Reliable in silico ranking of engineered therapeutic TCR binding affinities using MMPBSA and MMGBSA calculations.

Abstract Accurate and efficient in silico ranking of protein-protein binding affinities is useful for protein design with applications in biological therapeutics. One popular approach to rank binding affinities is to apply the molecular mechanics Poisson Boltzmann/generalized Born surface area (MMPB/GBSA) method to molecular dynamics trajectories. This provides a compromise between rapid but approximate scoring functions of single structures and more sophisticated methods such as free energy perturbation. Optimal MMPB/GBSA parameters tend to be system specific. Here, we identify protocols that enable reliable evaluation of the effect of mutations in a T-cell receptor (TCR) in complex with its natural target, the peptide-human leukocyte antigen (pHLA). The development of affinity-enhanced engineered TCRs towards a specific pHLA is of great interest in the field of immunotherapy. Our study highlights the importance of using a higher than default internal dielectric constant, especially in the case of charge changing mutations. Including explicit solvation and/or entropy corrections may deteriorate the ranking of single point variants due to the errors associated with these additions. For multi-point variants, however, these corrections were important for accurate ranking. We also demonstrate how potential outliers could be identified in advance by analyzing changes in the hydrogen bonding networks at the binding interface. Finally, using bootstrapping we show that as few as 5-10 replicas of short (4 ns) MD simulations may be sufficient for reproducible and accurate ranking of candidate TCR variants. Our work demonstrates that reliably ranking TCR variant binding affinities can be achieved at moderate computational cost. The protocols developed here can be applied towards in silico screening during the optimization of therapeutic TCRs, potentially reducing both the cost and time taken for biologic development. Author summary T-cell receptors (TCRs) recognise an extracellular presentation platform that displays peptides that can be derived from infectious agents, or mutated/dysregulated cellular proteins. Soluble TCRs can be engineered to bind with enhanced affinity and specificity towards a given protein-peptide complex displayed only on diseased cells (i.e., cancerous cells), providing the basis for potent biologic drugs. Here, we performed computer simulations that describe how atoms and molecules interact with one another with atomistic detail in order to predict how the binding affinity (i.e., how sticky the drug is to its target) changes upon mutation of the TCR. By systematically testing different settings and methodologies for our binding affinity calculations, we identified optimized protocols to more accurately predict how a given mutation (or set of mutations) will impact the binding affinity. We also identified differences in the optimal protocol depending on the number of mutations that separate the TCRs. Our resulting calculated affinities correlated very well with experimental results, showcasing the promise of incorporating our computational approach into the design of affinity-enhanced TCRs for therapeutic use. The protocols developed can lead to faster and less expensive development of effective TCR-based treatments for various cancers and other diseases.