AptaBind AI: Using Artificial Intelligence to predict aptamer target binding.

This project stems from a personal mission: helping my dad manage his celiac disease. Living with celiac means even a trace amount of gluten can be harmful, so reliable detection is vital. I focused on aptamers—think of them as "chemical Velcro." They are short DNA sequences shaped to stick to specific targets, like gluten proteins, acting as custom-built sensors for food safety. However, finding the right sequence is usually a slow, expensive guessing game. In applications like food safety and health, this uncertainty can lead to wasted time, cost, and unreliable results.

They are powerful tools for detection, but developing them is difficult. The current process relies on laboratory testing and published data that often varies depending on experimental conditions.

The central question of this project is:

Can patterns in DNA sequences be used to estimate binding strength before laboratory testing?

To explore this, I developed a computational system that analyzes key sequence features—such as length, GC content, and motifs—and uses them to estimate binding potential.

This project was also inspired by the need to make early-stage research more accessible. Many advanced prediction tools require specialized software or large datasets. This system focuses on building a simple, interpretable approach that can support early decision-making.

The goal was to create a predictive framework that helps identify promising aptamer sequences while recognizing the limitations of available data.

This Project was developed by designing a structured system that predicts aptamer-target binding using sequence based features and a bias-aware evaluation approach.

First, published aptamer data was collected from multiple studies available online, including DNA sequences, target type, binding values (Kd), and available experimental conditions. Because this data is often inconsistent across sources, once all the data was gathered, it was clean and standardized, with Kd values converted into comparable units (nM) and incomplete entries clearly identified (1 = present; 0 = absent).

Next, each sequence was analyzed using 3 biologically relevant features:

Sequence length

GC content (percentage of Guanine and Cytosine basis)

Presence of specific motifs

This features were selected based on both observed patterns and findings in the relevant literature.

Using this inputs a prediction system was created to estimate binding strength.

To address variability in experimental conditions, which were affecting the reliability of the predictions, a confidence scoring system was introduced. Each prediction is evaluated based on:

Availability of binding data (Kd)

Consistency of experimental condition (assay type, pH, temperature)

Completeness of metadata

Only then predictions are labeled with the confidence level, allowing results to be interpreted with appropriate caution.

This approach creates a bias-aware prediction pipeline that not only estimates binding, but also communicates how much that estimate can be trusted.

When I initially started researching aptamers I noticed commonly recurring patterns so when I was thinking of new possible project ideas I remembered the patterns and remembered that teachable machines learn through patterns. That's when I decided it was worth looking into.

This project led to the development of a structured system that takes into account that the data is inconsistent and adjusts how much it trusts the results when estimating how strongly an aptamer binds to its target.

The system analyzes aptamer DNA sequences using three main features:

Sequence length

GC content

Motif presence

These features were selected based on known biological relevance and  their potential relationship with binding strength. Because of personal observations confirmed by literature findings.

Using these inputs, the model generates:

A predicted outcome

A confidence level that reflects how reliable the prediction is based on the available data  given

Two major challenges identified during this project was that binding values (Kd)  vary depending on experimental conditions such as assay or buffer type temperature and more; and the inconsistency in the reported experimental conditions.

To address this I implemented a confidence scoring system. This means that instead of just assuming all the data points are equally reliable and consistent, each prediction is evaluated based on:

Availability of Kd values

Consistency of experimental conditions

Completeness of metadata

Presence of the key features identified earlier

This allows the system to not only generate predictions, but also indicate when results should be interpreted with caution. This will help scientists and researchers make smart and informed decisions on selecting aptamer sequences.

The initial analysis and observation showed:

Sequences with moderate to high GC content tended to be associated with stronger binding values

Sequence length influence stability

Certain motifs appeared more frequently in higher affinity sequences

The system was tested using different aptamer sequences and produced a combination of different outcomes. For an example:

A sequence predicted to have strong binding with high confidence

A sequence predicted to have weaker binding with high confidence

A sequence predicted to have strong binding with low confidence, meaning it was flagged as low confidence, where insufficient or inconsistent data limited reliability

This demonstrates that the system is designed not only to predict, but to recognise uncertainty, which is critical in real world scientific applications.

Overall, this shows that even with limited and inconsistent data, the system can still identify meaningful patterns and make useful predictions, while being honest about how reliable those predictions actually are.

This project explores whether patterns and published aptamer data can support early stage binding predictions. A key finding is that binding data is not always directly comparable. Differences in experimental conditions can significantly affect the reported Kd values, limiting direct interpretation. To address this, the system introduces a structured, confidence based approach. Instead of producing definitive results, it generates informed estimates and assigns a reliability score based on data comparability.

This model was developed using limited and inconsistent dataset, focussed on gliadin and aflatoxin targets. Therefore, results should be interpreted as early stage observations, not general conclusions across all aptamers.

Despite this, the approach demonstrates how simple sequence features such as length, GC content and motifs can help identify promising candidates before laboratory testing.

From a practical perspective, this framework may:

Support initial screening of aptamer candidates

Emphasize the importance of data quality

Provide a foundation for more advanced models

To conclude, the main contribution of this project is a proof of concept: that combining sequence analysis, quantitative reasoning and data awareness can support more informed early stage decisions in aptamer research.

This project is an early-stage predictive system with clear plans for refinement. To improve accuracy, I will expand the dataset and incorporate additional experimental variables like temperature. I plan to implement a more advanced machine learning and validate results in a laboratory. Developing a more user-friendly interface is also a priority. Ultimately, this system will evolve into a streamlined tool that helps scientists and researchers optimize aptamers more efficiently, accelerating the development of safety tools for conditions like celiac disease.

I would like to acknowledge the individuals who helped throughout this project.

I first thank the academic mentors who evaluated my project following the provincial science fair and provided targeted feedback that improved the clarity, structure, and communication of my poster.

I would especially like to thank Ana Diaz Fernández. Her published research provided the foundation for this project, and her early guidance and literature recommendations were invaluable. I am especially appreciative of her openness and encouragement, which provided needed motivation during key transitions.

I also thank CytoGroup for their flexibility and for offering materials to help assess feasibility during my initial project. For their support in the final stages of CWSF preparation, my thanks go to Lise and Cindy.

Finally, I want to express my deepest gratitude to my parents. Their unwavering patience and constant encouragement have been my greatest support throughout this entire journey.

AFB1-01: Amaya-González, S., et al. "Aptamer-Based Detection of Aflatoxin B1 in Food Matrices." Food and Chemical Toxicology 145 (2023).

AFB1-02: "Advances in Aflatoxin B1 Sensing via Optical Aptasensors." Spectrochimica Acta Part A (2025).

AFB1-03: "Chemical Communications: Aptamer-Based Detection." Royal Society of Chemistry (2025).

AFB1-04: "Aflatoxin B1 Aptamer (AFB1) Biotin." Antibodies.com (2024).

AFB1-05: Chen, X., et al. "Aptamers for Detection of Aflatoxins: A Review." Journal of Analytical Methods in Chemistry (2017).

AFB1-06: Guo, X., et al. "Selection, Identification, and Application of Aflatoxin B1 Aptamer." ResearchGate (2015).

AFB1-07: Wigier, P., et al. "Aptamer-Based Biosensors for Aflatoxin B1 Detection." Applied Sciences 9.11 (2019).

AFB1-08: Zhang, L., et al. "Aptasensor for Aflatoxin B1 Based on Functional Nanomaterials." Microchimica Acta 186 (2019).

AFB1-09: "Structure and Recognition Mechanism of a DNA Aptamer for Aflatoxin B1." bioRxiv (2022).

AFB1-10: "High-Affinity DNA Aptamers for Aflatoxin B1 Detection." PubMed (2020).

AFB1-11: "Sensors and Diagnostics: AFB1 Monitoring." Royal Society of Chemistry (2025).

AFB1-12: "Food Control: Aptasensing Technology for Mycotoxins." ScienceDirect (2024).

AFB1-13: Li, H., et al. "Aflatoxin B1 Aptamer and Application Thereof." Google Patents. Patent CN116656688A, 2023.

AFB1-14: Rhouati, A., et al. "Aptasensors for Aflatoxin B1 Detection in Food and Feed." Molecules 27.13 (2022).

AFB1-15: "Frontiers in Pharmacology: Mycotoxin Toxicology." Frontiers (2018).

AFB1-16: "Aptamer Sequences and Dissociation Constants for AFB1." PubMed (2024).

AFB1-17: "Aflatoxin B1 Aptamer Technical Data." Sigma-Aldrich (2024).

AFB1-18: "Aptamer Against Aflatoxin B1 Obtained by SELEX." SciSpace (2024).

AFB1-19: "Aflatoxin M1 Aptamer (AFAS3) Biotin." Antibodies.com (2024).

AFB1-20: "Polymers: Aptasensors for Environmental Monitoring." MDPI (2020).

AFB1-21: "Sustainable Chemistry and Pharmacy: Mycotoxin Detection." ScienceDirect (2023).

AFB1-22: "RSC Advances: DNA Aptamer Development." Royal Society of Chemistry (2020).

AFB1-23: "Lateral Flow Assay for Aflatoxin Detection." ResearchGate (2024).

AFB1-24: "Fluorescence Detection of Aflatoxin B1." PubMed (2019).

AFB1-25: "Sensors and Actuators B: Chemical - AFB1 Monitoring." ScienceDirect (2025).

AFB1-26: "Journal of Agricultural and Food Chemistry: Mycotoxin Analysis." ACS Publications (2015).

AFB1-27: "Sensors: Aptamer-Based Technology in Food Safety." MDPI (2025).

AFB1-28: "Fluorometric Determination of Aflatoxin B1." ResearchGate (2019).

AFB1-29: "Aptamers in Lateral Flow Assays." Base Pair Biotechnologies (2018).

AFB1-30: "Food Science: Detection of Aflatoxins." Journal of Food Science (2019).

AFB1-31: "ProQuest: Aptasensing for Food Safety." ProQuest (2024).

AFB1-32: "Selection of DNA Aptamers for Aflatoxin." PubMed (2015).

AFB1-33: "SSRN: Emerging Technologies in Aflatoxin Testing." SSRN (2025).

AFB1-34: "Scientific Reports: Aptasensors for Mycotoxin Detection." Nature Portfolio (2020).

AFB1-35: "HAL Open Science: Aptamer Research." HAL (2016).

AFB1-36: "PMC: Advanced Diagnostics for Aflatoxins." PubMed Central (2022).

AFB1-37: "Europe PMC: Toxicology of Aflatoxins." Europe PMC (2016).

AFB1-38: "Aflatoxin B1 levels, dietary exposure and cancer risk." PMC (2025).

AFB1-39: "Determination of Aflatoxins B1, B2, G1, and G2 concentrations." PMC (2025).

AFB1-40: "Aflatoxins: Source, Detection, Clinical Features and Prevention." MDPI (2023).

AFB1-41: "Aflatoxins: A Global Concern for Food Safety, Human Health." PMC (2017).

AFB1-42: "Determination of Aflatoxin B1 Levels in Organic Spices and Herbs." PMC (2013).

AFB1-43: "Aflatoxin levels in maize and peanut and blood in women and children." PMC (2019).

AFB1-44: "A Recent Overview of Producers and Important Dietary Sources of Aflatoxins." PMC (2021).

AFB1-45: "Bioremediation of Aflatoxin B1 by Meyerozyma guilliermondii." ResearchGate (2024).

Gliadin Series (Gli)

Gli-01: Amaya-González, S., et al. "Aptamer-Based Analysis of Gluten in Foods." PubMed (2015).

Gli-02: "Food Control: Gliadin Sensing in Food Products." ScienceDirect (2017).

Gli-03: Chen, X., et al. "Recent Advances in Aptasensors for Gluten Detection." Frontiers (2024).

Gli-04: "Celiac Disease Aptamer Gli1." Aptagen (2024).

Gli-05: "Aptamers for Detection of Gluten." Google Patents. Patent WO2013178844A1, 2013.

Gli-06: "Alpha-2 Gliadin Aptamer (Gli4) Biotin." Antibodies.com (2024).

Gli-07: "Amino Acid Sequence of a Typical Alpha-Gliadin." ResearchGate (2013).

Gli-08: Svobodova, M., et al. "Selection and Characterization of DNA Aptamers for Gliadin Detection." Biosensors 6.2 (2016).

Gli-09: "Gluten DNA Aptamer." Fusion Biolabs (2024).

Gli-10: "Development of a Competitive Aptasensor for Gliadin." Europe PMC (2013).

Gli-11: "Gliadin-Specific Aptamers for Food Safety." PMC (2024).

Gli-12: "Disposable Electrochemical Aptasensor for Gluten." SciSpace (2024).

Gli-13: "Analytica Chimica Acta: Gliadin Detection." ScienceDirect (2020).

Gli-14: Albanese, D. "Aptamer-Based Sensors for Food Quality." IRIS UNISA (2024).

Gli-15: "ACS Sensors: Advanced Gluten Monitoring." ACS Publications (2024).

Gli-16: "Label-Free Detection of Gliadin Mediator." ResearchGate (2013).

Gli-17: "Selection of Anti-Gluten DNA Aptamers." RIA Asturias (2024).

Gli-18: "Talanta: Aptasensing Technology for Allergens." ScienceDirect (2025).

Gli-19: "Aptamers for Detection of Gluten." Google Patents. Patent WO2021001784A1, 2021.

Gli-20: "Google Scholar: Research in Gliadin Sensing." Google Scholar (2024).

Gli-21: "Analytical Chemistry: High-Affinity Aptamers." ACS Publications (2015).

Gli-22: "Review of Allergen Analytical Testing Methodologies." Food Standards Agency (2024).

Gli-23: Amaya-González, S., et al. "Disposable Electrochemical Aptasensor for Gluten." Analytical and Bioanalytical Chemistry (2014).

Gli-24: Díaz-Fernández, A., et al. "Competitive Aptasensor for Gliadin: Experimental and Computational Study." Analytica Chimica Acta (2020).

Gli-25: Albanese, D., et al. "Submitted Manuscript regarding gluten detection." University of Salerno Institutional Repository (2023).

Gli-26: "Selection and Characterization of DNA Aptamers for Gliadin Detection." ACS Sensors (2024).

Gli-27: Svobodova, M., et al. "Label-free Detection of Gliadin Food Allergen Mediated by Real-time Apta-PCR." Analytical and Bioanalytical Chemistry (2014).

Gli-28: "Selection of Anti-gluten DNA Aptamers in a Deep Eutectic Solvent." University of Oviedo Repository (2024).

Gli-29: Pinto, A., et al. "Label-free detection of gliadin food allergen." Analytical Chemistry (2014).

Gli-30: "Electrochemical Aptasensing of Gliadin in Food Samples." Talanta (2025).

Gli-31: Svobodova, M., et al. "Aptamer selection for gliadin." Biotechnology Letters (2015).

Gli-32: "Methods and Materials for Detecting Gliadin." World Intellectual Property Organization (2021).

Gli-33: "Citations and Research Profile." Google Scholar (2024).

Gli-34: "Structural characterization of aptamers." Analytical Chemistry (2015).

Gli-35: "Review of Allergen Analytical Testing Methodologies." Food Standards Agency (2019).

Gli-36: Díaz-Fernández, A., et al. "Refinement and Truncation of DNA Aptamers Based on Molecular Dynamics Simulations." Journal of Chemical Information and Modeling (2025).

Gli-37: "Label-Free Detection of Gliadin in Food by Quartz Crystal Microbalance." PubMed (2017).

Gli-38: "An Effective Label-Free Electrochemical Aptasensor Based on Gold." PMC (2022).

Gli-39: "Empowering Gliadin Detection: A Visible-Code Semiquantitative Assay." PMC (2025).

"Celiac Disease Symptom Checklist." Celiac Canada (2024). celiac.ca

"Gluten-Related Disorders Menu." Celiac Canada (2024). celiac.ca

"Understanding Gluten Levels." National Celiac Association (2024). nationalceliac.org

"Ryegrain vs Barley Hulled Nutrition." Foodstruct (2024). foodstruct.com

"Gliadin - An Overview." ScienceDirect (2024). sciencedirect.com

"Triticeae - An Overview." ScienceDirect (2024). sciencedirect.com

"Seed Storage Protein - An Overview." ScienceDirect (2024). sciencedirect.com

"Amino Acid." Encyclopedia Britannica (2024). britannica.com

Berg, E. "What is Gluten?" YouTube (2023). youtube.com

"The Science of Gluten." YouTube (2020). youtube.com

Tye-Din, J. A., et al. "The Genetic Basis of Celiac Disease Pathogenesis." Frontiers in Nutrition 6 (2019). frontiersin.org

"Protein Structure." Nature Education (2014). nature.com

Wakim, S., and Grewal, M. "5.6: Protein Synthesis." Human Biology (2020). tru.ca

"Organic Compound." Encyclopedia Britannica (2024). britannica.com

"Chemical Compound - An Overview." ScienceDirect (2024). sciencedirect.com

"Protein Folding - An Overview." ScienceDirect (2024). sciencedirect.com

"Biomolecule." Encyclopedia Britannica (2024). britannica.com

"What is Gluten?" Celiac.org (2024). celiac.org

Mena, M. C., et al. "Comprehensive analysis of gluten-free cereals and pseudocereals." Food Chemistry (2024). sciencedirect.com

"Label-Free - An Overview." ScienceDirect (2024). sciencedirect.com

"Aptamer-based Biosensors for Food Safety." S-Space (2023). snu.ac.kr

Wang, L., et al. "Label-free and ultrasensitive detection of gluten." Nature Communications (2024). nature.com

"40nm Aptamer-Ready Gold Nanoparticle Conjugation Kit." Cytodiagnostics (2024).

Zhao, X., et al. "Recent Advances in Aptasensors for the Detection of Gluten." PMC (2023). nih.gov

Zhang, Y., et al. "Electrochemical Aptasensor for Gluten Detection in Food Samples." Analytica Chimica Acta (2024). sciencedirect.com

Goringer, H. U., et al. "In vitro selection of high-affinity nucleic acid ligands." Nucleic Acids Research 25.3 (1997). oup.com

Mayer, G. Nucleic Acid Aptamers: Selection, Characterization, and Application. Springer, 2015.

Amaya-Gonzalez, S., et al. "Aptamer-based analysis: A promising alternative for food safety control." TrAC Trends in Analytical Chemistry (2022).

Basu, A., et al. "A competitive aptamer-based biosensor for detection of gluten." Sensors and Actuators B: Chemical (2009).

Zhao, Q., et al. "Development of an aptamer-based method for gluten detection." Archives of Virology (2011).

Peters, C. J., et al. "Aptamers as therapeutic and diagnostic agents." Journal of Infectious Diseases (1999).

Schalk, K., et al. "Isolation and Characterization of Gluten Protein Types." Cereal Chemistry (2014).

Valdes, I., et al. "Innovative approach to gluten-free product certification." Journal of AOAC International (2012).

Lexhaller, B., et al. "Comparative analysis of gluten protein fractions." Journal of Cereal Science (2016).

Huang, X., et al. "Biosensors for the detection of gluten: A review." Food Control (2017).

Svobodova, M., et al. "Aptamers for the detection of gluten-derived peptides." Talanta (2011).

"AptBind Gold Nanoparticle Conjugation Kit." CD Bioparticles (2024).

"Aptamer Development Services." Novaptech (2024).

"Aptamers for Complement Therapeutics." Creative Biolabs (2024).

"Cytology and Diagnostic Solutions." CytoGroup (2024).

"Phosphate Buffered Saline (PBS) pH 7.4." Bioenno (2024).

"PBS Stock Solution (10X) pH 7.4." Cayman Chemical (2024).

"PBS, pH 7.4 (10X)." Thermo Fisher (2024).

"Unit Dose Glass Ampoule Swabs." James Alexander (2024).

Svobodova, M., et al. "Selection of 2'-F-modified RNA aptamers against gliadin." Letters in Applied Microbiology (2003).

Li, J., et al. "Advanced materials for energy-efficient sensing." ACS Energy Letters (2024).

"Phosphate Buffered Saline MSDS." Canvax Biotech (2024).