Calcium Binding Affinity in the Mutational Landscape of Troponin-C: Free Energy Calculation, Co-evolution modeling and Machine Learning

Computational protein science has made substantial headway, but accurately predicting the functional effects of mutation in Calcium-binding proteins (CBPs) on Ca2+ binding affinity proves obscure. The complexity lies in the fact that only sequence features or structural information individually offer an incomplete picture on their own. To triumph over this adversity, we introduce a pioneering framework that effortlessly integrates protein sequence evolution information, structural characteristics, and Ca2+ binding interaction properties into a machine learning algorithm. This synthesis has been carried out poised to significantly enhance accuracy and precision in the prediction of the Ca2+ binding affinity towards CBP variants. In our study, we have developed a Ca2+ binding affinity prediction model for various mutants of cardiac Troponin-C protein, to uncover the molecular determinants that contribute binding affinity across protein variants. Our method combines state-of-the-art practices, including a physics-based approach that uses relative binding free-energy (BFE) calculations to assess mutations with implicit polarization. Additionally, it incorporates the impact of evolutionary factors on protein mutations through a theoretical deep mutational scan using a statistical probability model. Support Vector Regression (SVR) algorithms have been used to predict Ca2+ binding affinity based on sequence information, structural properties, and interactions of water molecules with Ca2+ in the EF-hand loop. Our model demonstrates high accuracy and can potentially be generalized for other calcium-binding proteins to predict the effects of point mutations on Ca2+ binding affinity for CBPs.