The Field Flashcards
get to know the people, labs, and papers that define the field I'm in
Bonnie Berger
Mathematics professor at MIT and head of the Berger Group.
Papers:
- Topsy-Turvy (PPI predictor)
Debora Marks
Computational biologist who runs the Marks Lab in the Department of Systems Biology at Harvard Medical School. Lab vision is to computationally address three critical challenges: protein conformational plasticity in health and disease, genome-wide evaluation of mutations on disease likelihood, antibiotic resistance and personal drug response, and synthetic protein design.
Papers:
- ProteinGym
David Baker
Runs the Baker Lab (Institute for Protein Design) at the University of Washington.
Papers/Contributions:
- The entire Rosetta platform
DSMBind
Jin et al., Broad Institute, 2023.
An unsupervised binding energy prediction framework that does not need experimental binding data for training. This energy-based model estimates the likelihood of a protein complex via SE(3) denoising score matching (DSM). It is for both protein-protein and protein-ligand predictions. DSMBind maximizes the log-likelihood (minimizes the energy) of crystal structures in a training set.
SE(3) in protein modeling tasks
SE(3) can be used to superimpose protein structure for comparative analysis; describe dynamics of protein molecules in simulations where they move and interact; make predictions about protein folding, interactions, and conformational changes
Tranception
Marks lab, 2022
A novel transformer architecture leveraging autoregressive predictions and retrieval of homologous sequences at inference to achieve SOTA fitness prediction performance
ProteinGym
Marks lab, 2022
An extensive set of multiplexed assays of variant effects, introduced in the Tranception paper.
DeepSequence
Marks lab, 2017
A generative, unsupervised VAE-based method for biological sequences. Given a multiple sequence alignment as input, it can be used to predict accessible mutations, extract quantitative features for supervised learning, and generate libraries of new sequences satisfying apparent constraints.
