UD-ATOM Hackathon unites minds for breakthroughs in drug discovery
Article by Julia LaPolice | Photos courtesy of Rebecca Lein and Nikhil Rao
The University of Delaware (UD) recently hosted the first-ever Accelerating Therapeutics for Opportunities in Medicine (ATOM) Hackathon, aimed at enhancing the computational resources of the ATOM platform, an open platform for molecular and drug discovery. Organized by UD’s AI Center of Excellence (AICoE) and Data Science Institute (DSI), in collaboration with ATOM, the Frederick National Laboratory for Cancer Research (FNLCR), Lawrence Livermore National Lab (LLNL) and the University of California, San Francisco (UCSF), this three-day event brought together students, researchers and faculty to tackle drug discovery challenges using AI-driven molecular modeling.
Innovating drug discovery with AI
Held at the University of Delaware’s new FinTech Innovation Hub on STAR Campus from July 18-20, 2024, the hackathon was a vibrant, collaborative event where participants worked intensively over a short period to solve technological challenges. According to event organizers, “Over the three days, attendees and mentors worked together to learn to use ATOM Modeling PipeLine (AMPL), create predictive models, run generative AI to create new potential drugs, improve the ATOM code with new features and bug fixes and even run newly developed models in the FDA-operated PrecisionFDA cloud environment.”
Recognizing excellence: Winning teams and their innovations
Decay Detectives: Technical Complexity Award
The Decay Detectives focused on the role of a hyperparameter of neural networks known as weight decay. This regularization technique aims to simplify models to enhance their generalization to test data.
Team members:
- Mentor: Vineeth Gutta, Ph.D. student, UD, computer and information sciences
- Tatiana Acello-Cuellar, Ph.D. student, UD, physics and astronomy
- Janiya Davis, undergraduate student, Bowie State, biology
- Giorgi Gvalia, Ph.D. student, UD, computer and information sciences
- Lauren Olsen, Ph.D. student, UD, computer and information sciences
- Jay Shah, Ph.D. student, UD, chemical engineering
CompTox: Problem Statement Execution Award
Team CompTox explored whether predictive and generative modeling of toxicity could enhance drug discovery throughput. Using the AMPL platform and Generalized Generative Molecular Design (GGMD), they developed a predictive model to estimate LD50 (lethal dose, 50%) as a proxy for toxicity and created new molecules with desirable toxicity properties. They also addressed ethical concerns regarding the potential misuse of AI in drug discovery, particularly the creation of highly toxic compounds.
Team members:
- Mentor: Pinyi Lu, FNLCR
- Logan Hallee, Ph.D. student, UD, bioinformatics
- Colin Horger, undergraduate student, UD, biomedical engineering
- Tom Le, undergraduate student, UD, biomedical engineering
- Niko Rafailidis, Ph.D. student, UD, bioinformatics
- Nikhil Rao, NCAR
Team CodeGuardians: Creativity and Originality Award
The CodeGuardians used static analysis tools to the AMPL software to identify and correct potential errors, improving its overall quality and usability. By addressing issues reported by contestants during setup and operation, these improvements are expected to facilitate smoother onboarding for new users and increase adoption rates.
Team members:
- Mentor: Herman Singh, MassMatrix
- Joel Duah, undergraduate student, Bowie State, biology
- Vasavi Prasanna Kurapati, graduate student, UD, data science
- Nihaal Chowdary Surpani, graduate student, UD, data science
- Saieda Ali Zada, undergraduate student, UD, computer and information sciences
Team Amanda: Remarkable Impact Award
Team Amanda enhanced AMPL’s code base by adding new model types and functionality, including the addition of Gaussian Process models (GPyTorch) and ImbalancedLearn models to AMPL’s model zoo.
Team members:
- Mentor: Amanda Paulson, UCSF
- Natalya Armenta, undergraduate student, Bowie State, biology
- Nikhil Dhanankam, Ph.D. student, UD, computer and information sciences
- Peng He, Incyte
- Andrew Kallai, undergraduate student, UD computer and information sciences
Team Paulson: Honorable Mention
Team Paulson received an honorable mention for developing a predictive model for Drug-Induced Liver Injury (DILI) using cell and biochemical assay data.
Team members:
- Mentor: Amanda Paulson, UCSF
- Wengang Zhang, NIH
- Jaden Clarke, undergraduate student, Delaware State, Biology
- Divy Gabbireddy, undergraduate student, Penn State, computer and information sciences
- Julie Nguyen, Ph.D. student, UD, bioinformatics and data science
- Kien Nguyen, Ph.D. student, UD, computer and information sciences
- Lalitha Priya Sripathi, undergraduate student, UD, business analytics
Building connections across institutions
As highlighted by event organizers, “The 80+ participants sought to learn about, use and improve ATOM. Seven universities were represented, including the University of Delaware, Delaware State University, Bowie State University, Northeastern University, Pennsylvania State University, University of Southern California and UC Berkeley. FNL, UC San Francisco and Lawrence Livermore National Laboratory scientists and staff played a key mentor role for the hackathon teams.”
UD graduate students participated from the departments of Computer and Information Sciences, Physics, Chemistry and Biochemistry, and Chemical Engineering, as well as the data science master’s and the bioinformatics master’s programs. UD undergraduate students represented the departments of Computer and Information Sciences, Biological Sciences, Biomedical Engineering, Electrical and Computer Engineering, and Applied Mathematics. The hackathon provided a valuable platform for UD students to connect with students and mentors from other universities, industries and organizations, fostering new relationships.
Industry and academia collaborate for open science
Sunita Chandrasekaran, a co-organizer of the event and associate professor in the Department of Computer and Information Sciences and co-director of the AICoE, added, “It was excellent that pharma, industries and national labs, namely Incyte, NVIDIA, MassMatrix and Lawrence Livermore National Laboratory scientists, could also be part of this hackathon as participants and mentors. It was truly industry, government and academia working together to improve open-source capabilities to advance open science.”
The event also marked a significant first engagement with the Historically Black Colleges and Universities community to raise awareness of ATOM’s capabilities. Students from Delaware State University and Bowie State University actively participated, preparing in advance and fully engaging in the event.
Acknowledging our generous sponsors
The University of Delaware extends its gratitude to Labware, TechImpact, Collaborative Drug Discovery, National Institute for Innovation in Manufacturing Biopharmaceuticals, Amazon Web Services, DSI, Center for Bioinformatics and Computational Biology (CBCB) and the Graduate College for their sponsorship support of the hackathon.
Special thanks to NVIDIA for their complimentary Deep Learning Tutorial credit giveaways and BVP Coffee Co. for their coffee during the reception. An additional shoutout to UD’s Department of Computer and Information Sciences, CBCB, MathWorks and the Center for Economic Education and Entrepreneurship for their wonderful support in the form of swag giveaways.