Deciphering Long-Range Effects of Mutations: An Integrated Approach Using Elastic Network Models and Protein Structure Networks

Abstract

Understanding the impact of genetic variants on protein structure and function is essential for deciphering disease mechanisms. The MAVISp framework offers a systematic approach for evaluating structural effects, including variants with long-range impact. In this study, we critically evaluate and refine the LONG_RANGE module of MAVISp, leveraging data from over 400 proteins to optimize parameters for detecting significant response sites. We implement a systematic filtering workflow integrating allosteric free energy, distance constraints, solvent accessibility, and pocket localization to prioritize biologically relevant variants. We benchmarked the results against experimental data from deep mutational scans to identify the optimal combination of thresholds and filtering steps for assessing the impact of allosteric variants at response sites. Our analysis reveals that a 5.5 Å distance threshold, based on atomic distances, effectively minimizes the occurrence of local contacts in the allosteric map while preserving long-range effects. To address the limitations of the elastic network model for predicting allosteric free energy changes in non-globular proteins, we propose introducing three different metrics to assess protein globularity within the MAVISp framework, thereby supporting the design of the trimming to be applied to the input structure. Furthermore, we illustrate the potential of incorporating molecular dynamics simulations and algorithms for path analysis to confirm pairs of allosteric mutation sites and response sites involved in distal communication. Overall, we established a robust and scalable workflow for detecting allosteric protein variants, offering insights into structural communication and disease-associated protein variants.