Competitive Workflow: novel software architecture for automating drug design.

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Bioinformatics

Background:

• Industrialization of laboratory processes in drug discovery generates vast data.
• Current decision-making relies on limited expert individuals, creating bottlenecks.
• Information management systems and computer-aided molecular design show promise but are limited by human decision-making.

Purpose of the Study:

• To introduce Competitive Workflow, a distributed multi-agent system for automating expert decision-making.
• To extend workflow architectures by modeling tacit expert knowledge for pathway selection.
• To review related workflow management systems and multi-agent approaches.

Main Methods:

• Development of a distributed multi-agent system (Competitive Workflow).
• Integration of workflow architectures with models of expert tacit knowledge.
• Application of the 'Discovery Bus' implementation to meta-quantitative structure-activity relationship analysis.

Main Results:

• Competitive Workflow automates decision-making, addressing a key bottleneck in drug discovery.
• The system models expert knowledge for selecting alternative pathways within workflows.
• Demonstrated application in meta-quantitative structure-activity relationship analysis.

Conclusions:

• Automating expert decision-making through systems like Competitive Workflow can significantly enhance drug discovery productivity.
• Distributed multi-agent systems offer a scalable solution for complex decision processes.
• The 'Discovery Bus' implementation shows practical utility in advanced cheminformatics tasks.