CovSite : A High-Throughput Blind Covalent Screening Framework for Reactive Site Detection

Area of Science:

• Computational chemistry
• Drug discovery
• Biochemistry

Background:

• Targeted covalent inhibitors represent a promising therapeutic class but face limitations in computational screening.
• Current methods require prior knowledge of the target site and suffer from low throughput, hindering early drug discovery.

Purpose of the Study:

• To develop CovSite, a blind covalent screening tool for identifying reactive residues across a protein's surface.
• To enhance the accuracy, throughput, and generalizability of computational covalent screening.

Main Methods:

• CovSite utilizes protein structure and electrophile SMILES with a four-filter pipeline: nucleophile identification, solvent accessibility, deprotonation prediction, and reactivity ranking.
• The tool was validated against a large dataset of covalent protein-ligand complexes, including cysteine-targeted ones.

Main Results:

• CovSite achieved 98.5% blind target site hit rate on cysteine-targeted complexes, exceeding existing tools' accuracy.
• It reduced the search space by 97.8% and demonstrated broad nucleophilic residue coverage (cysteine, serine, threonine, lysine, histidine, tyrosine).
• Screening a 200-residue protein takes only 2-3 minutes, showcasing high throughput.

Conclusions:

• CovSite addresses critical gaps in covalent screening throughput, generalizability, and accuracy.
• It serves as a platform technology for computationally guided covalent drug discovery, enabling novel inhibitor generation for understudied targets.