Leveraging Image-Derived Phenotypic Measurements for Drug-Target Interaction Predictions

Area of Science:

• Biochemistry
• Pharmacology
• Computational Biology

Background:

• Protein kinases are key drug targets in cancer therapy, regulating cell proliferation.
• Efficient identification of novel small molecules for kinase inhibition remains a challenge in drug discovery.

Purpose of the Study:

• To develop a novel in silico drug discovery approach for identifying kinase targets impacting nuclear receptor signaling.
• To leverage high-content analysis (HCA) imaging data for kinase target identification.

Main Methods:

• Generated high-throughput imaging data from an siRNA human kinome screen on engineered cells.
• Extracted population-median and bag-of-words (BoW) descriptors from imaging data.
• Applied single-task learning, multi-task learning, and collaborative filtering for drug-kinase-target interaction prediction.

Main Results:

• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.86 for target-based drug discovery.
• Obtained an AUC of 0.79 for ligand-based drug discovery.
• Demonstrated the utility of imaging-based information in virtual screening.

Conclusions:

• Imaging data can significantly enhance existing virtual screening methods.
• This approach offers a more time and cost-efficient drug discovery process.
• The developed computational method shows promise for identifying novel kinase inhibitors.