distinguish between persuasive phishing samples (like nearperfect Netflix clones in class 12) and legitimate sites in the 'other' category by analyzing comprehensive feature combinations. The model's dual classification capability was demonstrated when it correctly identified both specific brand impersonations (14 phishing classes) and provided binary judgments on phishing versus legitimate content, achieving 86.44% accuracy on test samples, including challenging cases such as professional-looking Bank of America phishing pages (class 5) that had previously deceived human reviewers. Figure 6. Confusion matrix of the resultFigure 7. Summary graph of the performance of the model
8 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025When the model attempted to identify which,specific brand was being copied (such as fake Amazon or Adobe sites), it performed well most of the time. As shown in Figure 7, it correctly identified 18 fake Adobe sites. When just asked the model \"Is this website fake or real?\" (instead of naming the brand), it got it right 86.44% of the time. Figure 7 shows that most fake and real sites were successfully separated, indicating that the model could easily distinguish between them.Here are the reasons why these matters. High recall, it catches almost all fake sites (good for safety!), 9.3% false alarms = It sometimes mistakes real sites for fake ones (bothering, but safer than missing fakes), F1 score (86.5%) = It is both precise (not guessing randomly) and thorough (not missing many fakes).ConclusionThis study demonstrated the effectiveness of text and image analysis with machine learning for robust phishing detection. The BERT model achieved exceptional accuracy (99.14%) in identifying spam through NLP, while the CNN model obtained 86.44% accuracy in detecting visual inconsistency in phishing websites. These models address the dual challenges of textual and visual phishing tactics, offering a layered defense mechanism. The BERT model excels in parsing suspicious language patterns in emails and messages, while the CNN enhances security by scrutinizing authentic websites. They served as complementary defenses: BERT as a high-precision text filter and CNN as a visual verifier, creating a robust multilayered security approach.Future WorkFuture research could examine the integration of real-time detection capabilities and expand the dataset to include emerging phishing tactics across various platforms, such as mobile apps and social media. Improving the CNN model to identify highly sophisticated visual mimics better and reduce false positives would enhance accuracy. Additionally, combining these models with behavioral analysis and/ or user interaction patterns could create a more comprehensive anti-phishing system. Exploring lightweight versions of these models for edge devices could also extend their practical use in cybersecurity.AcknowledgmentThis research was made possible through the FIESTA grant, a collaboration between UCNJ Union College of Union County, NJ, and Felician University. We appreciate the continuous support from the STEM Division. The authors thank Katelyn Snyder and Curran Murphy from ALC for their review of the article.Contact InformationCedric Desinor [email protected] Singh [email protected] Chow [email protected] Fenelon rudolphalan-todley.f@ow l.ucc.eduMadhusudhana Reddy GR madhusudhanar. [email protected], F. C., Bozkir, A. S., & Aydos, M. (2018). Phish-IRIS: A new approach for vision-based brand font prediction of phishing web pages via compact visual descriptors. In 2018 2nd International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT)(pp.1–4). https://doi.org/10. 1109/ISMSIT.2018.8567299Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) (pp. 4171–4186). https://arxiv.org/abs/1810.04805 Patat, R. (2023, April 21). Phishing detection using a transformer [Source code]. https://github.com/rajupatat17/Phising-Detection-using-Transformer/
9 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Smith, L. N. (2015). Cyclical learning rates for training neural networks. https://arxiv.org/abs /1506.01186Smith, N. (2022). Phishing detection with BERT [Source code]. https://github.com/EstAlvB/Phishing-Detection-with-BERT
10 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Unmasking Reality: Distinguishing Genuine Content from AI-Generated FacadesAngello Bravo, Christian Chavez, Astin David, Mohammed HalaiMentor: Academic Specialist Krishna Kumaari JanakiramanSTEM Division, UCNJ Union College of Union County, NJAbstract - Over the past decade, Artificial Intelligence(AI) has advanced to a point where it can generate various types of media that closely resemble human-created content. This study examined how effectively individuals can recognize AI-generated media, especially when it appears indistinguishable from genuine content. We developed a test using an online form builder to evaluate participants' ability to distinguish between AI-generated and human-created media, including audio, images, video clips, and written text. A sample of 100 students from diverse majors at UCNJ took part in the survey. By analyzing their responses, we aimed to identify patterns in AI recognition, such as which types of media are more difficult to differentiate. We also examined the influence of academic discipline, age, and daily interactions with AI on recognition skills. Ultimately, this research seeks to assess the role of artificial intelligence in contemporary digital society and its implications for media consumption. Introduction In recent years, the rapid advancement of Artificial Intelligence (AI) has led to the creation of content that is increasingly indistinguishable from human-generated material. From hyper-realistic images and human-like voices to natural-sounding text and functioning code, AI now shows a remarkable ability to mimic human output across different formats (Frank et al., 2023). As this technology becomes more accessible and widespread, the challenge of telling apart AI-generated and human-created content raises important questions about authenticity, trust, and digital literacy. This research article examines the issues by investigating individuals' ability to recognize and differentiate between AI-generated and human-produced media, highlighting the implications for today’s digital society and the way we consume media.Similar to recent research (Waltzer et al., 2024), this study examined people's ability to recognize AI-generated content by presenting student participants at UCNJ with various media types, including images, text, audio, videos, and code, and asking them to determine whether a human or an AI created each. By analyzing their responses, the team aimed to find patterns in human perception, identify where AI-generated content is most and least convincing, and gain insights into the public’s awareness of AI’s capabilities.Materials and Methodology The study employed Tally, an online form builder, to design a research test with an estimated duration of 10-15 minutes. The test was structured into four distinct sections: audio, image, video, and text. ParticipantsInitially, participants provided demographic information, including their age, major, and the level of their daily engagement with AI.Test StructureEach section of the test featured three questions. Within each question, participants encountered one piece of AI-generated media alongside different forms of real media, such as podcasts, news headlines, and profile photos. The primary objective for participants was to identify which option they believed to be AI-generated. After completing each section, participants rated their confidence in their answers on a scale from 1 to 5.
11 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025IT-related Code SectionAn additional component of the test was directed towards students majoring in IT-related fields, specifically Computer Science and Cybersecurity. This section required participants to disclose their years of programming experience and tasked them with identifying two Python code snippets that produced the same output, with one being AIgenerated.Difficulty AssessmentUpon completion of the test, all participants were asked to rate the overall difficulty of identifying AI on a scale from 1 to 5. They were also invited to indicate which section they found the most challenging.Participant FeedbackDuring and after the test, some participants voluntarily provided insights regarding their thought processes, rationale, and opinions while answering the questions. Additionally, several participants chose to offer informal verbal feedback about their overall experience with the test, further enriching the qualitative data collected.Results and DiscussionI. DemographicsTo visualize and gain a better understanding of the sample demographics, each of the 100 participants in the experiment first entered their age, major, and level of daily engagement with AI before answering any questions.Out of 100 students, a 59% majority reported being non-IT-related majors. These fields varied from Engineering, Psychology, Sports Management, Business, Liberal Arts, Biology, and more. The other 41 students were majoring in IT-related fields, including Computer Science, Cybersecurity, and Data Science. A clear majority of 72 students were between the ages of 18 to 21, although there was still a small number of students beyond that, including a handful of outliers in their 30s, 40s, 50s, and even 60s. Additionally, an absolute majority of participants reported being at least moderately engaged or more with AI tools daily, with only 13% rating their level of engagement as 1 or 2, the rest being fairly distributed between 3, 4, or 5 (Figure 1). Figure 1. Distribution of participants’ daily engagement with AIUltimately, the research team aimed to represent the diverse student body at UCNJ by engaging with students from various backgrounds across campus. The goal was to include individuals enrolled in both unrelated courses and those related to technology and AI to analyze later whether exposure influenced performance. Unsurprisingly, most students, aged 18-21, reflected the traditional college path of enrolling after high school graduation. However, the sample also included the minority of students who returned to school later in life. The number of students of various ages reporting high daily engagement with AI is significant, especially since this trend was consistent across different majors. Many participants shared personal stories while answering, often mentioning that they used ChatGPT to write essays, solve math problems, or obtain direct answers for multiple-choice questions.II. Audio SectionEach of the three questions for the audio section included two recordings under a minute long, one of which was AI-generated and the other real. Participants had to decide which one they believed was generated by AI. The first question compared a narration in Spanish with a voicemail in English
12 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025listing grocery items, the second contrasted a podcast introduction with a live talk from a math teacher, and the third compared a TED Talk speaker with a narrator. Although the audio samples in each question differ—either in language or context—the goal was to see whether familiarity with the language or setting, rather than the content, affects participants’ ability to distinguish between AI and human-produced audio. Given the increasing use of AI-generated recordings, testing various types of recorded content can help identify potential biases in recognizing AI-generated content.The first question’s results showed that only a slim 54% majority could identify which of the two was fake, which was the Spanish audio. The results of the second question were more skewed, with only 23% correctly identifying the live talk as AI. Like the second question, the results of the third question were also skewed, but in the opposite direction, with 70% correctly guessing the narrator’s voice as AI (Figure 2).Figure 2. Results for all Audio Section questionsThe results from each of these three questions, as shown in Figure 2, reflect three different types of outcomes also evident throughout the results of the other test sections. The first outcome is evident in the results of the first question, where there was a relatively even split, with slightly more than half of the respondents choosing correctly and slightly less than half choosing incorrectly. The second and third outcomes, reflected by the results of the second and third questions, show skewed results in both directions, with a vast majority of participants either able or unable to identify the AI recording. This may be attributed to specific characteristics of the recordings themselves. For Question 1, participants found elements of each one’s tone, pronunciation, pace, and flow to be suspicious, making the choice less clear and the results evenly split. In contrast, for Questions 2 and 3, one recording appeared significantly more suspicious than the other—regardless of whether it was generated by AI or not—causing most participants to select that option and resulting in a skewed outcome.III. Image SectionAll three questions in the image section displayed two photos of different people, and participants were asked to identify which of the two photos was generated by AI.Figure 3. AI-generated image for Image SectionEach result for the image section questions was skewed in one way or another. For the first question, 75% correctly guessed the AI image, and 68% guessed correctly for the third question. However, for the second question, only 26% were able
13 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025to identify which of the two images was AI (Figure 4).Figure 4. Results for Image Section Questions 1 and 2Out of all the other sections, the results for the images were the most skewed, with each question having a clear majority of at least 70% who either guessed correctly or incorrectly. To differentiate the two images as much as possible, some participants notably tried to point out exaggerated details in the background, color, photo resolution, facial expressions, and facial texture, perhaps overthinking their answers, leading to mixed results in the end. This pattern was actual across majors as no level of familiarity with AI or exposure to more technology-related courses made any meaningful difference in the results.IV. Video SectionEach question on the video section displayed two different animated/moving images of the same thing: a tiger, an eye, and a tarsier. For the results of this section, 58% guessed correctly for the tiger question, 60% guessed correctly for the eye question, but only 41% correctly identified the AI-generated tarsier for that question (Figure 5). Similarly to the image section, each result for this section displayed a decently vast majority choosing either right or wrong, albeit not as extreme as the results for the images. Participants, like those in the image section, attempted to differentiate between the two videos by examining details in the background, color, resolution, and texture, as well as the movement of each figure.Figure 5. Results for Video Section Questions 1 and 2Text SectionThe first question of the text section was a comparison of quotes from four different historical figures. These quotes included three real ones from Virginia Wolff, Oscar Wilde, and Frida Kahlo, alongside a fake one disguised as Edgar Allen Poe: “The sorrow we carry is the true weight of the heart, and it is only through its release that we may find peace.” The second question followed the same format, but instead of quotes from historical figures, it featured political news headlines from various media corporations. These included The New York Times, CNN, NPR, and a fake one from The Washington Post, “Biden urges action on climate change as Congress debates major reform” – published on February 25, 2024. The third and final question was a comparison of only two passages: an AI-generated paragraph, “The sky was a dull shade of gray,
14 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025thick with the promise of rain, yet the streets remained dry. In the distance, the faint hum of an old train echoed, reminding Nora of forgotten promises. She had not planned to return to this place. However, now that she was here, everything felt impossibly familiar – like a dream half-remembered, teetering on the edge of memory,” and a small excerpt from Jane Austen’s Pride and Prejudice.As shown in Figure 5, the results for each question were fairly divided among all options. For Question 1, only 29% identified the Poe quote as AI, while 71% incorrectly chose one of the other three options. Question 2 saw another majority guess go wrong, with the correct option being the least chosen (18 students). A slight majority of 58% correctly identified the AI-generated passage.Figure 6. Results for Text Section Questions 1 and 2Remarkably, more than half of the participants chose the text section as the most challenging medium for AI to identify out of all the other sections. As they answered the questions, most participants lacked a basis to differentiate between the options. For Question 1, in a few cases, participants notably tried to base their answers on previously held assumptions they had of any of the four historical figures. One participant, anecdotally, who correctly identified Poe’s “quote,” justified her answer by saying how Poe is generally known for saying much darker things and affirmed that he would never say something as light as the option presented to her. For Question 2, as in Question 1, participants attempted to justify their answer by drawing on assumptions they held about media and politics, trying to determine whether the headline's wording and bias matched their knowledge and beliefs about the news corporation. Most participants correctly guessed the AI passage; a popular justification for their answer was the single use of a dash in the paragraph, which has recently become a common indication of AI-generated writing (Wang et al., 2022).V. Code SectionForty-one out of the 100 students were majoring in IT-related fields and were thus eligible for the code section of the test. This section only asked for the number of years of programming experience and to differentiate between two functions for a Fibonacci sequence (Figure 7).Figure 7. Code Section functions (bottom one is AIgenerated)Most of the forty-one IT major participants reported having only 2 or fewer years of programming experience, primarily because they were beginners when they started their majors. Only 14%
15 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025had three or more years of experience, while one individual had 10 years of experience. 56% of the 41 participants correctly identified the recursive function shown in Figure 7 as AI, continuing the trend of randomness observed throughout the test results. The purpose of this section was to determine whether students with programming experience could identify AI better than non-IT majors. However, the results show that these students did not differ in their recognition patterns. Difficulty and randomness were common, regardless of whether students had one or ten years of programming experience ( Ren et al., 2022).ConclusionThe study examined UCNJ students' ability to distinguish between AI-generated content and human-created content across different media types. Results indicated inconsistencies, particularly in the text, suggesting that AI-generated content is becoming increasingly sophisticated. Participants seemed to rely more on their prior knowledge than on recognizable features of the media, contributing to these inconsistencies. The findings underscore the importance of enhanced education and critical evaluation of AI-generated content to address the challenges of identifying misleading media.Future WorkFuture studies could broaden the participant pool to include individuals from other schools and non-collegiate settings for diversity. Additionally, exploring participants' recognition of AI-generated content in everyday contexts, like social media, and investigating the criteria they use in their decisionmaking could provide valuable insights into the characteristics that affect the identifiability of such content.AcknowledgementThis research was made possible through the FIESTA grant, a collaboration between UCNJ Union College of Union County, NJ, and Felician University. We appreciate the continuous support from the STEM Division. The authors thank Katelyn Snyder and Curran Murphy from ALC for their review of the article. We also acknowledge the use ofMicrosoft Editor for grammar and syntax enhancement while preparing this manuscript.Contact InformationKrishna Kumaari Janakiraman: [email protected] Bravo: [email protected] Chavez: [email protected] David: [email protected] Halai: [email protected], J., Herbert, F., Ricker, J., Cispa, L., Eisenhofer, T., Fischer, A., Dürmuth, M., & Cispa, T. (2023). A representative study on human detection of artificially generated media across countries. Cryptography and Security, 1. https://arxiv.org/ pdf/2312.05976Ren, D., Tagg, A. J., Wilcox, H., Roland, D. (2024). Identification of human-generated vs AI-generated research abstracts by health care professionals.JAMA Pediatrics. 178(6), 625–626. doi:10.1001/Jama pediatrics. 2024.0760.Waltzer, T., Pilegard, C. & Heyman, G. D. (2024). Can you spot the bot? Identifying AI-generated writing in college essays. International Journal for Educational Integrity 20(11), 1–18. https://doi.org/10.1007/s40979-024-00158-3Wang, L., Zhou, L., Yang, W., & Yu, R. (2022). Deepfakes: A new threat to image fabrication in scientific publications? Patterns, 3(5), 100509. https://doi.org/10.1016/j.patter.2022.100509
16 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Efficacy of Herbal and Spice Extracts at Different Concentrations Against Gram-positiveand Gram-negative BacteriaJackie Morocho, Wlla AlomariMentor: Academic Specialist Dr. Mirza M. Baig STEM Division, UCNJ Union College of Union County, NJAbstract: The rise of antibiotic resistance necessitates the development of alternative antimicrobial agents. This study examined the antibacterial activity of herbal and spice extracts, including garlic, ginger, turmeric, cinnamon, clove, mustard, pepper, coriander, cumin, bay leaf, oregano, sage, rosemary, thyme, black pepper, and fenugreek, against both Gram-positive and Gramnegative bacteria such as Staphylococcus saprophyticus, Citrobacter freundii, Klebsiella aerogenes, Enterococcus durans, Staphylococcus capitis, Staphylococcus epidermidis, and Escherichia coli. Using the diffusion method, the results indicated that all extracts exhibited antibacterial activity, with garlic and cinnamon displaying the most substantial effect against Grampositive bacteria. Ginger and turmeric showed moderate activity, while cloves were particularly effective against Gram-positive bacteria. These findings highlighted the potential of herbal extracts as natural antimicrobial agents, underscoring the need for further research to isolate active compounds and investigate their clinical applications.IntroductionThe rising prevalence of antibiotic-resistant bacterial strains is a global health emergency, driving the search for alternative antimicrobial options. Antibiotic resistance in bacteria such as Staphylococcus aureus, Enterococcus faecalis, Enterococcus faecium, Citrobacter freundii, Escherichia coli, Staphylococcus saprophyticus, Staphylococcus epidermidis, and Klebsiella pneumoniae has dramatically diminished the effectiveness of standard treatments. This resistance has resulted in higher rates of illness and mortality (Almeida et al., 2021). The situation highlights the urgent need for innovative strategies to combat bacterial infections, particularly those caused by multidrug-resistant strains. Herbal and spice extracts, which have been used in traditional medicine for a long time, hold promise as natural alternatives to synthetic antibiotics due totheir antibacterial properties (Lamy, 2011). Numerous studies have documented the antibacterial activity of these plant-based compounds. Many herbs and spices contain bioactive molecules such as alkaloids, flavonoids, and essential oils, which target bacterial cell walls, disrupt metabolic processes, and inhibit protein synthesis (Ahmed & Wang, 2021; Almeida et al., 2021). Although individual herbs and spices are known for their antibacterial effects, there is still limited research systematically comparing their effectiveness against a broad range of bacterial strains, especially those that are Gram-positive. Previous studies have shown the antibacterial potential of herbs like garlic and ginger, which have demonstrated activity against pathogens such as S. aureus and E. coli (Suman et al., 2020; Tan & Vanitha, 2017). Research addresses the growing problem of antibiotic resistance and explores the potential contribution to developing new antimicrobial therapies (Smith et al., 2020). By identifying the most effective plant-based extracts, we can gain valuable insights into alternative treatments. This could lead to the incorporation of these natural compounds into modern medicine (Jones & Patel, 2019; Williams & Lee, 2019).
17 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Despite the existing research on antibacterial properties, there have been relatively few studies comparing the effectiveness of a broader range of prepared herbal and spice extracts at various concentrations. The objective of this study is to explore the effectiveness of these spices at different serial dilutions. Also, to determine the dilution at which bacteria are most responsive to their inhibitory effect and at which they remain unaffected. This aspect of study design was not addressed in previous research. Additionally, the study will evaluate the effectiveness of spices against several bacterial strains, including E. coli and S. epidermidis. By testing these extracts at different concentrations, the study aims to determine the most effective herbal and spice extracts for fighting both Gram-positive and Gram-negative bacterial infections. Materials and MethodologyA. Collection of spicesThe spices garlic, ginger, turmeric, cinnamon, cloves, mustard, peppers, coriander, cumin, bay leaves, oregano, sage, rosemary, thyme, black pepper, and fenugreek were obtained in powdered form from a local market.B. Preparation of Spices ConcentrationSterile water was used as a solvent to soak the discs in a sterile environment, serving as the negative control. Organic solvents, such as ethanol or acetone, were not used because it was assumed these solvents might alter the concentrations and potentially increase the inhibitory response from bacteria. Keeping the discs immersed in a sterile environment was an ideal choice for accurately assessing the actual inhibitory effect. A total of 16 spices were measured, with 1 gram of each used to prepare a 10% stock solution. Dilutions were then made at concentrations of 1:1, 1:10, 1:100, and 1:1000 for each spice. Sterile discs (6 mm in diameter) were placed in each dilution and allowed to absorb completely. C. Preparation for InoculationThe Tryptic Soy Agar (TSA) plates were inoculated with the bacterial culture and allowed to dry for a few minutes. Discs containing the spice extract were then placed in each of the designated triangles to indicate the concentration levels. The plates were subsequently incubated for 24 to 48 hours at 37°C. A minimum incubation period of 48 hours is required for accurate results and proper data collection.Results and DiscussionData Collection and Analysis:A standard laboratory procedure (Bauer et al., 1966) was employed to evaluate the antibacterial activity of spice extracts using the disc diffusion method. A commonly used technique was utilized to assess antimicrobial activity, providing a clear visual demonstration of inhibition zones. This approach ensures the reliability and reproducibility of results by reducing experimental error. Optimal growth conditions were maintained for the bacteria, enabling a precise assessment of the inhibition zones. A quantitative measure of antibacterial potency was obtained, enabling comparison acrossdifferent extracts. Figure 1: The zone of inhibition for S. epidermidis treated with different concentrations of sage (A) and for S. capitis treated with different concentrations of clove (B).The study's results were supported by a systematic evaluation of inhibition zones for each extract-bacterium combination, revealing apparentvariations in antibacterial activity across the seven bacterial strains. Comparing different extract concentrations (10%, 5%, 1:1, 1:10, 1:100, 1:1000) helped identify concentration-dependent trends,
18 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025with some extracts showing consistent antibacterial effects across multiple dilutions. The findings highlighted important differences in efficacy among the spice extracts, with some demonstrating strong inhibitory effects. In contrast, others showed little tono activity, emphasizing the selective antibacterial properties of the tested compounds. This analysis offers valuable insights into the antimicrobial properties of these extracts, supporting their potential applications in future therapeutic or preservative development. The study's results are further validated by the strong antibacterial effects of ginger, cinnamon, fenugreek, and sage, which consistently produced larger inhibition zones compared to other extracts, highlighting their superior antimicrobial capabilities. Ginger exhibited potent activity against C. freundii, whereas cinnamon demonstrated notable effectiveness against E. coli and K. aerogenes, particularly at higher concentrations, suggesting the presence of potent bioactive compounds. Fenugreek and sage also displayed broad-spectrum inhibitory effects, reinforcing their potential as natural antimicrobial agents. These results align with previous studies on spicederived phytochemicals, supporting the idea that certain plant extracts have significant antibacterial properties (Delaquis, 2001). The responses that depend on concentration further confirm the dose-efficacy relationship, highlighting the importance of optimizing extract formulations for maximum antimicrobial effect. Overall, these findings emphasizethe therapeutic potential of these spices and justify further research into their active components and mechanisms of action.The results of this study effectively highlight a significant finding regarding turmeric's consistent antibacterial activity across all bacterial strains and concentrations, distinguishing it from other extracts and indicating its potential as a broad-spectrum antibacterial agent (Cowan, 1999). The analysis also provides insight into the differing susceptibility of Gram-negative and Gram-positive bacteria to spice extracts, attributing this variation to structural differences in their cell walls. The data reveal a notable difference in average inhibition zone sizes between Gram-positive and Gram-negative bacteria (Figure 2). Overall, the data analysis offers a clear and logical interpretation of the findings, providing valuable insights into the antibacterial properties of spice extracts.Figure 2: Average inhibition zone sizes (measured in mm) in Gram-positive vs. Gram-negative bacteria.A key finding of the study was the concentration-dependent activity of the extracts. The extracts at a 10% concentration produced significantly larger inhibition zones compared to those at lower concentrations (1:10, 1:100, 1:1000), indicating a positive relationship between concentration and antibacterial effectiveness, as shown in Figures 3a and 3f. The data effectively present and interpret the study's findings on the concentration-dependent activity of spice extracts, clearly emphasizing the positive relationship between extract concentration and antibacterial effectiveness. The findings are supported by references to specific figures, which enhance credibility. Thetrend remains consistent across most spice-bacteria combinations, indicating a strong effect (Smith et al., 2020). The interpretation of potency is logical, considering that extracts are more effective at lower concentrations and are considered more potent. The study acknowledges variability in results, noting that some extracts showed minimal activity. This adds credibility and lends a logical interpretation to the findings, highlighting areas for detailed quantitative analysis to support further research investigation.
19 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Figure 3a: Inhibitory effect of different home spice extracts on E. coli at various concentrationsFigure 3b: Inhibitory effect of different spice extracts on C. freundii at various concentrationsFigure 3c: Inhibitory effect of different spice extracts on Klebsiella aerogenes at various concentrationsFigure 3d: Inhibitory effect of different spice extracts on S. capitis at various concentrationsFigure 3e: Inhibitory effect of different spice extracts on S. saprophyticus at various concentrationsFigure 3f: Inhibitory effect of different spice extracts on S. epidermidis at various concentrations
20 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025ConclusionThis study investigated the antibacterial effectiveness of 16 herbal and spice extracts against seven bacterial strains, including both Gram-positive and Gram-negative bacteria, using the disc diffusion method to measure inhibition zones. Significant antibacterial activity was observed in spice extracts, such as garlic, clove, and oregano, particularly against Gram-positive bacteria, including S. capitis and S. epidermidis. Conversely, some extracts showed minimal or no effects, suggesting bacterial resistance or a lack of active compounds. This pattern aligned with earlier studies on other herbal extracts (Cowan, 1999; Jones & Patel, 2019; Jahanshir, 2023).Higher concentrations of extracts, such as 10%, produced larger inhibition zones only in the responsive extracts. In comparison, lower concentrations (1:100 and 1:1000) were less effective or inactive, even in the responsive extracts, demonstrating the concentration-dependent nature of antimicrobial activity. This pattern was observed in all extracts (Suman et al., 2020).Future WorkFurther research is needed to optimize extract preparation and explore their efficacy in vivo for potential applications in healthcare or food preservation. Overall, this study highlights the promise of spice extracts as safe and cost-effective antimicrobial agents.AcknowledgmentThis research was made possible through the FIESTA grant, a collaboration between UCNJ Union College of Union County, NJ, and Felician University. We appreciate the continuous support from the STEM Division. The authors thank Katelyn Snyder and Curran Murphy from ALC for their review of the article. Additionally, some of the content in this article was generated with AI tools, and we are grateful to Gamma App for using this technology. Contact InformationMirza Baig: [email protected] Morocho: [email protected] Wlla Alomari: [email protected] ReferencesAhmed, T., & Wang, C.-K. (2021). Black garlic and its bioactive compounds on human health diseases: A review. Molecules (Basel, Switzerland), 26(16), 5028. Almeida, M., et al. (2021). Plant-based antimicrobial agents: Efficacy against antibiotic-resistant bacteria. Phytotherapy Research, 35(7), 3728–3742. Bauer, A. W., W. M. M. Kirby, J. C. Sherris, & M. Turck. (1966). Antibiotic susceptibility testing by a standardized single disk method. Am. J. Clin. Pathol. 36:493-496.Cowan, M. M. (1999). Plant products as antimicrobial agents. Clinical Microbiology Reviews, 12(4), 564–582. https://doi.org/10.1128/CMR.12.4.564Delaquis, P. J., et al. (2001). Antimicrobial activity of individual and mixed fractions of dill, cilantro, coriander, and eucalyptus essential oils. International Journal of Food Microbiology. https://www.sciencedirect.com/science/article/pii/S0168160501007346Jahanshir, M., et al. (2023). Effect of clove mouthwash on the incidence of ventilator-associated pneumonia in intensive care unit patients: A comparative randomized triple-blind clinical trial. Clinical Oral Investigations, 27(7), 3589–3600. Jones, M. G., & Patel, D. B. (2019). Antibacterial properties of plant-based extracts: A systematic review. Journal of Natural Products, 82(3), 591–600. Lamy, E., et al. (2011). Isothiocyanate-containing mustard protects human cells against genotoxins in vitro and in vivo. Mutation Research, 726(2), 146–150.
21 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Smith, P. A., et al. (2020). Antibiotic resistance in gram-positive and gram-negative pathogens: An increasing threat to public health. Journal of Antimicrobial Chemotherapy, 75(8), 2049–2058. Suman, S., et al. (2020). Antibacterial activity of essential oils from ginger, garlic, and turmeric. Journal of Medicinal Plants, 14(1), 24–30. Tan, W. H., & Vanitha, J. (2017). Antimicrobial properties of garlic and ginger extracts against E. coli and S. aureus strains. International Journal of Pharmaceutical Sciences and Research, 8(4), 1337–1342.Williams, T. M., & Lee, H. J. (2019). The role of plant-based antimicrobial compounds in modern therapeutic strategies. Frontiers in Pharmacology, 10, 253.
22 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Discrepancy Between Cost and Effectiveness of Cleaning Agents: Role of Active IngredientsMansura Abubakar, Jenna Milfort, Emmanuel Alexander MartinezMentor: Academic Specialist Dr. Olubisi T. Ashiru STEM Division, UCNJ Union College of Union County, NJAbstract - This study examined the effectiveness of six common all-purpose household cleaning agents used across the United States in eliminating harmful bacteria, specifically Enterococcus durans, Staphylococcus capitis, Citrobacter freundii, and Escherichia coli.Evaluating the antibacterial properties of cleaning agents is crucial due to their vital role in maintaining hygiene and preventing the spread of disease. This research used the Kirby-Bauer disk diffusion method for assessment. Results were measured by the zones of inhibition around disks infused with different cleaning agents on each bacterial lawn. Additionally, the purchase prices and active ingredients of all-purpose household cleaning agents were analyzed to see if there was a relationship between antibacterial effectiveness and cost. Our findings showed that all-purpose cleaning agents containing hydrogen peroxide as the active ingredient produced the largest zones of inhibition across all four bacterial lawns tested. Interestingly, it was also one of the most affordable among the six cleaning agents analyzed.IntroductionCleaning agents are crucial in every American home. These products are commonly used to maintain cleanliness and hygiene in households, workplaces, and public spaces. They play an important role in daily routines by helping to keep surfaces clean and free from dirt, debris, and, most importantly, harmful microorganisms (American Cleaning Institute, 2022). Cleaning supplies, including soaps, degreasers, and disinfectants, are essential for preventing the spread of microscopic pathogens such as Clostridium difficile and Escherichia coli. These bacteria are often found on food, surfaces, and in restrooms, and they can pose serious health risks if not properly removed.According to Statista, Americans use many cleaning products each year. A study by Statista projected that the U.S. market for these cleaning products will reach about $41.15 billion by 2025 (Statista, 2023; n.d.). This estimate highlights their economic significance. It also underscores the importance of cleaning products in daily life. Consistent and proper disinfection practices effectively protect individual users. These steps also support the broader goal of maintaining healthy environmental hygiene and public health (American Cleaning Institute, 2022).The widespread use of cleaning products underscores the importance of understanding how effective these agents are in truly protecting public health. An improved formulation of active ingredients, including hydrogen peroxide, bleach, ammonia, and alcohol, can effectively target and neutralize harmful microbes, thereby helping to reduce the spread of disease. Recent studies have shown that hydrogen peroxide and sodium hypochlorite are effective disinfectants against various pathogens. Disinfectants such as hydrogen peroxide and chlorine have demonstrated a 4-log10 reduction in microbial populations on various surfaces (Tao et al., 2024). This reduction corresponds to a 10,000-fold decrease in viable microorganisms, representing a 99.9% reduction of microbes.This research aimed to identify which all-purpose household cleaning product offers the most effective bacterial removal. This will support datadriven, informed decision-making. The six all-purpose household cleaning brands analyzed are Clorox, Lysol, Mrs. Meyer’s, Bona, Method, and Seventh Generation. Their antibacterial effectiveness was measured quantitatively against Grampositive and Gram-negative bacteria. Additionally, we examined whether a correlation exists between the antimicrobial effectiveness and the purchase price of these six cleaning products.
23 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Materials and Methodology A total of four bacterial species were tested to assess the antibacterial effectiveness of six all-purpose cleaning agents. The two Gram-positive bacteria were Enterococcus durans and Staphylococcus capitis, while the other two were Gram-negative bacteria, Citrobacter freundii and Escherichia coli. The quadrant streak plate method was used to isolate single colonies from stock cultures of all four bacterial species. Each colony was aseptically transferred to tryptic soy broth (TSB) to grow pure cultures under controlled conditions. These pure cultures were then used to prepare separate bacterial lawns on large Mueller-Hinton agar plates (150 x 15 mm) for further testing (Bauer et al., 1966).The Kirby-Bauer disk diffusion method was used to assess the antibacterial effectiveness of six all-purpose cleaning agents. Briefly, sterile 6 mm discs were individually soaked with each cleaning product. The soaked discs were then aseptically placed onto the prepared bacterial culture lawns. Each bacterial lawn plate was divided into three sections, with one infused disc placed in each. The plates were then incubated at 37°C for 24 hours. After incubation, the zones of inhibition (clear areas with no visible bacterial growth) were measured in millimeters to assess the antibacterial effectiveness of each all-purpose household cleaning agent. All experiments were performed in duplicate and repeated once to ensure consistency and reproducibility.The purchase prices of the six household cleaning agents were collected from three retail stores across different regions of New Jersey. Two of the stores belonged to the same retail chain. By analyzing these prices, we can gain a better understanding ofconsumer options and market trends in household cleaning products.Figures 1A- D: S. capitis (A & B) and E. coli (C & D) lawns showing zones of inhibition around infused discsResults and DiscussionThe discs infused with the all-purpose cleaning agent from the Seventh Generation brand consistently produced the largest zones of inhibition against all four bacterial species. This indicates that it was the most effective cleaning agent in terms of overall antibacterial activity (Figures 1B, 1D, and Table 1). Mrs. Meyer’s brand-infused discs created the smallest zones of inhibition on the lawns of E. durans, C. freundii, and E. coli (Figure 1C & Table 1). Meanwhile, the Method brand-infused discs produced the smallest zones of inhibition on the lawn of S. capitis (Figure 1B & Table 1). Table 1: Average zones of inhibition in mmThe zones of inhibition on Gram-positive bacterial lawns were larger than those on Gram-negative lawns. This indicates that Gram-positive species are more susceptible to the cleaning agents than Cleaning Agents E. durans S. capitis C. freundii E. coliLysol 19 25 6.5 8.5Clorox 20 25 8 9Mrs. Meyer's 8 19.5 6.05 6Bona 16 35 16.5 15Method 13 13 9.25 7.57th Generation 33 47.5 32 20A BC D
24 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Gram-negative species. This difference might result from variations in cell wall composition and structural features between the two bacterial groups. Although the Lysol and Clorox brands had the same active ingredient, their antibacterial effectiveness varied against three of the four bacterial species tested (Tables 1 and 2). The only exception was on the S. capitis lawn, where both produced identical zones of inhibition (Figure 1A and Table 1). In contrast, the Bona and 7th Generation brands, which also contain the same active ingredient, showed different levels of antibacterial effectiveness against all four bacterial species (Tables 1 and 2).Table 2: Cleaning agent prices & active ingredientThe purchase prices of different cleaning agent brands varied across the three stores, except for Mrs. Meyer’s and Bona. The price for Mrs. Meyer’s brand remained the same in all three stores, while the Bona brand was only available in two stores (Table 2). Although the purchase price does not directly determine antimicrobial effectiveness, the chemical makeup of each cleaning agent is crucial to its performance in cleaning. Alkyl-dimethyl-benzyl-ammonium-chloride, the active component in both Lysol and Clorox products, is a quaternary ammonium compound (QAC) with broad-spectrum antimicrobial properties. However, differences in their concentration and the length of the carbon chains in the alkyl groups between the two brands may account for variations in their antibacterial results.Mrs. Meyer’s brand did not specify its active ingredient. It stated that it contained essential oils, which may have antibacterial effects through naturally occurring bioactive compounds such as terpenes and phenolics. In contrast, the other two cleaning agents – Bona and 7th Generation – both used hydrogen peroxide as their primary active ingredient. These two products demonstrated the highest antibacterial effectiveness among the six cleaning agents (Figures 1B and 1D; Tables 1 and 2). Incubation time is another factor that affects the results of antimicrobial assays. This study used a 24-hour incubation period to ensure consistent development of inhibition zones. Shorter incubation times, such as 4 hours, have been reported to produce preliminary results. However, such shortened durations may produce results that differ significantly from those observed after 24 hours (Pankey & Sabath, 2004).ConclusionThis study found that among six all-purpose household cleaning agents, the Seventh Generation brand was the most effective against four bacterial species, while Method and Mrs. Meyer's were the least effective. Staphylococcus capitis was the most susceptible bacterium, with Escherichia coli and Citrobacter freundii being the most resistant, depending on the active ingredient.Active ingredients, such as hydrogen peroxide, significantly enhanced antibacterial effectiveness, whereas products without identifiable active ingredients performed poorly. Moreover, higher cost did not correlate with better performance. These findings underscore the importance of reading product labels and selecting cleaning agents based on scientific evidence, rather than relying on branding or price.Future WorkFuture work for this experiment could involve expanding the range of household cleaning agents and bacterial species tested. This would help identify additional effective compounds, aside from hydrogen peroxide, that have strong antibacterial
25 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025properties. Further studies could also examine hydrogen peroxide's effectiveness against a broader variety of bacteria, including species like C. freundii and E. coli to evaluate alternative active ingredients.LimitationsA broader survey of retail outlets is necessary to obtain a more comprehensive comparison of household cleaning agent prices. Additionally, this study did not evaluate the safety profiles of these cleaning agents on different skin types. As a result, the most effective cleaning agent might not be the best option for sensitive or varied skin conditions. This information does not constitute an endorsement.AcknowledgmentThis research was made possible through the FIESTA grant, a collaboration between UCNJ Union College of Union County, NJ, and Felician University. We thank Miss. Omano Akenami, Ms. Karen Ryan, Mrs. Tamiko Carman, and UCNJ STEM Division, for their continuous support. The authors thank Katelyn Snyder and Curran Murphy from ALC for their review of the article.The authors would like to acknowledge the use of OpenAI’s ChatGPT for language editing and assistance in improving the clarity of this article.Contact InformationDr. Olubisi Ashiru: [email protected] Abubakar: [email protected] Milfort: [email protected] Martinez: [email protected] Cleaning Institute. (2022). Cleaning product industry overall economic impact: Nearly $200 billion, 700,000 jobs, says new ACI Report.Retrieved May 11, 2025, from https://www.cleaninginstitute.org/newsroom/2022/cleaning-productindustry-overall-economic-impact-nearly-200-billion-700000-jobs-saysAmerican Cleaning Institute (n.d.). Home Page.Retrieved May 11, 2025, from https://www.cleaninginstitute.org/Bauer, A. W., Kirby, W. M., Sherris, J. C., & Turck, M. (1966). Antibiotic susceptibility testing by a standardized single disk method. American Journal of Clinical Pathology, 45(4), 493–496. https://academic.oup.com/ajcp/article-abstract/45/4_ts/493/4821085Centers for Disease Control and Prevention. (2024). When and how to clean and disinfect a facility. U.S. Department of Health & Human Services. Retrieved May 11, 2025, from https://www.cdc.gov/hygiene/about/when-and-how-to-clean-anddisinfect-a-facility.htmlPankey, G. A., & Sabath, L. D. (2004). Clinical relevance of bacteriostatic versus bactericidal mechanisms of action in the treatment of Gram-positive bacterial infections. Clinical Infectious Diseases, 38(6), 864–870. https://academic.oup.com/cid/article/38/6/864/320723Statista. (n.d.). Household cleaners – United States. Retrieved May 11, 2025, from https://www.statista.com/outlook/cmo/home-laundrycare/household-cleaners/united-statesStatista. (2023). Cleaning products industry in the U.S. - statistics & facts. Retrieved May 11, 2025, from https://www.statista.com/topics/127 7/cleaning-products-industry-in-the-us/Tao, C., Tang, X., Gan, Y., Qin, Y., Yang, S., & Huang, F. (2024). Investigation of the disinfection efficiency of commercial hydrogen peroxide, chlorine dioxide, and chlorine disinfectant on different surfaces. American Journal of Veterinary Research, 85(8), ajvr . 24.03.0079. https://doi.org/10.2460/ajv r.24.03.0079
26 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Heat-Induced Gene Expression in Tomato Plants:A Molecular Response to Climate ChangeWlla Alomari, Gabriela Duran, Celeste BlanksMentor: Academic Specialist Sanaz Oghlidos STEM Division, UCNJ Union College of Union County, NJ Abstract - Global warming has significantly affected plant growth and productivity by increasing environmental heat stress. This study investigated how Solanum lycopersicum (tomato) responds to elevated temperatures at the molecular and physiological levels. We examined the expression of stress response genes and physiological indicators, including relative water content, chlorophyll concentration, and antioxidant activity. Plants exposed to heat stress exhibited reduced root length, altered water retention, and a slight decline in chlorophyll content. The antioxidant activity in stressed plants increased over time, indicating the activation of defense mechanisms. Despite an overall increase in total RNA, gel electrophoresis confirmed stable expression levels of housekeeping genes (GAPDH and Actin), suggesting that the higher RNA content was likely due to enhanced transcription of stress response genes. These findings support the conclusion that heat stress triggers complex adaptive responses in tomato plants, providing valuable insights into their resilience under changing climate conditions.Introduction The gradual increase in Earth’s temperature, caused by global climate change, is profoundly affecting ecosystems worldwide, particularly agricultural systems. These changes are disturbing essential environmental conditions needed for optimal crop growth, including soil health, water supply, and overall crop yields (Mittler, 2002). Extreme heat events can cause plants to wilt, hinder photosynthesis, and decrease harvests, which stresses global food security and economic stability. Therefore, understanding the physiological and molecular effects of heat stress on plants has become a crucial area of scientific study.To address these challenges, researchers are focusing on key indicators of how plants respond to increasing temperatures, such as relative water content (RWC), chlorophyll levels, and antioxidant capacity. RWC measures the water status of plant tissues compared to their fully saturated state and serves as a reliable marker for drought tolerance and cell hydration (Weatherley, 1950). Chlorophyll levels are directly linked to photosynthetic efficiency and plant health, often decreasing under stress and resulting in visible signs, like chlorosis (Taiz et al., 2015). Oxidative stress, another consequence of heat exposure, leads to the production of reactive oxygen species (ROS). Antioxidants, such as enzymes like superoxide dismutase and compounds like vitamin C, help neutralize ROS, thereby safeguarding cellular components (Mittler, 2002).At the molecular level, heat stress is known to trigger the production of heat shock proteins (HSPs), especially HSP70, which act as molecular chaperones to help proteins fold correctly and prevent denaturation under stressful conditions (Vierling, 1991). Recent advances in molecular biology techniques, including RNA isolation, cDNA synthesis, and PCR-based gene expression analysis, now allow for the accurate measurement of stressinduced molecular responses. This study examines the effects of prolonged high temperatures on Solanum lycopersicum (tomato) by analyzing changes in RWC, chlorophyll levels, antioxidant capacity, and gene expression. By combining physiological data with molecular findings, this research aims to provide a comprehensive understanding of the plant’s responses to heat stress at multiple levels, offering valuable insights to improve crop resilience in the face of global warming.
27 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Materials and MethodologyI. Plant Growth and Experimental DesignSeeds of Solanum lycopersicum (tomato) were planted in Rockwool mixed with Vermiculite and kept in a controlled growth chamber at 22–25°C for two weeks. The seedlings were subsequently transferred to floating platforms in a nutrient-rich solution for an additional two weeks. Plants were then individually transferred to Erlenmeyer flasks and divided into two groups: a control group and an experimental group. Both groups were maintained at 22–25°C for three weeks. Afterward, the control group stayed at the same temperature for an additional nine days, while the experimental group was exposed to heat stress at 55°C for the same duration. Leaf samples were collected at four time points for physiological, biochemical, and molecular analyses.II. Relative Water Content (RWC) Fully expanded, healthy leaves were harvested and immediately placed in plastic bags to minimizewater loss. The fresh weight (FW) was measured using an analytical balance. The samples were then submerged in distilled water at room temperature in the dark for 24 hours to reach turgid weight (TW). TW was recorded after gently blotting the leaves with paper towels to remove surface moisture. The samples were dried at 70°C for 48 hours to determine the dry weight (DW). Relative water content (RWC) was calculated as described by Barrs and Weatherley (1962) using the following method: RWC (%) =FW−DW/TW−DW×100 III. Chlorophyll Extraction and AnalysisFifty milligrams of each collected leaf sample were weighed and ground in a mortar with a pinch of sand and 3 mL of acetone until a homogeneous mixture was formed. The mixture was then filtered through cheesecloth into a test tube to remove debris. An additional 4 mL of acetone was added to each tube, increasing the total volume to 7 mL. The absorbance was measured at 630 nm, 645 nm, 665 nm, and 750 nm using a spectrophotometer. Chlorophyll concentration was calculated using the following formula:Concentration = [(11.6 X A665 nm) - (0.14 x A630 nm) - (1.3 x A645 nm)].IV. Preparation of Lysate for Antioxidant Assay Approximately 100 mg of frozen leaf tissue per sample was ground into a fine powder using a mortar and pestle. To each sample, 1 mL of cold phosphate-buffered saline (PBS) was added per 100 mg of tissue, and the mixture was homogenized by pipetting and vortexing. The homogenate was centrifuged at 13,000 × g for 15 minutes at 4°C. The supernatant, containing antioxidants, was collected and stored at –80°C until further analysis. Total antioxidant capacity was measured using the SigmaAldrich Antioxidant Assay Kit, following the manufacturer's instructions.V. RNA Extraction and cDNA SynthesisTotal RNA was extracted from frozen plant tissues using the Mag-MAX Plant RNA Isolation Kit (Thermo Fisher Scientific), following the manufacturer’s protocol. To eliminate potential genomic DNA contamination, DNase I treatment was performed. RNA concentration was measured using a Nanodrop, with acceptable A260/A280 ratios close to 2.0.For cDNA synthesis, 2 µL of purified RNA was mixed with oligo (dT) primers and reverse transcriptase enzyme using the Thermo Fisher HighCapacity cDNA Reverse Transcription Kit. Reverse transcription was carried out according to the manufacturer's instructions, and the resulting cDNA samples were stored at –20°C for future analysis.VI. Gel Electrophoresis for Semi-Quantitative Gene ExpressionGene expression analysis focused on three housekeeping genes: Actin, GAPDH, and HSP70. Conventional PCR was performed using cDNA from both control and heat-stressed plants. PCR conditions included an initial denaturation step, followed by 30 to 35 cycles of denaturation, annealing at 58°C for 30 seconds, and extension steps suitable for the amplicon size. PCR products were separated
28 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025on a 2% agarose gel alongside a 1000 bp DNA ladder. Band intensities representing gene expression levels were compared between control and heatstressed samples. Both groups were expected to display similar band intensities for the housekeeping genes, while the heat-stressed samples were anticipated to show increased HSP70 expression, indicated by higher band intensity.Results and discussionI. Root Length Heat stress significantly hampers plant growth, particularly affecting the root system. In heatstressed tomato plants, root length decreased and roots became more fragile (Figure 1), unlike the steady development observed in the control plants. These results suggest that higher temperatures inhibit root elongation.Functionally, a shorter root length reduces water and nutrient uptake, weakens plant anchorage, and makes plants more vulnerable to environmental stress. This finding aligns with previous research, which has shown that root architecture is susceptible to heat and is crucial for stress resilience (Calleja-Cabrera et al., 2020). Figure 1: Comparison of root length: control (left) and experimental (right)II.Relative Water ContentInterestingly, despite significant root damage, the experimental group exhibited a slightly higher leaf RWC compared to the control (Figure 2). This may suggest a compensatory mechanism for conserving water under heat stress, possibly through reduced transpiration via stomatal closure (Mansfield & Atkinson, 1990). Elevated RWC could also result from cellular adjustments such as osmotic regulation or enhanced water retention (Blum, 2017).Figure 2: Bar graphs of the relative water content of the control and experimental groupsThese findings suggest that Solanum lycopersicum may employ adaptive strategies to maintain leaf hydration despite root damage, thereby supporting overall stress tolerance (Jongdee et al.,2002).Figure 3: Bar graph of the chlorophyll concentration of the control and experimental groupsIII. Chlorophyll ConcentrationChlorophyll content was similar between the control and heat-stressed groups (Figure 3). This suggests that the duration or intensity of heat exposure may not cause significant chlorophyll degradation, which is often associated with prolonged or more severe stress conditions.Chlorophyll breakdown usually indicates a delayed response to environmental stress. Its stability in this situation may mean that Solanum lycopersicum has early tolerance mechanisms that help preserve photosynthesis during moderate heat stress. These results highlight the complexity of physio-
29 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025logical responses to heat, suggesting that chlorophyll level alone may not always be a reliable early indicator of thermal stress. IV. Antioxidant ActivityAntioxidant capacity was measured at three time points: T0, T2, and T6, using the Sigma-Aldrich Antioxidant Assay Kit to assess the plant’s ability to counteract oxidative stress induced by heat. T0 reflects the baseline before heat exposure; T2 and T6 correspond to 2 and 6 days after the start of heat treatment in the experimental group, respectively. At T0, the control group and the experimental group showed similar antioxidant activity, indicating a similar initial oxidative status (Figure 4). However, after heat treatment, the experimental group showed a 110% increase in antioxidant levels at T2 and a 59% increase at T6 compared to the control.Figure 4: Bar graphs of the antioxidant level of both the control and experimental groupsThese results indicate an evident upregulation of antioxidant defenses in response to heat stress. Elevated temperatures increase the production of reactive oxygen species (ROS), which can harm cellular components if not neutralized. The observed increase in antioxidant activity suggests the activation of protective mechanisms, including both enzymatic and non-enzymatic antioxidants, which mitigate oxidative damage and preserve cellular function (Gill & Tuteja, 2010).V. RNA ExtractionAt the conclusion of the experiment, total RNA was extracted from both control and heatstressed Solanum lycopersicum (tomato) plants to assess their molecular response to prolonged exposure to high temperatures. RNA integrity was confirmed through nanodrop analysis and agarose gel electrophoresis, ensuring the samples were suitable for downstream molecular applications such as cDNA synthesis and PCR.Both groups produced significant amounts of RNA. However, the experimental group showed higher RNA yields compared to the control group (Figure 5). This increase in total RNA concentration in heat-stressed plants may indicate an upregulation of transcriptional activity in response to heatinduced stress, as plants activate various stress-responsive genes under such conditions (Suzuki et al., 2014; Wang et al., 2003).Figure 5: Bar graphs of the total RNA concentration of both the control and experimental groupsUnder heat stress, plants often increase the expression of protective genes, including those encoding heat shock proteins (HSPs), antioxidants, and other molecular chaperones, which are crucial for maintaining protein structure and cellular stability (Mittler, 2002; Vierling, 1991).These molecular adjustments are a crucial component of the plant’s comprehensive strategy to mitigate the detrimental effects of heat and oxidative stress. The differences in RNA quantity observed between the control and experimental groups support the hypothesis that heat stress triggers complex molecular changes at the transcrip-
30 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025tional level. These changes ultimately aim to maintain cellular homeostasis and enhance stress tolerance.VI. cDNA Synthesis, Polymerase Chain Reaction, and Gel ElectrophoresisFollowing RNA extraction, cDNA synthesis was completed and used for PCR amplification of HSP70, ACTIN, and GAPDH using gene-specific primers. Gel electrophoresis revealed clear, consistent bands for the housekeeping genes in both the control and heat-stressed groups (Figure 6), confirming the quality of the cDNA and the successful amplification. However, HSP70 was not detected, possibly due to insufficient heat stress, suboptimal primers, or inadequate PCR conditions. The stable expression of the housekeeping gene indirectly suggests that the increased total RNA in the experimental group may reflect elevated transcriptional activity of other heat-induced genes. While HSP70 gene was undetectable, the results support the activation of other stress-responsive pathways in Solanum lycopersicum. Further analysis is needed to clarify these molecular responses.Figure 6: Gel electrophoresis analysis of multiplex PCR: Lane 1 - DNA Ladder, Lane 2 - Control Plant One, Lane 3 - Experimental Plant One, Lane 4 -Control Plant Two, Lane 5 - Experimental Plant Two, Lane 6 - DNA Ladder.ConclusionThis study demonstrated that Solanum lycopersicum (tomato) plants respond to heat stress with multiple levels of physiological and molecular changes. Although HSP70 expression was not detected via gel electrophoresis, increases in total RNA yield and antioxidant activity suggest heightened transcriptional activity and activation of stress-response pathways. Physiological adjustments, such as improved leaf water retention and a slight reduction in chlorophyll levels, further highlight adaptive responses to elevated temperatures. Additionally, visible changes, such as altered root structure and improved water conservation, support the presence of whole-plant adaptive mechanisms. Collectively, these findings provide a comprehensive understanding of tomato plant responses to heat stress and establish a foundation for future research to enhance crop resilience in the context of climate change.Future WorkFuture research could focus on identifying genes that are upregulated in response to heat stress, especially in plant varieties that naturally resist extreme environmental conditions. A deeper understanding of the regulatory mechanisms governingthese stress-responsive genes will be crucial for developing targeted genetic interventions. This knowledge could form the basis for enhancing heat tolerance in more vulnerable crops through advanced molecular techniques, including gene editing. The genetic resilience of crops will be crucial for sustaining agricultural productivity and ensuring food security in the face of global warming and increasingly unpredictable environmental conditions.Acknowledgment This research was made possible through the FIESTA grant, a collaboration between UCNJ Union College of Union County, NJ, and Felician University. We appreciate the continuous support from the STEM Division. The authors thank Katelyn Snyder and Curran Murphy from ALC for their review of the article—special thanks to Karen Ryan and Tamiko Carman for their unwavering support throughout the project. We also thank Psychogenics for their valuable assistance in primer design and support with the antioxidant assay. The authors
31 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025acknowledge using ChatGPT for grammar and syntax refinement while preparing this manuscript.Contact InformationSanaz Oghlidos: [email protected] Duran: [email protected] Alomari: [email protected] Blanks: [email protected], H. D., & Weatherley, P. E. (1962). A re-examination of the relative turgidity technique for estimating water deficits in leaves. Australian Journal of Biological Sciences, 15(3), 413–428.https://doi.org/10.1071/BI9620413Blum, A. (2017). Osmotic adjustment is a primary mechanism of drought stress adaptation, supporting plant production. Plant, Cell & Environment, 40(1), 4–10. https://doi.org/10.1111 /pce.12800Calleja-Cabrera, J., Boter, M., Oñate-Sánchez, L., & Pernas, M. (2020). Root growth adaptation to climate change in crops. Frontiers in Plant Science, 11, 544. https://doi.org/10.3389/fpls .2020.00544Gill, S. S., & Tuteja, N. (2010). Reactive oxygen species and antioxidant machinery in abiotic stress tolerance in crop plants. Plant Physiology and Biochemistry, 48(12), 909–930. https://doi .org/10.1016/j.plaphy.2010.08.016Jongdee, B., Fukai, S., & Cooper, M. (2002). Leaf water potential and osmotic adjustment as physiological traits to improve drought tolerance in rice. Field Crops Research, 76(2–3), 153–163.https://doi.org/10.1016/S0378-4290(02)00036-7Mansfield, T. J., & Atkinson, C. J. (1990). Stomatal behaviour in water-stressed plants. In R. G. Alscher & J. R. Cumming (Eds.), Stress responses in plants: Adaptation and acclimation mechanisms(pp. 241–264). Wiley-Liss.Mittler, R. (2002). Oxidative stress, antioxidants and stress tolerance. Trends in Plant Science, 7(9), 405–410. https://doi.org/10.1016/S1360-1385(02)02312-9Sigma-Aldrich. (n.d.). Antioxidant assay kit. Retrieved from https://www.sigmaaldrich.comSuzuki, N., Rivero, R. M., Shulaev, V., Blumwald, E., & Mittler, R. (2014). Abiotic stress and plant responses from the whole-plant to the cellular level. Plant Physiology, 164 (2), 450 – 462. https://doi.org/10.1104/pp.113.23142 3Taiz, L., Zeiger, E., Møller, I. M., & Murphy, A. (2015). Plant physiology and development (6thed.). Sinauer Associates.ThermoFisher Scientific. (n.d.-a). High-capacity cDNA reverse transcription kit. Retrieved fromhttps://www.thermofisher.comThermo Fisher Scientific. (n.d.-b). MagMAX™plant RNA isolation kit. Retrieved fromhttps://www.thermofisher.comVierling, E. (1991). The roles of heat shock proteins in plants. Annual Review of Plant Physiology and Plant Molecular Biology, 42(1), 579–620.https://doi.org/10.1146/anurev.pp.42.060191.003051Wang, W., Vinocur, B., & Altman, A. (2003). Plant responses to drought, salinity and extreme temperatures: Towards genetic engineering for stress tolerance. Planta, 218(1), 1–14.https://doi.org/10.1007/s00425-003-1105-5Weatherley, P. E. (1950). Studies in the water relations of the cotton plant. I. The field measurement of water deficits in leaves. New Phytologist, 49(1), 81–97. https://doi.org/10.1111/j.14 698137.1950 .tb051x
32 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Natural Food Preservation Using Household Ingredients: A Chemistry-Based InvestigationDeclan James Cassidy, Emely ChangMentor: Academic Specialist Shahrzad TaghdissiSTEM Division, UCNJ Union College of Union County, NJAbstract: This experiment assessed the chemical effectiveness of natural preservatives—alt (NaCl), vinegar (acetic acid), fresh garlic, oil, superfood blend, lemon juice (citric acid), and powdered citric acid—in delaying spoilage in various ingredients over 7 days at room temperature (37°C). It tested the durability of different food groups when exposed to these substances. Spoilage was evaluated based on changes in color, odor, texture, and microbial growth. The study also explored the role of phenolic compounds and oxidative enzymes in spoilage mechanisms. The results supported the use of natural, high-concentration acids, such as vinegar and citric acid, as well as salt, as effective and low-cost alternatives to synthetic preservatives for short-term food storage. Additionally, the results indicated that garlic and lemon juice were the least effective, exhibiting the most significant microbial growth.Introduction Synthetic preservatives, such as sodium benzoate and potassium sorbate, are commonly used to slow down spoilage; however, growing concerns over health and toxicity have prompted a search for natural, safer alternatives (Gyawali & Ibrahim, 2014). Natural substances, such as sodium chloride (table salt), acetic acid (found in vinegar), and citric acid, possess well-documented antimicrobial and antioxidant properties. Their modes of action include altering osmotic pressure, lowering pH, and inhibiting enzymatic activity. Additionally, understanding the chemistry of phenolic oxidation and microbial growth inhibition provides insight into why certain natural substances are effective in delaying food spoilage. Food spoilage is a biochemical process primarily driven by microbial activity and enzymatic oxidation, particularly in fruits rich in phenolic compounds. Apple slices, for instance, are most known to undergo enzymatic browning due to the oxidation of polyphenols (such as chlorogenic acid and catechins) catalyzed by the enzyme polyphenol oxidase (PPO), resulting in quinones that polymerize to form brown pigments (Shahidi & Zhong, 2010). This study conducts a detailed analysis of the effectiveness of various preservatives from a chemical perspective. It aims to clarify their specific roles in food chemistry and how they contribute to the preservation of food products. By examining the chemical properties and mechanisms of action of these preservatives, the research seeks to provide a deeper understanding of their impact on food safety, shelf life, and overall quality.Materials and Methodology The materials used in the study included tablegrade iodized salt (NaCl), white vinegar (5% acetic acid solution), fresh lemon juice (roughly 5–8% citric acid), powdered citric acid, powdered berry superfood blend, chopped fresh garlic, food-safe oil spray, Buddig sliced deli turkey, Taylor Farms chopped salad mix, canned pineapple chunks, flour tortillas, weighing boats, measuring spoons, Erlenmeyer flasks, a pH meter, food-safe storage wrap, and a daily observation log.
33 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025 Eight groups of ingredients underwent different treatments, including a control group with no treatment, a salt treatment with one teaspoon of table salt, a vinegar treatment with one tablespoon of white vinegar, a lemon juice treatment with one tablespoon of lemon juice, a garlic treatment where ingredients were rubbed and covered with freshly chopped garlic, an oil treatment with one tablespoon of food-safe spray oil, a citric acid treatment with one tablespoon of powdered citric acid, and a berry superfood blend treatment with one tablespoon of powdered berry superfood blend (Fig. 1).All samples were sealed in individual weighing boats and stored at room temperature. Observations were conducted daily for 7 days, monitoring indicators of chemical and microbial spoilage, including enzymatic browning (which indicates PPO activity and phenol oxidation), mold growth (suggesting fungal contamination), textural degradation, and odor development (linked to microbial metabolism) (Fig. 2). After the 7-day trial, when signs of microbial growth or spoilage were apparent, the samples were transferred into Erlenmeyer flasks, each filled with an equal amount of water to create solutions for each group. The pH of each solution was then measured using a pH meter, with a higher pH indicating microbial growth and activity (Fig. 3).Results and DiscussionI. ResultsThe ingredients that proved most successful at the end of this experiment were table salt, vinegar, and citric acid. Table salt (NaCl) proved highly effective, delaying spoilage by drawing out moisture through osmotic dehydration, thereby preventing microbial growth and preserving texture with no visible browning. Vinegar was also effective, leveraging its acidity to minimize browning and inhibit mold growth, though texture retention was only moderate (Fig. 1). Figure 1: Eight treatment groups, each with four replicates, used the same food items: sliced turkey, chopped salad mix, canned pineapple chunks, and tortillas. One group served as the untreated control. Citric acid also proved effective as a preservative, outperforming lemon juice due to its higher acidity. It prevented both browning and microbial growth while maintaining a moderate texture (Fig. 1). Although lemon juice contains antioxidants and has a low pH, it exhibited limitations due to its lower citric acid concentration, which led to browning and mold formation by Day 3.Ingredients such as garlic and the superfood berry blend demonstrated moderate effectiveness in preserving food quality. Garlic is recognized for its antimicrobial properties, primarily due to compounds such as allicin. These compounds successfully delayed mold growth until Day 4; however, garlic was not effective in preventing browning or maintaining the food's texture over that time. On the other hand, the superfood berry blend showed results similar to those of the moderate group. Its lower acidity contributed to browning and texture degradation, making the fruit appear less fresh. Although the berry blend did not lead to visible mold, changes in color and texture indicated it was less effective at preserving the food's overall quality.
34 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025Table 1: Results of the factors tested and observed over 7 days.The control group experienced rapid enzymatic browning by Day 2 and visible mold growth by Day 3. This indicated poor preservation, especially for fruits and vegetables, which deteriorated quickly. While the meat and tortilla avoided bacterial growth, this was primarily due to dehydration rather than practical preservation. Similarly, oil, while capable of creating a barrier to air and moisture, was largely ineffective on its own under the experimental conditions, allowing mold growth and poor texture retention by Day 3 (Fig. 2).Figure 2: Samples were placed in Petri dishes and incubated for bacterial growth.These results were determined by observing and recording visible changes in each group over the first 7 days, including any shifts in structure, signs of microbial growth, or discoloration. After this initial test, the pH of each sample was measured after 7 days to detect any changes in acidity, which could serve as an additional indicator of food-safetyspoilage.DiscussionFood spoilage in this study was mainly caused by enzymatic browning and microbial activity, both of which are influenced by environmental factors such as pH, oxygen levels, and the presence of antimicrobial or antioxidative agents. The primary biochemical process involved is the oxidation of phenolic compounds by polyphenol oxidase (PPO), which leads to the formation of quinones that polymerize into brown melanin pigments. Additionally, microbial growth is often encouraged by neutral pH, available moisture, and the absence of inhibitory agents. The qualities are found in ingredients such as vinegar and salt, which are among the most effective preservatives. Table salt is preserved mainly through osmotic dehydration. By creating a hypertonic environment, salt draws water out of microbial cells, impairing their function and stopping their growth. Although salt does not directly inhibit PPO, its ability to reduce water activity is sufficient to prevent both enzymatic browning and microbial spoilage in most ingredients (Sapers et al., 1989).Vinegar (acetic acid) also proved to be one of the most effective agents due to its low pH, which denatures microbial enzymes and PPO by disrupting their protein structures. The acidic environment also interferes with phenolic substrate oxidation by reducing available oxygen and altering redox potential (Lee et al., 2017). As a result, vinegar-treated samples exhibited minimal browning and no visible microbial growth over the 7 days.Figure 3: Measuring the pH of the samples with the most bacterial growth.Citric acid provided a more controlled and concentrated source of acidification compared to
35 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025lemon juice. As a powerful chelating agent, citric acid also binds transition metals such as Fe²⁺ and Cu²⁺, which can accelerate oxidative reactions. This dual action significantly slowed both enzymatic browning and microbial spoilage, although not as effectively as vinegar, due to differences in penetration and solubility across the sample surfaces. Lemon juice, though also acidic and rich in antioxidants, was less effective than vinegar because of its slightly higher pH and lower citric acid concentration (Fig. 3). While it initially inhibited PPO activity and slowed microbial growth, it was insufficient to prevent the eventual development of browning, mold, and odor, especially in ingredients more prone to spoilage.Conversely, control samples underwent rapid spoilage, particularly evident in fruits and vegetables, due to the absence of inhibitory agents. PPO activity and microbial metabolism were unimpeded, leading to visible browning and fungal growth within 48–72 hours. The meat and tortilla, though dehydrated and showing less visible spoilage, still provided a baseline for comparative degradation. The lack of preservation agents allowed phenolic oxidation and fungal colonization to proceed unchecked. The last set of ingredients, although effective in certain aspects of preservation, lacked a crucial quality necessary for successful results. The berry superfood blend, although containing some antioxidant components, lacked the necessary acidity and chelating strength to prevent spoilage effectively. Its limited ability to suppress PPO or inhibit microbial metabolism led to ongoing textural degradation and partial browning, indicating only a mild preservative effect. Oil treatment functioned as a physical barrier, reducing oxygen exposure and moisture exchange. However, without additional chemical preservative properties or refrigeration, the oil was insufficient to inhibit microbial or enzymatic activity. As a result, significant mold growth and tissue breakdown occurred, demonstrating the limited effectiveness of oil as a standalone preservative under ambient conditions. Lastly, Garlic contains organosulfur compounds, most notably allicin, which exhibit broadspectrum antimicrobial activity by disrupting bacterial cell membranes and metabolic processes. While garlic treatment successfully suppressed mold growth, it lacked sufficient acidity or antioxidative capacity to inhibit PPO, leading to noticeable browning and softening of plant tissues.ConclusionThe experiment demonstrated that acid-based natural preservatives are chemically effective at delaying spoilage in various food items by inhibiting enzymatic oxidation of phenols, suppressing microbial growth, and utilizing sodium’s dehydrating possessions. Citric acid, acetic acid, and NaCl were the most effective at maintaining the integrity of the ingredients compared to the control group. These findings support their use as reliable short-term preservation agents in food chemistry, offering a natural, low-toxicity alternative to synthetic additives. This study also provided insight into commonly used preservatives, such as lemon juice, garlic, a superfood mix, and oil, which were found to be less effective or ineffective at preventing bacterial growth in food. The chemical principles of enzyme inhibition, pH manipulation, and phenol stabilization underscore their importance in food preservation science. These methods offer natural, affordable conservation options that may reduce waste and decrease the use of harsh chemicals in food production.Future WorkIn the future of food preservation, studies like these can pave the way for sustainable and consumer-friendly options to extend the longevity of our food. The use of natural and inexpensive preservatives may one day lead to a decrease in chemical usage and food processing, as well as a reduction in food waste for the average person.AcknowledgmentThis research was made possible through the FIESTA grant, a collaboration between UCNJ Un-
36 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8| No. 1 | Fall 2025ion College of Union County, NJ, and Felician University. We appreciate the continuous support from the STEM Division. The authors thank Katelyn Snyder and Curran Murphy from ALC for reviewing the article.Contact InformationShahrzad Taghdissi: [email protected] Cassidy: [email protected] Chang: [email protected] ReferencesGyawali, R., & Ibrahim, S. A. (2014). Natural products as antimicrobial agents. Food Control, 46, 412–429. https://doi.org/10.1016/j.foodcont.2014. 05.047Lee, C. Y., Whitaker, J. R., & Schwarz, S. (2007). Enzymatic browning and its prevention. In J. Whitaker & D. Voragen (Eds.), Handbook of Food Enzymology (pp. 267–280). CRC Press.Queiroz, C., Lopes, M. L. M., Fialho, E., & Valente-Mesquita, V. L. (2008). Polyphenol oxidase: Characteristics and mechanisms of browning control. Food Reviews International, 24(4), 361–375. https://doi.org/10.1080/87559120802089364Sapers, G. M., Miller, R. L., & Smith, N. L. (1989). Inhibition of enzymatic browning in apple with ascorbic acid derivatives, cysteine, and sodium sulfite. Journal of Food Science, 54(4), 997–1002. https://doi.org/10.1111/j.1365-2621.1989.tb0740 2.xShahidi, F., & Zhong, Y. (2010). Novel antioxidants in food quality preservation and health promotion. European Journal of Lipid Science and Technology, 112(9), 930-940. https://doi.org/ 10.1002/ejlt.20100004
2 UCNJ Union College of Union County, NJ | Undergraduate Research Journal Volume 8 | No. 1 | Fall 2025Union County Board of County CommissionersUCNJ Union College of Union County, NJBOARD OF TRUSTEES BOARD OF GOVERNORSLourdes M. Leon Commissioner – ChairwomanJoseph C. Bodek Commissioner – Vice ChairmanJames E. Baker, Jr., CommissionerMichèle S. Delisfort, CommissionerSergio Granados, CommissionerBette Jane Kowalski, CommissionerAlexander Mirabella, CommissionerKimberly Palmieri-Mouded, CommissionerRebecca L. Williams, Commissioner Victor M. Richel, ChairRafael J. Betancourt, Esq., Vice ChairDr. Margaret M. McMenamin, PresidentBrian Campbell, UCNJ ’95George A. Castro, IIDaniel J. Connolly, CPAMiguel A. FigueredoJeffrey H. Katz, Esq.Miguel A. MerinoDaryl PalmieriEmily L. Root, MA, CCC-SLPKamran Tasharofi, M.D.Katherine Mejía Reyes UCNJ ’24, Student RepresentativeMichael M. Horn, Esq.,Legal CounselJeffrey H. Katz, Esq., ChairF. Jim Della Sala, Vice ChairDr. Margaret M. McMenamin, PresidentMelinda Ayala UCNJ ’11Lawrence D. Bashe Rafael J. Betancourt, Esq.Brian Campbell UCNJ ’95Deborah Enix-RossKathleen A. Fitzpatrick UCNJ ’75Stephen F. Hehl, Esq. UCNJ ’75Donna M. Herran UCNJ ’85Edward J. Hobbie, Esq.Gary S. HoranRichard J. MalcolmCarl J. NaporMatthew R. Nazzaro UCNJ ’04Paul T. O’Neill UCNJ ’84Sandra RiceJohn RichelVictor M. RichelJames Roundtree IIIRobert J. SloanJohn M. TorielloLisa Vecchione UCNJ ’03Allan L. WeisbergHugh C. WelshMary M. Zimmermann UCNJ ’01