A machine learning and live-cell imaging tool kit uncovers small molecules induced phospholipidosis

Area of Science:

• Cell Biology
• Drug Discovery
• Toxicology

Background:

• Drug-induced phospholipidosis (DIPL) involves lysosomal phospholipid accumulation, leading to adverse clinical effects.
• DIPL can interfere with phenotypic responses in functional studies using chemical probes, impacting research accuracy.

Purpose of the Study:

• To develop robust methods for early prediction and quantification of phospholipidosis (PL) during drug discovery.
• To evaluate the PL-inducing potential of chemical probes and their relevance in target validation.

Main Methods:

• A high-content live-cell imaging approach was employed to assess PL.
• Machine learning models were trained and evaluated using a comprehensive set of public compounds.
• SHapley Additive exPlanations (SHAP) were used for model interpretation.

Main Results:

• The developed method successfully evaluated chemogenomic and lysosomal modulation libraries.
• Analysis revealed that closely related molecules, including chemical probes and their controls, can exhibit different PL induction potentials.
• The study highlights the critical need for identifying PL in chemical biology.

Conclusions:

• A versatile high-content live-cell imaging and machine learning approach can effectively predict and quantify DIPL.
• Accurate PL assessment is crucial for reliable target validation and robust chemical biology research.
• This method aids in early identification of potential drug liabilities and ensures the quality of chemical probes.