Modeling Biomolecular Interactions with Generative Models
Name
corso-gcorso-phd-eecs-2025-thesis.pdf
Description
Thesis PDF
Size
25.88 MB
Format
Adobe PDF
Checksum (MD5)
8f07fb1399786d38de8b61a41e8aa3c7
Author(s)
Corso, Gabriele
Advisor(s)
Jaakkola, Tommi S.
Barzilay, Regina
Date Issued
May 2025
Publisher
Massachusetts Institute of Technology
Abstract
In 2021, DeepMind’s AlphaFold2 revolutionized single-chain protein structure prediction achieving atomic accuracy, solving a longstanding challenge in biology. However, understanding biomolecular interactions, a critical problem for advancing drug discovery and biological research, remained unsolved. This thesis presents our research to redefine the machine learning approach to this problem, modeling structures with a new generative paradigm and tailoring the neural architectures and learning tasks to the specific challenges that arose. These ideas combined with significant engineering efforts led us to develop a class of open-source models from DiffDock to the recent Boltz-1. These have significantly pushed our ability to understand biomolecular interactions, they have been widely adopted in industry and academia to help with drug development and protein design and they have opened the door to new research paradigms to push biological research further.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
