Meta-Bayesian Optimisation for Cell Culture Media Development

Multus Biotechnology uses Bayesian optimisation through MediOP to optimise diverse cell culture campaigns. However, each campaign starts with limited use of historical data. Differences between campaigns prevent simple data pooling, making effective knowledge transfer a key challenge.

The research problem is how to turn these small, heterogeneous and partially overlapping campaigns into a transferable, uncertainty-aware prior that makes each new campaign faster, cheaper and more likely to reach a better optimum, while remaining interpretable to scientists.

Five research questions will be addressed. RQ1 - Representation: how should ingredients, cell lines and objectives be represented so a surrogate can be shared across different input and output sets, and can mechanism-inspired features improve transfer over learned embeddings? RQ2 - Meta-learned surrogates: can a model trained across the portfolio provide calibrated uncertainty for a new cell line after few observations and reduce the rounds needed to reach target performance? RQ3 - Design-space inference: can the model recommend which ingredients and concentration bounds to use, and how does this compare with expert-specified spaces? RQ4 - Portfolio-level acquisition: how should experimental selection balance immediate improvement in the current campaign against information that benefits future campaigns, under cost and laboratory-throughput constraints? RQ5 - Human in the loop: how should recommendations and uncertainty be presented so scientists can interrogate, override and refine them, and can this improve outcomes?

The proposed approach frames media development as multi-task Bayesian optimisation across heterogeneous, partially overlapping tasks. A probabilistic surrogate will be meta-learned from Multus's portfolio and used as the prior for each new campaign. Work package 1 will create a canonical ingredient vocabulary, cell-line descriptors and harmonised readouts; establish single-task, warm-start and transfer baselines; and define leave-one-campaign-out evaluation. Work package 2 will investigate deep-kernel, ensemble and hierarchical surrogates that adapt as observations arrive before live evaluation.

Work package 3 will infer design spaces by recommending ingredients and bounds, extend learned-search-space methods to partially overlapping variables, and handle different objective sets through multi-output surrogates with missing outputs and multi-objective acquisition under cost constraints. Work package 4 will develop acquisition functions that value information for current and future campaigns and study how scientists use, challenge and improve recommendations.

Evaluation will combine retrospective leave-one-campaign-out experiments with prospective validation on live campaigns, where the new pipeline will run alongside Multus's current process on the same laboratory schedule. Success will mean fewer rounds to reach the same performance, a better optimum for the same budget, calibrated uncertainty, and recommendations judged clear and interpretable by scientists. The methods could also apply to portfolios of costly experiments in formulation, catalysis, materials and process development.

Desirable Student Background:

An MSc or strong first degree in machine learning, statistics, computer science, engineering or a quantitative life-science subject. The student should be comfortable with probabilistic modelling, including Gaussian processes and Bayesian inference, and with Python. An interest in biology is welcome, but wet-lab experience is not required.

How to apply:

Please apply under the University of Manchester application portal and select 'PhD in Artificial Intelligence' (OAA Applicant Portal).

When applying, you'll need to specify the full name of this project, the name of your supervisor, if you already having funding or if you wish to be considered for available funding through the university, details of your previous study, and names and contact details of two referees.

Your application will not be processed without all of the required documents submitted at the time of application, and we cannot accept responsibility for late or missed deadlines. Incomplete applications will not be considered.

You will be asked to upload the following supporting documents:

• Final Transcript and certificates of all awarded university level qualifications
• Interim Transcript of any university level qualifications in progress
• CV
• Supporting statement: A one or two page statement outlining your motivation to pursue postgraduate research and why you want to undertake postgraduate research at Manchester, any relevant research or work experience, the key findings of your previous research experience, and techniques and skills you've developed. (This is mandatory for all applicants and the application will be put on hold without it).
• Contact details for two referees (Please ensure referee email addresses are official university or workplace accounts, as we may need to verify the references.)
• English Language Certificate (if applicable)

We actively encourage applicants from diverse career paths and backgrounds and from all sections of the community, regardless of age, disability, ethnicity, gender, gender expression, sexual orientation and transgender status.

We also support applications from those returning from a career break or other roles. We are dedicated to supporting work-life balance and offer flexible working arrangements to accommodate individual needs. Our selection process is free from bias, and we are committed to ensuring fair and equal opportunities for all applicants.

We are dedicated to supporting work-life balance and offer flexible working arrangements to accommodate individual needs, including a part-time study option.

Join us and help grow the UK's national capability in artificial intelligence, part of the UK's modern industrial strategy. Everyone is welcome if you are passionate about uncertainty quantification, decision making with humans in the loop and decision making for machine learning systems and their applications in Science and Engineering, and looking to develop your research and innovation skills at the doctoral level to further your career in the AI sector.

Equality, diversity and inclusion is fundamental to the success of The University of Manchester, and is at the heart of all of our activities. We know that diversity strengthens our research community, leading to enhanced research creativity, productivity and quality, and societal and economic impact. (Equality, diversity and inclusion | The University of Manchester)

Funding Notes

Fully funded AI UKRI CDT 4 year program; The Project will be based in University of Manchester.

The UKRI AI CDT in Decision Making for Complex Systems is part of the TechFirst offering stipends of £30,000 for students eligible for home fee status. The start date is September 2026.

This project is with Multus.

References

1. Hashizume, T. and Ying, B.-W. (2024). Challenges in developing cell culture media using machine learning. Biotechnology Advances, 70.
2. Hashizume, T., Ozawa, Y. and Ying, B.-W. (2023). Employing active learning in the optimization of culture medium for mammalian cells. npj Systems Biology and Applications, 9, 20. https://doi.org/10.1038/s41540-023-00284-7
3. Garnett, R. (2023). Bayesian Optimization. Cambridge University Press.
4. Swersky, K., Snoek, J. and Adams, R. P. (2013). Multi-task Bayesian optimization. NeurIPS 26.
5. Bonilla, E. V., Chai, K. M. and Williams, C. K. I. (2008). Multi-task Gaussian process prediction. NeurIPS 20.
6. Feurer, M., Letham, B., Hutter, F. and Bakshy, E. (2018). Practical transfer learning for Bayesian optimization. arXiv:1802.02219.
7. Wistuba, M. and Grabocka, J. (2021). Few-shot Bayesian optimization with deep kernel surrogates. ICLR.
8. Li, Y. L., Rudner, T. G. J. and Wilson, A. G. (2023). A study of Bayesian neural network surrogates for Bayesian optimization. arXiv:2305.20028; ICLR 2024.
9. Perrone, V. et al. (2019). Learning search spaces for Bayesian optimization. NeurIPS 32.
10. Volpp, M. et al. (2020). Meta-learning acquisition functions for transfer learning in Bayesian optimization. ICLR.
11. Liu, T., Astorga, N., Seedat, N. and van der Schaar, M. (2024). Large language models to enhance Bayesian optimization (LLAMBO). ICLR.