Identification of lysosomotropism using explainable machine learning and morphological profiling cell painting data.

Area of Science:

• Pharmacology and Drug Discovery
• Computational Chemistry
• Cell Biology

Background:

• Lysosomotropism is a crucial drug property with broad pharmaceutical relevance, affecting compounds with varied structures and targets.
• While lipophilicity and basicity are commonly associated with lysosomotropism, not all compounds with these properties exhibit it.
• Accurate prediction of lysosomotropism is essential for optimizing drug development and understanding drug-target interactions.

Purpose of the Study:

• To establish cell painting assay (CPA) as a reliable method for identifying lysosomotropic compounds.
• To investigate the physicochemical drivers of lysosomotropism beyond simple lipophilicity and basicity.
• To leverage machine learning for predicting and understanding the molecular basis of lysosomotropism.

Main Methods:

• Utilized morphological profiling via cell painting assay (CPA) to phenotype compounds.
• Applied matched molecular pair analysis (MMPA) to identify key structural determinants.
• Employed explainable machine learning (XML) models to analyze compound properties and predict lysosomotropism.

Main Results:

• Identified that only 35% of compounds with expected physicochemical properties displayed lysosomotropism.
• Matched molecular pair analysis (MMPA) did not reveal specific substructures responsible for lysosomotropism.
• Explainable machine learning (XML) highlighted higher lipophilicity, basicity, molecular weight, and lower topological polar surface area as key inducers.

Conclusions:

• Cell painting assay (CPA) serves as a robust surrogate for identifying lysosomotropism.
• Lysosomotropism is a complex phenomenon influenced by a combination of physicochemical properties, not solely lipophilicity and basicity.
• Explainable machine learning (XML) provides valuable insights into the molecular drivers of lysosomotropism, aiding drug design.