Projects available

Interpretable Multi-Scale Integration Framework for Biological Systems: Towards a Virtual Cell

Application Deadline: August 17th 2025

Background & Significance 

Recent advances in machine learning and foundation models have revolutionised biological research, enabling the development of models that can generalise learned representations of cells in various states. These models have shown promise in areas such as protein folding, protein interactions, gene regulation, gene expression across cell types and disease states, and cellular phenotypes. However, most existing models operate in isolation, lack cross-modality integration, and are often opaque boxes—limiting interpretability and trust in their predictions.

Biological systems are inherently multi-scale, involving complex temporal and spatial dependencies and feedback loops. To advance our understanding and enable actionable insights—especially in clinical research—there is a critical need for interpretable, integrative models that connect molecular, genetic, and cellular levels. This project aims to address these challenges by developing and validating frameworks that integrate predictions across biological scales and provide mechanistic insights into model decision-making.

Research Objectives and Approach

The project will proceed through the following key steps:

Framework Development: Interpretable Multi-Scale Integration

• Design an architecture to integrate predictions from diverse biological models (e.g., molecular, genetic, cellular) within a foundation model.
• Implement interpretability techniques to clarify model decisions.
• Validate biological plausibility of interpretability output and develop strategies to resolve conflicting predictions.
• Identify and curate high-quality, multi-scale datasets suitable for model training and validation.
• Establish mechanisms for bidirectional information exchange between biological scales and quantify each model’s contribution.

Prototype Development: Interpretable Virtual Cell Framework

• Build a unified, interpretable simulation platform that integrates multi-scale models at the cellular level.
• Validate the virtual cell framework using both experimental and synthetic datasets.
• Leverage mechanistic insights from the framework to generate and prioritize new biological hypotheses for experimental testing.
• Collaborate with experimentalists to iteratively refine the framework based on empirical feedback.

Collaboration and Training Environment

This project will be a collaborative effort between Queen Mary University of London, a top-ranked univerisity for research and MSD (Merck Sharp & Dohme), world’s top-five pharmaceutical company. Students will benefit from exposure to both academic and industry research environments, gaining a comprehensive perspective on the challenges and opportunities in computational biology from both fundamental and translational viewpoints. This unique partnership will provide valuable experience in applying advanced computational methods to real-world biomedical problems, enhancing candidate’s scientific and professional prospects.

Impact 

This research will advance computational biology and AI interpretability by: 

• Providing robust tools for understanding complex biological systems across multiple scales.
• Enabling the discovery of novel biological mechanisms through interpretable, integrated models.
• Enhancing trust, transparency, and reliability in AI-driven biological research.
• Laying the groundwork for a new generation of AI models that bridge predictive power and mechanistic insight, transforming biology into a more predictive and actionable science.

This project is in conjunction with MSD.

Eligibility and Applying 

This project is part of the UKRI/BBSRC AI for Drug Discovery Progamme, and successful candidates will join a cohort of students working on complementary projects in the AI for Drug Discovery space.

We are looking for highly motivated individuals who are passionate about contributing to new discoveries in drug discovery bioscience through the application of the latest techniques in AI and data science. Ideal candidates will have a grounding in both a natural science and data science, e.g. through a Master's degree or work experience in a subject such as bioinformatics or computational chemistry. Alternatively, you may have, for example, a first-class degree in computer science followed by biochemistry experience, or vice versa (qualifications and evidence thereof must be obtained before October 2025). You will be confident in performing data wrangling and analysis in a language such as Python, R or C++. Effective communication skills are essential.

We particularly encourage students from groups that are currently underrepresented in postgraduate science research, including black and minority ethnic students and those from a socio-economically disadvantaged background.

The Studentship will cover UK tuition fees, UKRI stipend (currently £21,637) and a consumable allowance for a period of 4-years (pro-rata for part-time), and is open to candidates who meet the UKRI eligibility criteria. This typically means the candidate will have unrestricted access on how long they can remain in the UK (i.e. are a British National, have settled, or pre-settled status, have indefinite leave to remain etc.) and have been living in the UK for the 3 years immediately prior to studentship starting. Candidates who would be classed as International are unfortunately not eligible for this opportunity.

Visit the Apply Pages for further details on how to submit an application. 

Supervisors:  

Professor Venet Osmani - Professor of Clinical AI and Machine Learning, Digital Environment Research Institute, QMUL

Dr Kirill Shkura - Systems Biologist and Data Scientist, MSD

Soumya Ghosh - Director of Machine Learning, MSD

Project Partner: MSD