[Event Debrief] Silicon Valley Artificial Intelligence Hackathon

[Event Debrief] Silicon Valley Artificial Intelligence Hackathon

Prepared by: Wei Wang, Associate Bioinformatics Scientist, National Marrow Donor Program®

I’m delighted to have been sponsored by Healthcare.mn, Minnesota’s home for healthcare innovation and startups, to attend the first AI Genomics hackathon at Google’s Launchpad in San Francisco. The event was kicked off by the SVAI + Accel.ai + NF2 Project. It focused on using artificial intelligence (AI) and machine learning (ML) methods to analyze the high quality whole genome data contributed by Onno Faber, obtained from samples of his healthy tissue and from his neurofibromatosis type 2 (NF2) induced tumor.

Being part of the National Marrow Donor Program® (NMDP)/Be The Match®, an organization with nearly 30 years of experience in immunogenetics and bioinformatics, patient and donor coordination, cell collection, cell supply chain logistics, and regulatory compliance, I’m especially interested in applying new technologies, such as AI and immunotherapy, to develop effective treatments to save patient’s lives. A six-person team was formed on Saturday morning and aimed to apply ML to predict the interaction between human leukocyte antigen (HLA) and tumor-derived peptides known as neoantigens. Specifically, the team members used a tool called persistent homology to study X-ray crystallography structures of HLA proteins, using resulting distance matrices of HLA antigen recognition sites (ARS) as input to a feedforward Artificial Neural Network(ANN) for binding affinity predictions. Meanwhile, I used a pre-built pipeline to simulate antigen processing to generate a set of peptides which were encoded by Onno’s tumor mutations. Those peptides were then inputted into a developed ML based prediction tool for identifying the HLA-restricted neoantigen.

The team presented “ML based Prediction of Immunogenic Neoantigens for Immunotherapy” before the hackathon closed. Our results indicate that topological methods enable better performance in ML based prediction. We believe that a better prediction of peptide-HLA interaction can help bioinformaticians precisely predict neoantigens for potential immunotherapy of Onno’s NF2 condition.

Team worked at the hackathon and presented the results of neoantigen prediction.

Leave a Comment