Morphological Data Analysis: From Descriptor Development to Predictive Modeling

Area of Science:

• Computational biology
• Cheminformatics
• Machine learning in drug discovery

Background:

• Morphological fingerprints capture cellular phenotype from image data.
• These fingerprints can be used for downstream computational analysis.
• Machine learning models can leverage these fingerprints for predictive tasks.

Purpose of the Study:

• To explore the computational processing of morphological fingerprints.
• To demonstrate their application in compound similarity search and activity prediction.
• To provide a guide for leveraging these fingerprints in drug discovery.

Main Methods:

• Data preparation including ingestion, standardization, and normalization.
• Computation of similarity searches using morphological and structural fingerprints.
• Building and evaluating a machine learning model for activity prediction.

Main Results:

• Successful application of morphological fingerprints for similarity searches.
• Demonstration of machine learning model for predicting estrogen receptor activity.
• Insights into model tuning, testing, and interpretation were provided.

Conclusions:

• Morphological fingerprints are valuable for computational analysis in drug discovery.
• These fingerprints facilitate compound similarity searches and activity prediction.
• The chapter provides a practical guide using Jupyter notebooks for implementation.