Wavelet-Enhanced Interaction Entropy Analysis of Pyrithiamine Binding to Thiamine Pyrophosphokinase

Accurate estimation of protein ligand binding free energies remains challenging for flexible systems in which conformational heterogeneity leads to nonstationary sampling and broad interaction energy distributions. In such cases, the interaction entropy (IE) approach may exhibit poor convergence due to violation of the quasi-harmonic assumptions underlying the cumulant expansion. Here, molecular dynamics simulations were performed to examine the binding of pyrithiamine to thiamine pyrophosphokinase (PDB ID: 2F17), a system characterized by loop-driven structural flexibility. Binding free energies were evaluated using the MMPBSA framework combined with IE analysis. The interaction energy trajectory displays significant low-frequency conformational drift arising from coupled ligand protein motions, resulting in non-Gaussian fluctuations. To reduce this instability, a wavelet-based decomposition was introduced to separate slow structural evolution from fast equilibrium fluctuations. The extracted high-frequency component exhibits near-Gaussian behavior and improves the numerical stability of the entropy estimate within this system. Incorporation of the filtered entropy yields binding free energies consistent with experimental observations. These results illustrate that wavelet-based separation of time scales can provide a practical means of stabilizing IE-based entropy estimates in flexible protein ligand complexes.