Name: Kris Bouchard
Pronouns: he/him/his
Biography:
Kristofer Bouchard is the leader of the Computational Biosciences Group. He is a neuroscientist and data scientist in the Scientific Data Division and Biological Systems & Engineering Division.
Institution/Lab: Lawrence Berkeley National Laboratory
Website: https://crd.lbl.gov/divisions/scidata/computational-biosciences/
SRP Collaboration Topic/Title: Foundation model for proteins
Field or research area: AI and biology
Please select all the topical areas that apply to your project:
Computational Science Applications (i.e., bioscience, cosmology, chemistry, environmental science, nanotechnology, climate, etc.)
Brief Abstract:
Foundation models are massive (>1B parameters) deep networks that have been pre-trained on large and diverse data sets and can then be transfer-learned to other relevant tasks. Salient examples include GPT-4, DALL-E2, etc., While this class of models has been impactful in industrial applications, foundation models for science are nascent. Proteins are fundamental units of biological processes. This project will build of diverse extent work across LBL to create a foundation model that integrates all known protein structure and sequences. This protein foundation model could then be used as the basis for, e.g., generating novel protein sequences with enhanced functionality, inferring atomic structure from SAXS data, or inferring the function of newly discovered proteins.
Desired relevant skills, background, or interests:
pytorch, large-language models, transformers, graph neural networks,
Other comments:
Do any special requirements apply? other
Other, specify: none
Keywords:
machine learning, biology, data management, foundation models, simulations,
Lightning Talk Title: The Computational Biosciences Group