Interpretable machine learning uncovers structural determinants of Wnt-Wntless binding specificity from atomistic...

Area of Science:

• Molecular Biology
• Computational Biology
• Biophysics

Background:

• Wnt proteins are crucial for cell development and signaling.
• Wnt proteins interact with the Wntless (Wls) protein for proper trafficking.
• Understanding Wnt-Wls interactions is vital but challenging due to Wnt protein diversity.

Purpose of the Study:

• To elucidate the molecular mechanisms of Wnt-Wls binding specificity across different Wnt proteins.
• To identify key residue interactions governing Wnt-Wls binding affinity and trafficking efficiency.
• To develop a generalizable computational approach for analyzing protein-protein interactions.

Main Methods:

• Atomistic molecular dynamics simulations were employed to model Wnt-Wls interactions.
• Supervised machine learning, specifically Random Forest classification, was used to analyze binding data.
• A local structure alignment algorithm and two-stage clustering were implemented for feature selection and comparison.

Main Results:

• The study identified key residue pairs, both known and novel, that dictate Wnt-Wls binding specificity.
• Binding specificity arises from distributed interactions across the protein interface.
• Interpretable machine learning effectively uncovered crucial biophysical interactions.

Conclusions:

• The integrated computational strategy successfully elucidates Wnt-Wls binding mechanisms.
• This data-driven approach can guide experimental validation and therapeutic targeting for Wnt pathway modulation.
• The methodology is generalizable to other complex protein-protein interaction systems.