PhD Fellowships in Neuro-symbolic Machine Learning for Biology and Drug Design

Project Overview

This project involves two PhD fellowships focused on neuro-symbolic machine learning, an innovative approach that combines symbolic reasoning with neural networks to tackle complex problems in biology and drug discovery. The research aims to advance AI models for biological sequence analysis and protein-ligand interaction interpretation to improve understanding of diseases and drug design.

What You Will Do

PhD 1 involves developing probabilistic circuits as alternatives to variational autoencoders for protein family modeling, including scalable constrained decoding that respects biological constraints. This includes a research stay at the University of Edinburgh.

PhD 2 aims to create a neuro-symbolic AI framework integrating biomedical literature and 3D molecular structures, targeting G protein-coupled receptors to identify novel drug binders. This project is part of the Center for Pharmaceutical Data Science Education (CPDSE) and offers interdisciplinary training.

Expected Outcomes

The research will produce novel neuro-symbolic AI models capable of reliable, interpretable, and constraint-aware predictions in computational biology and drug discovery. These outcomes have potential to accelerate rational drug design and improve therapeutic discovery pipelines.

Why This Matters

The research addresses urgent needs to enhance AI systems in biology to be more interpretable and biologically faithful, crucial as foundation models grow in complexity. By bridging symbolic reasoning with neural methods, the projects can revolutionize biochemical modeling and drug discovery, ultimately impacting human health.