Room: Kilburn Building, Lecture Theatre 1.1

9:30-10:00 Arrivals (includes coffee)

10:00-10:10 Welcome, 

10:10-11:10 Experimental Design in the Age of GenAI Henry Moss, University of Lancaster

11:10-12:10 Accelerating Chemical Discovery with Knowledge Guided Bayesian Optimisation Xenofon Evangelopoulos, University of Liverpool

12:10-13:30 Lunch

13:30-14:30 Global Acquisition Optimization for Structured Graph Bayesian Optimization Jixiang Qing, Imperial College London

14:30-15:00 Tea Break

15:00-16:00 Bayesian experimental design for antibody discovery Alexandra Gessner, AstraZeneca

Abstracts

Experimental Design in the Age of GenAI

Abstract

Generative AI stands to revolutionize how experiments are conceived, orchestrated, and iterated. This talk outlines a general framework that can harness generative AI within experimental design loops, then examines the widely adopted approach of Latent Space Bayesian Optimization. Our proposed methodology diverges from conventional frameworks by decoupling the discriminative (surrogate) and generative models, rather than tightly integrating Gaussian Processes with Variational Autoencoders. Through independent training and a straightforward Bayesian update, we enable an efficient sampling strategy that identifies candidate structures with high scores.

Accelerating Chemical Discovery with Knowledge Guided Bayesian Optimisation

Abstract

Tackling global challenges like sustainability and antimicrobial resistance requires faster discovery of new materials and drugs. However, this process is often slow due to complex experimentation and decision making. Bayesian optimisation (BO) offers a sample efficient approach, but struggles with the non smooth, high dimensional nature of scientific design spaces. In this talk, I will show how integrating domain knowledge into BO through ML enriched multi model optimisation and hypothesis driven constraints can reshape the search landscape and accelerate discovery. I will present recent developments in human and LLM in the loop optimisation and conclude with key challenges and future directions in knowledge guided scientific discovery.

Global Acquisition Optimization for Structured Graph Bayesian Optimization

Abstract

Bayesian optimization over graph structured domains has emerged as a powerful paradigm for tasks such as molecular design and neural architecture search (NAS). While noticeable progress has been made on constructing expressive graph based Gaussian process (GP) surrogates, acquisition function optimization central to BO performance remains underexplored in discrete, combinatorial graph spaces, particularly under structural constraints. This talk addresses the challenge of global acquisition optimization in such settings. We present recent work that encodes both the graph search space and shortest path graph kernels via mixed integer programming (MIP), enabling acquisition functions such as LCB to be globally optimized. Empirical results on molecular design and NAS benchmarks demonstrate the effectiveness of principled acquisition optimization in graph structured BO.

Bayesian experimental design for antibody discovery 

Abstract

Experimentation in antibody discovery is inherently slow and costly, making in silico (computational) strategies essential for designing wet lab experiments. Recent breakthroughs in AI, particularly the rise of structure prediction models, have rapidly gained widespread attention and are fundamentally transforming computational protein design. Yet, to fully leverage the growing arsenal of computational oracles, efficient decision making is critical when selecting antibody candidates for experimental validation. Bayesian experimental design offers a principled and robust framework for navigating the vast combinatorial space of sequence mutations while balancing key properties, such as binding affinity, humanness, and developability. We will highlight how multi objective Bayesian optimization leverages diverse in silico oracles for simultaneous optimization of multiple antibody characteristics. This approach bridges deep learning advances with Bayesian models to accelerate the next generation of biotherapeutics.