AI-based prediction of SARS-CoV-2 mutations
The process of SARS-CoV-2 mutation may affect the effectiveness of the vaccines. In our presentation, I will describe a pipeline which generates a big set of peptide candidates for vaccines. We are relying on the GISAID dataset and two machine learning tools: netMHCpan and netCHOP. After obtaining that dataset, we tackle the problem of optimization. This is done by choosing the most conservative peptides, which are also representative of the majority population. Additionally, our project includes a model virus evolution. During Researchathon – a hackathon organized by Stanford University, in cooperation with a team of experts, we developed mutation-sensitive scores of peptides by using XGBoost library. At Covid Genomics, we are trying to improve the usability of LSTM and we are eager to propose our solutions.
Bio
Grzegorz Preibisch is a double degree student. He finished 4th year of Medicine at the Medical University of Warsaw and 3rd year of Mathematics at the University of Warsaw. He is interested in Stochastic Processes, Machine Learning and how they can apply in Medicine. He is also a co-founder of Covid Genomics – a startup with a mission of solving the problems of the SARS-CoV-2 pandemic.