MATCHMAKERS - Solving T-cell receptor recognition and design via integrated high-throughput screening and structural, functional and computational approaches

ABSTRACT: UCSF
Understanding how T cell receptors (TCRs) see tumor antigens presented by MHCs is necessary to fully
understand how the immune system recognizes tumor antigens, and to reap the full potential of antigen-specific
immunotherapy. To achieve this goal, a quantum leap forward is required in which the revolutionary advances in
machine learning are combined with a large volume of structure, function, data on matched TCR-pMHC pairs.
The development of accurate predictors of TCR-antigen recognition will be dependent on the creation and
integration sequencing-based datasets with high-throughput structural and functional insights. Our proposal,
submitted as a CRUK/NCI Grand Challenge team (MATCHMAKERS) will combine researchers with expertise in
immunology, methods development, structural biology, and computation to enable generalized prediction and
design of TCR recognition. This work will be spread across four Work Packages (WPs):
WP1: Large-scale generation of TCR-pMHC pairs from naturally occurring sources. We will build datasets of
naturally occurring TCR-pMHC pairs. Our team will use an array of approaches to collect these datasets, from
humans and from mouse models, and in the context of both cancer and immunity more generally.
WP2: Ultra-high throughput TCR-pMHC matching using molecular engineering. Efforts to create general models
will require a broader array of data than feasible to collect from natural TCR systems. We will use an array of
synthetic approaches developed by our team to comprehensively match TCRs with pMHCs to train
computational models.
WP3: Large-scale structural and biochemical analyses of TCR-pMHC interactions. A key to our team’s vision is
to match interaction datasets with high throughput structural and functional insights. A deep understanding of
how the TCR contacts with MHC helices control function and orientation will be essential for training and testing
computational models.
WP4 AI-based prediction and design of TCR-pMHC interactions. We will integrate our data to train next-
generation algorithms capable of generally predicting and designing TCR-pMHC interactions. These predictions
will proceed through a reiterative testing and feedback circuits for further model optimization.

Grant Number: 1OT2CA297383-01
NIH Institute/Center: NIH
Principal Investigator: Peter Bruno