CACHE Challenge #1. Team participation

Scientific problem

CACHE Challenge #1 asked participants to identify binders for the WDR domain of LRRK2 when no known ligands and only an apo protein structure were available.

Motivation

The blinded challenge provided a prospective test of computational hit-finding methods because submitted compounds were purchased and experimentally measured by the organizers.

My role

• Developed and applied CReM-opt for molecular generation and local optimization.
• Applied StreaMD in the MD/MM-GBSA part of the workflow.
• Performed Glide docking and contributed to strategy discussions and analysis of the results.
• Co-author of the collaborative CACHE publication.

Methodology

• CReM-based de novo generation and optimization of candidate molecules.
• Physicochemical and structural filtering followed by consensus docking.
• MD-based analysis and MM/GBSA rescoring of prioritized compounds.
• Similarity-based search of the Enamine REAL database to retrieve commercially accessible analogs of prioritized designs.
• Experimental testing was performed independently by the CACHE organizers.

Results

• Our Round 1 workflow produced 8 SPR binders among 82 compounds tested; one compound of interest was additionally supported by 19F-NMR.
• Hit expansion produced two additional binders, with the best measured KD of 71 µM.
• The workflow was among the third-place teams in the aggregated CACHE Challenge #1 ranking.

Software

• Python
• RDKit
• CReM
• StreaMD
• AutoDock Vina
• Gnina
• Glide
• GROMACS
• gmx_MMPBSA