#33 · Single-cell

Molecular profiles are being linked to individual clinical outcome predictions

The common thread is linking high-dimensional molecular measurements to individual clinical endpoints. irAEGIS predicts immune-related adverse event risk from pre-treatment scRNA-seq using a pathway-guided, interpretable hierarchical model. MANTIS applies an artificial neural network to genome-wide DNA methylation from 2,786 patients to predict time-dependent survival and treatment effects in childhood brain tumors. ImmuneTwins personalizes a shared single-cell virtual-cell model per individual to forecast immune trajectories. CALM, by contrast, is an integrated single-cell and bulk RNA-seq atlas of AML with clinical annotations and a T-cell signature, and is primarily descriptive rather than predictive. The new element relative to descriptive profiling is explicit individual-level prediction of response, toxicity, and survival, though the degree of predictive validation varies across the set.