Stable-Shift fits a low-rank transcriptional response basis from training perturbations and uses graph convolution on biological context (STRING, network, expression stats, GO) to predict coordinates for unseen genes, reaching 0.592 cosine similarity on K562 Perturb-seq data versus 0.569 for GEARS.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
q-bio.GN 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Stable-Shift: Biologically Structured Prediction of Transcriptional Responses to Unseen Gene Perturbations
Stable-Shift fits a low-rank transcriptional response basis from training perturbations and uses graph convolution on biological context (STRING, network, expression stats, GO) to predict coordinates for unseen genes, reaching 0.592 cosine similarity on K562 Perturb-seq data versus 0.569 for GEARS.