Computer-aided selection of compounds for antitumor screening: validation of a statistical-heuristic method

Area of Science:

• Computational chemistry
• Pharmacology
• Drug discovery

Background:

• The National Cancer Institute (NCI) employs a mouse tumor prescreen to evaluate potential anticancer compounds.
• Accurate selection methods are crucial for optimizing the prescreening process and identifying promising drug candidates.
• Previous statistical-heuristic methods require validation for their efficacy in compound selection.

Purpose of the Study:

• To validate a statistical-heuristic computational method for selecting compounds for the NCI mouse tumor prescreen.
• To compare the predictive performance of this computational method against human expert (chemist) evaluation.

Main Methods:

• A large dataset of approximately 35,000 compounds from the NCI collection was ranked by predicted probability of anticancer activity.
• A second study involved a chemist and a computer algorithm independently rating nearly 1,000 previously unselected compounds for activity.

Main Results:

• The statistical-heuristic method successfully identified active compounds, with 34% of active compounds found in the top 10% of the ranking.
• The computational method and the human chemist demonstrated comparable yields of biologically active compounds.
• Despite similar yields, the agreement between the computational method's and the chemist's compound selections was relatively low.

Conclusions:

• The statistical-heuristic method is a valid and effective tool for prioritizing compounds in cancer research prescreening.
• Computational approaches can achieve comparable results to expert human evaluation in identifying active compounds.
• Further research may explore combining computational predictions with expert judgment to enhance selection accuracy and agreement.