RevPert: ranking candidate drivers of transcriptomic state transitions via gallery-native reverse perturbation
Cellular state transitions underlie adaptation, ageing and disease, yet prioritizing catalogued genetic perturbations whose expression signatures match an observed transcriptomic shift remains difficult. Most models predict phenotype from a nominated intervention, whereas genetic inverse benchmarks are largely restricted to within-screen identity recovery. Here we introduce RevPert, a…
The article introduces RevPert, a reverse perturbation model designed to rank genetic candidate drivers of transcriptomic state transitions. This model combines signed Pearson connectivity with a learned residual to prioritize catalogued genetic perturbations based on their expression signatures, matching observed transcriptomic shifts.
RevPert outperformed matched baselines in Replogle Essential Perturb-seq (for four lines) and LINCS-KO screens (for ten lines). When applied to public drug-resistance contrasts in hepatocellular carcinoma (HCC) and chronic myeloid leukemia (CML), RevPert's dual-arm ranking placed pre-specified disease anchors higher on the expected arms compared to ranking the same signatures by differential-expression magnitude alone.
This model integrates within-screen reverse ranking with a screen-external signed-geometry check, which calibrates literature anchors and is not meant for held-out recovery.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
