Harvard’s Program in Quantitative Genomics brought researchers together to discuss AI for modelling cells, improving cancer research and collaborating with scientists. The most striking prospect is a generative “virtual cell” that could let researchers simulate biological interventions before testing them in a lab.
Watch Desk analysis
What happened
The two-day conference, hosted on 29–30 October by Harvard T.H. Chan School of Public Health, featured about 15 experts and drew 130 participants, according to a report by the school’s news office. Mohamed bin Zayed University of Artificial Intelligence and Carnegie Mellon’s Eric Xing discussed using generative AI to create models of cells that simulate how cells or tissues might respond to interventions.
Stanford’s Sylvia Plevritis described combining multiscale modelling with multimodal AI, bringing together clinical records, imaging, pathology and molecular data to investigate cancer vulnerabilities and inform treatment. Harvard Medical School’s Marinka Zitnik discussed AI agents working with scientists on analysis, hypothesis generation and test design, including agents that could organise themselves into teams for longer-running research.
Why it matters
These are research directions, not evidence that virtual cells or autonomous research teams are ready to replace laboratory work. But they point to a shift from AI as a tool for processing results towards systems that may help decide what to investigate next.
The conference report also gives a useful reality check: Zitnik highlighted the need for suitable data and biological context, accurate agents and laboratory experiments alongside AI. In science, a convincing simulation is a lead to test, not a result to frame on the wall.
Our read
The most interesting part is the range: AI could help connect patient data, model biological systems and support the work of forming and testing hypotheses. That is a more ambitious scientific role than simply speeding up analysis, but the useful measure will be whether these approaches produce findings that hold up outside the model.
What to watch
- Whether virtual-cell models are tested against laboratory observations.
- Whether combining clinical, imaging and molecular data leads to measurable improvements in cancer research.
- How AI research agents are evaluated for accuracy, reproducibility and useful scientific discoveries.
Discussion spark: Should AI systems be allowed to propose and run scientific experiments with limited human oversight, or should researchers approve every step?
Sources and evidence
- Source update (5 November 2026, 18:04 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.