Investigating Errors in Alchemical Free Energy Predictions Using Random Forest Models and GaMD

State-of-the-art <i>in silico</i> ΔΔ<i>G</i> predictions for antibody-antigen complexes achieve an accuracy of ±1 kcal/mol. While this is sufficient for high-throughput screening or affinity maturation, it is insufficient for assessing the criticality and impact of post-translational modifications (PTMs) during clinical development. PTMs that impair binding by >50% pose a major risk to achieving the desired therapeutic bioactivity and must be controlled within defined limits to ensure product quality. A 50% loss in the dissociation constant (<i>K</i>_D) corresponds to a ΔΔ<i>G</i> of +0.5 kcal/mol, thus requiring a ±0.5 kcal/mol accuracy threshold for <i>in silico</i> predictions to become practically actionable in clinical phases. In this work, we use conventional molecular dynamics thermodynamic integration (cMD-TI) to generate ΔΔ<i>G</i> predictions and develop an error analysis approach using random forest (RF) models and end-state Gaussian accelerated molecular dynamics (GaMD). This approach provides untargeted insight into inadequate sampling of key degrees of freedom using only cMD-TI and end-state GaMD. We identify bulky side-chain undersampling and violation of energetically relevant interatomic interactions as major sources of error, and our GaMD-based error corrections lead to >1 kcal/mol improvements in accuracy in our most erroneous cases. When applied to a set of 13 predictions, the GaMD-based error correction reduced the root-mean-square error (RMSE) from 1.06 ± 0.22 to 0.70 ± 0.18 kcal/mol. This work introduces the application of alchemical free energy predictions to estimate PTM impacts on bioactivity and investigates the current errors that limit their practical use in clinical development.