Strategies for robust, accurate, and generalizable benchmarking of drug discovery platforms

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Benchmarking is an important step in the improvement, assessment, and comparison of performance drug discovery platforms technologies. We revised existing benchmarking protocols our Computational Analysis Novel Drug Opportunities (CANDO) multiscale therapeutic platform to improve utility performance. optimized multiple parameters used candidate prediction assessment with these updated protocols. CANDO ranked 7.4% known drugs top 10 compounds for their respective diseases/indications based on drug-indication associations/mappings obtained from Comparative Toxicogenomics Database (CTD) using parameters. This increased 12.1% when mappings were Therapeutic Targets Database. Performance indication was weakly correlated (Spearman correlation coefficient >0.3) size (number associated indication) moderately (correlation >0.5) compound chemical similarity. There also moderate between new original assessing per each protocol. results dependent source mapping used: a higher proportion indication-associated recalled 100 (TTD), which only includes FDA-approved associations (in contrast CTD, drawn literature). created compbench, publicly available head-to-head protocol that allows consistent different platforms. Using this protocol, we compared two pipelines repurposing within CANDO; primary pipeline outperformed another similarity-based still development clusters signatures Gene Ontology terms. Our study sets precedent complete, comprehensive, comparable platforms, resulting more accurate predictions.

Language: Английский

Strategies for Robust, Accurate, and Generalisable Benchmarking of Drug Discovery Platforms

Published: Dec. 24, 2024

Benchmarking is an important step in the improvement, assessment, and comparison of performance drug discovery platforms technologies. We revised existing benchmarking protocols our Computational Analysis Novel Drug Opportunities (CANDO) multiscale therapeutic platform to improve utility performance. optimized multiple parameters used candidate prediction assessment with these updated protocols. CANDO ranked 7.4% known drugs top 10 compounds for their respective diseases/indications based on drug-indication associations/mappings obtained from Comparative Toxicogenomics Database (CTD) using parameters. This increased 12.1% when mappings were Therapeutic Targets Database. Performance indication was weakly correlated (Spearman correlation coefficient _>_0.3) size (number associated indication) moderately (correlation _>_0.5) compound chemical similarity. There also moderate between new original assessing per each protocol. results dependent source mapping used: a higher proportion indication-associated recalled 100 (TTD), which only includes FDA-approved associations (in contrast CTD, drawn literature). created compbench, publicly available head-to-head protocol that allows consistent different platforms. Using this protocol, we compared two pipelines repurposing within CANDO; primary pipeline outperformed another similarity-based still development clusters signatures Gene Ontology terms. Our study sets precedent complete, comprehensive, comparable platforms, resulting more accurate predictions.

Language: Английский