Mechanics-informed graph neural operators for multimodal spatial decoding of mechanochemical pathways in arterial remodelling
Code: BBSRC-DFA_2026_19
Primary Supervisor: Prof Thomas Iskratsch
Email: t.iskratsch@qmul.ac.uk
Institute: School of Engineering and Materials Science
Secondary Supervisor: Dr Sathiskumar Anusuya Ponnusami
Email: s.a.ponnusami@qmul.ac.uk
Institute: School of Engineering and Materials Science
Abstract:
Vascular smooth muscle cells (VSMCs) take a central role in the regulation of vascular tone, but also the ageing associated changes to arterial composition, structure and function. In the normal arterial wall, VSMCs are contractile. However, in response to changing chemical and mechanical signals they change their phenotype, leading to a deregulation of cellular functions. Our in vitro experiments demonstrated that mechanical forces including extra cellular matrix (ECM) stiffness and age-associated increases in blood pressure alone are sufficient to stimulate the phenotypic conversion. To investigate the process in vivo, we collected tissues from animal models of vascular ageing and analyse these using mass-spectrometry imaging for lipidomic and proteomic analysis (both with 10µm spatial resolution). These data sets will be combined with spatial mechanical information from Brillouin microscopy (localised stiffness and viscosity with sub-micrometer resolution), Raman Spectroscopy (chemical changes) and correlative fluorescence staining for different biomarkers.
This PhD project will use AI to drive the integration and interrogation of these multi-modal spatial datasets to investigate the pathways that are connecting the different mechanical stimuli (especially localised extracellular matrix stiffness) to an ageing cell phenotype and vice versa, the localised lipid and proteomic changes to the alteration of extracellular and intracellular mechanics. The outcome will be a fully integrated data set that can be used for analysis of arterial ageing dependent mechanotransduction pathways and future hypothesis generation. It will further develop novel mechanics-regularised neural operator architectures for learning structured relationships between spatial mechanical and molecular fields and lastly quantify bidirectional mechanochemical coupling and identify mechanosensitive pathways underlying VSMC phenotype transitions.
Lay Summary:
This research project investigates how Vascular Smooth Muscle Cells (VSMCs)—the cells responsible for controlling blood flow—transform from healthy, functional components into alternative phenotypes during aging. While these cells normally maintain arterial health, age-associated physical stresses like high blood pressure and increased artery stiffness trigger a shift in their behavior. This project aims to decode the relationship between a cell's physical environment and its internal chemistry. To achieve this, the project will make use of spatial data sets that investigate the molecular, chemical and mechanical properties in vivo and have or will be collected in a currently running study. Because these datasets are incredibly complex, the project employs Artificial Intelligence to integrate this "multi-modal" information. The AI will bridge the gap between mechanical forces and biological changes, identifying the specific pathways that contribute to cellular changes in arterial ageing. Ultimately, this work will produce a comprehensive digital map of how cells sense and respond to physical stress and especially develop new tools for analysis of such multi-dimensional data sets.
Aims and Objectives:
This PhD project will develop a novel mechanics-informed AI framework for multimodal spatial field learning to elucidate mechanochemical pathways driving vascular smooth muscle cell (VSMC) phenotypic switching during arterial ageing. We have collected multi-modal spatial data sets (mechanical mapping, Raman spectroscopy data, immunostaining, mass-spectrometry imaging for spatial lipidomics). The objective is the spatiotemporal integration of and learning on these datasets to identify new mechanosensing pathways and mechanisms.
This will be achieved through the following specific aims:
Construct a spatially aligned mechanochemical atlas of ageing arterial tissue by integrating mechanical, chemical and molecular imaging modalities.
Develop novel mechanics-regularised neural operator architectures for learning structured relationships between spatial mechanical and molecular fields.
Quantify bidirectional mechanochemical coupling and identify mechanosensitive pathways underlying VSMC phenotype transitions.