Generative Models for Therapeutic Discovery 2026
Learning molecules, targets and cell-state responses
University of Cambridge · Computer Laboratory
29–30 September 2026
Cambridge, United Kingdom
About the workshop
Recent advances in generative artificial intelligence are reshaping therapeutic discovery, moving beyond the design of individual molecules toward models that can connect chemical structure, biological targets, disease mechanisms and cellular responses. At the same time, the rapid growth of high-dimensional biomedical data — including multi-omics, perturbation screens, single-cell profiling, imaging and clinical datasets — creates new opportunities to learn representations of biological systems that are both predictive and mechanistically informative.
This workshop will bring together researchers working at the interface of machine learning, computational biology and drug discovery to discuss how generative and relational models can support the discovery of new therapeutic hypotheses. Topics will include molecular generation, target identification, perturbation-response modelling, cell-state prediction, multimodal data integration, foundation models for biology, and evaluation strategies for biomedical generative systems.
A central theme of the workshop is the shift from generating candidate molecules in isolation to learning structured relationships between molecules, targets, pathways, cell states and disease contexts. By connecting these levels, generative models may help prioritize interventions, predict therapeutic effects, and reveal mechanisms underlying treatment response. The workshop aims to foster discussion on methodological advances, practical applications and open challenges in building reliable, interpretable and biologically grounded models for therapeutic discovery.
Topics
Molecular generation
Generative models for molecular design, optimization and prioritization.
Targets and mechanisms
Learning relationships between targets, pathways, disease states and therapeutic hypotheses.
Cell-state responses
Perturbation-response modelling, single-cell profiles and cellular state prediction.
Multimodal biomedical data
Multi-omics, imaging, clinical data and foundation models for biology.
Evaluation and validation
Benchmarking, interpretability, biological grounding and reproducibility.
Relational modelling
Structured models connecting molecules, targets, cells, pathways and patients.
Key information
| Item | Details |
|---|---|
| Dates | 29–30 September 2026 |
| Location | University of Cambridge, Computer Laboratory |
| Format | Two-day in-person workshop |
| Registration | Opening soon |
| Speakers | To be announced |
Prof. Pietro Liò
University of Cambridge, Cambridge, UK
Dr. Dario Righelli
University of Cambridge, Cambridge, UK
University of Padova, Padua, Italy
Prof. Cristian Taccioli
University of Cambridge, Cambridge, UK
University of Padova, Padua, Italy