ZURICH 14 July 2022. In a landmark study published in the journal Science, Lucas Pelkmans and members of his group reveal how the signaling activity of individual human cells can be accurately predicted by the subcellular, cellular, and multicellular state of these cells using machine learning approaches. This variability was long thought to be largely stochastic or unpredictable, and demonstrates that spatial context across scales contains a large amount of hidden information to predict single-cell activities. The study also shows that spatial context across scales determines the highly heterogeneous responses of cells to treatment with two drugs used in clinical trials, namely Avutometinib (a first-in-class dual MEK/RAF inhibitor that allosterically inhibits BRAFV600E, CRAF, MEK, and BRAF) and MK-2206 (an orally active allosteric AKT inhibitor). This enabled for the first time a data-driven prediction of the whole spectrum of single-cell responses to these two drugs.
