Automated protein-protein structure prediction of the T cell receptor-peptide major histocompatibility complex

The T Cell Receptor (TCR) recognition of a peptide-major
histocompatibility complex (pMHC) is a crucial component of the adaptive
immune response. The identification of TCR-pMHC pairs is a significant
bottleneck in the implementation of TCR immunotherapies and may be
augmented by computational methodologies that accelerate the rate of TCR
discovery.  The ability to computationally design TCRs to a
target pMHC will require an automated integration of next-generation
sequencing, homology modeling, molecular dynamics (MD), and TCR ranking.
We present a generic pipeline to evaluate patient-specific, sequence-based
TCRs to a target pMHC. The most expressed TCRs from 16 colorectal cancer
patients are homology modeled to target the CEA peptide using Modeller and
ColabFold. Then, these TCR-pMHC structures are compared by performing an
automated molecular dynamics equilibration. We find that Colabfold
generates starting configurations that require, on average, an ~2.5X
reduction in simulation time to equilibrate TCR-pMHC structures compared
to Modeller. In addition, there are differences between equilibrated
structures generated by Modeller and ColabFold.  Moreover, we
identify TCR ranking criteria that may be used to prioritize TCRs for
evaluation of in vitro immunogenicity.