Exploring the spatial effects influencing the EGFR/ERK pathway dynamics with machine learning surrogate models.

Area of Science:

• Cell biology
• Biophysics
• Computational biology

Background:

• Cellular fate is governed by intracellular pathways.
• Cell shape and structure impact molecular diffusion and interactions.
• Understanding spatial effects on intracellular regulation is crucial.

Purpose of the Study:

• To apply machine learning (ML) to a spatial model of epidermal growth factor receptor (EGFR) signaling.
• To investigate the influence of spatial parameters on intracellular pathway activation.
• To develop computationally efficient surrogate models for complex cell signaling.

Main Methods:

• Developed and trained ML models (neural networks, random forests, generalized linear models) using 10,000 numerical simulations.
• Simulated EGFR signaling under varying conditions: diffusion speeds, inactivation rates, cell/nucleus sizes, and cell structures.
• Calculated cumulative activation of molecules and transcription factors.

Main Results:

• ML models, particularly neural networks and random forests, achieved minimal mean square error (MSE).
• Smaller cell/nucleus radii, higher diffusion coefficients, and lower inactivation rates increase transcription factor activation.
• ML predictions were validated by numerical simulations.

Conclusions:

• ML provides a computationally efficient method to study spatial effects on intracellular pathways.
• Spatial parameters significantly modulate transcription factor activation.
• These ML models can be integrated into multiscale tumor growth models to reduce computational cost.