ScaffComb: A Phenotype-Based Framework for Drug Combination Virtual Screening in Large-Scale Chemical Datasets

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Combinational therapy is crucial for overcoming cancer drug resistance observed with monotherapy.
• Advancements in deep learning and pharmacological data enable drug pair prediction, but library sizes are limited.
• Existing virtual screening methods struggle with large-scale drug combination discovery.

Purpose of the Study:

• To introduce ScaffComb, a novel framework for large-scale virtual screening of drug combinations.
• To integrate phenotypic information with molecular scaffolds for enhanced drug discovery.
• To identify potent drug combinations and discover new synergistic mechanisms.

Main Methods:

• ScaffComb framework development integrating phenotypic information into molecular scaffolds.
• Validation using the US Food and Drug Administration (FDA) dataset to reidentify known drug combinations.
• Application to screen large chemical databases like ZINC and ChEMBL.

Main Results:

• Successful reidentification of known drug combinations using the FDA dataset.
• Screening of ZINC and ChEMBL databases yielded novel drug combinations.
• Demonstrated ability of ScaffComb to discover new synergistic mechanisms.

Conclusions:

• ScaffComb is the first phenotype-based virtual screening method for large-scale drug combination discovery.
• The framework effectively identifies potent drug combinations and potential synergistic mechanisms.
• ScaffComb advances the field of computational drug discovery for cancer therapy.