Unravelling Signalling Heterogeneity using DEEP Learning and MECHANIstic Modelling

Introduction

Signalling enables cells to respond to external cues, but the inherent heterogeneity of individual cell responses, essential for multicellular organization, complicates disease treatment. Heterogeneity arises from drivers at system and molecular scales, intertwined through feedback loops, making quantitative understanding and prediction challenging.

Objective

I will address this by pioneering transformative computational methods that predict phospho-signalling responses by integrating deep learning with mechanistic modelling to integrate systems and molecular scales.

Methodology

By using unbiased pattern recognition of deep learning models, I will learn cell states and simple phosphorylation rate laws. These will be combined with mechanistic models, integrating biological knowledge, to build simple and interpretable models that predict signalling responses from baseline omics profiles across distinct time-resolved and perturbational conditions.

Application

I will apply these methods to investigate drivers of heterogeneity in receptor tyrosine kinase (RTK) and rat sarcoma (RAS) signalling, in response to growth factors and targeted inhibitors in cancer cell lines. I will validate the approach by reprogramming patient-derived organoids using model-proposed inhibitor combinations.

Impact

The proposed research will advance our fundamental understanding of signalling regulation and co-regulation with cellular states. Given the vital role of RTK and RAS signalling in human health, it also holds the potential for translational impact.

Broader Implications

More broadly, the proposed computational methods are versatile and could be applied to a broad range of biological and non-biological systems.