Algorithm predicts how cells respond to interventions
What happens inside a cell when you switch off a gene or add a compound? Until now, you had to run the experiment to find out.
Simulating the full complexity of a living cell on a computer remains far beyond current capability. But researchers at Altos Labs have developed a machine learning algorithm that predicts gene expression, the level of activity of individual genes in a cell, following specific interventions. The work was published as a preprint on arXiv.
The system functions as an end-to-end predictive model. You specify an intervention, and the algorithm outputs a predicted gene activity profile for the cell. That has practical value for testing hypotheses about aging, disease, and potential treatments without running every experiment in the lab. The researchers describe this as a meaningful advance over existing approaches to biological simulation.
Why this is harder than protein folding
Models like AlphaFold solve a single well-defined problem: predicting protein structure from sequence. Predicting cellular responses to interventions is far more complex. A cell contains thousands of genes interacting through dense regulatory networks. The Altos Labs algorithm focuses specifically on causal prediction across this space, trained on large datasets of known cellular responses to perturbations.
The model’s ability to generalise to new, untested interventions is the critical question. The preprint reports promising accuracy, but independent replication is needed before these results can be considered established.
What it means for longevity research
Many longevity interventions work by altering gene expression patterns. A reliable predictive model could help identify the most promising candidates faster, before testing them in living organisms. That shortens the cycle from hypothesis to experimental validation. For a field constrained by the cost and time of animal studies, this kind of computational tool could be genuinely useful, provided its predictions hold up under scrutiny.