Synthetic Activators of Cell Migration Designed by Constructive Machine Learning.

Area of Science:

• Computational chemistry
• Machine learning in drug discovery
• Molecular design

Background:

• Constructive machine learning (ML) generates novel data instances.
• ML models can learn patterns from existing chemical structures.
• Cell migration is a critical process in development and disease.

Purpose of the Study:

• To apply constructive ML for de novo design of small molecules.
• To generate novel compounds predicted to modulate cell migration.
• To synthesize and test computationally designed molecules.

Main Methods:

• A recurrent neural network was trained on known cell-migration modulators.
• The ML model generated new candidate molecules.
• Synthesized molecules were tested in a spheroid cell migration assay.
• Functional activity was assessed in medulloblastoma cells.

Main Results:

• Two designed molecules were synthesized and tested.
• The novel small molecules significantly increased medulloblastoma cell migration.
• The results validate the ML model's predictive capability.

Conclusions:

• Constructive machine learning is effective for de novo drug design.
• Computationally generated molecules can possess desired functional properties.
• This method advances the design of druglike molecules for therapeutic applications.