An Integrated Machine Learning Approach Delineates Entropic Modulation ofα-synuclein by Small Molecule

Abstract

AbstractAberrant misfolding and progressive aggregation of the intrinsically disordered protein (IDP),α-synuclein (αS), is associated with the aetiology of several neurodegenerative diseases. However, the discovery of potential small-molecules targetingαS often hits a bottleneck due to the lack of a designated binding pocket in its structurally heterogeneous ensemble. Here, by integrating Markov State Modelling approach with a deep-learning work-flow namelyβ-Variational Autoencoder (β-VAE), we statistically dissect a large conformational ensemble ofαS in neat water and aqueous solution of fasudil (small molecule of interest) into a set of mutually distinct metastable states. We find that fasudil interacting withαS leads to the formation of significantly higher number of conformational states than that in water, akin to an ensemble-expansion mechanism. The ensembles possess mutually diverse intra-protein contact maps and display strong conformation-dependence in residue-wise interaction with the small molecule. A thermodynamic analysis reveals that fasudil influences the structural repertoire ofαS by tuning protein backbone entropy, without perturbing entropic ordering of surrounding solvent. Together, the investigation demonstrates the distinctive effect of a drug molecule on an IDP, elucidating entropic modulation and ensemble expansion as a mechanism for conformational changes.