Derisking Affinity Optimization for Macrocycles and Cyclic Peptides: High-Precision Free Energy Simulations across Five Diverse Targets.

Abstract

Macrocycles and cyclic peptides represent a compelling therapeutic modality for engaging historically challenging biological targets, yet their inherent structural complexity and laborious synthetic demands pose significant hurdles in drug discovery. Given the high resource investment required for macrocyclic synthesis, the integration of rigorous, physics-based computational methods provides a critical advantage for accurately prioritizing design candidates. While Free Energy Perturbation (FEP) has emerged as a transformative method for affinity prediction, its application to complex, beyond-rule-of-five (bRo5) macrocycles demands specialized enhanced sampling protocols and careful consideration of receptor conformational states to effectively navigate their multidimensional conformational landscapes. Here, we present a retrospective validation of the FEP+ framework across five diverse macrocyclic and cyclic peptide inhibitor series: KRAS, PCSK9, MCL-1, JAK2, and Cyclin A/B. Encompassing over 230 unique peptidic and nonpeptidic analogues, our analysis of this consolidated data set demonstrates robust predictive accuracy (global pairwise RMSEΔΔG = 1.06 kcal/mol) and reveals critical insights into the complex interplay between ligand preorganization, hydration dynamics, and binding energetics. The results demonstrate reliable intraseries rank-ordering and robust absolute accuracy across an experimental dynamic range exceeding 10 kcal/mol (>7 orders of magnitude in binding affinity), establishing FEP+ as an effective computational method for derisking and accelerating the discovery of clinically viable bRo5 therapeutics.