DoRIAT: A Bayesian framework for interpreting and annotating TCR-pHLA docking runs.

Christos Maniatis, Zahra Ouaray, Chengkai Xiao, Thomas P.E. Dixon, Charlie Naylor, James R Snowden, Dr Michelle Teng, Dr Jacob Hurst

Keywords: GP, Molecular Docking, Bayesian, Structure Prediction, Docking, TCR, TCR-pHLA

TL;DR: A Bayesian framework for interpreting and annotating TCR-pHLA docking runs.

Abstract: The advent of sequence-to-structure deep-learning models has transformed the protein engineering landscape by providing an accurate and cost-effective way to determine crystal structures. Despite their accuracy, deep-learning predictions tend to offer limited insights into protein dynamics. To improve conformational exploration in the context of T cell receptor (TCR) docking, we have developed a machine learning pipeline that combines deep-learning predictions with molecular docking. In this report, we introduce $\textbf{Do}$cking $\textbf{R}$un $\textbf{I}$ntepretation and $\textbf{A}$nnotation $\textbf{T}$ool (DoRIAT). In contrast to frameworks that score models based on interface interactions, DoRIAT uses a set of parameters that summarize the binding conformations. We use DoRIAT to annotate TCR-pHLA docking runs, score the resulting complexes, identify models close to the native crystal structure, and create ensembles of models with similar binding conformations. Our results demonstrate that the single structural model selected by DoRIAT as the closest representation of the crystal structure lies within the top $10$ docked models, ranked by Root Mean Squared Distance (RMSD), in approximately $80\%$ of $\text{HLA-A}^{\ast}\text{02}$ complexes considered.

Release To Public: Yes, please release this paper to the public

Submission Number: 7