Stable-Shift: Biologically Structured Prediction of Transcriptional Responses to Unseen Gene Perturbations

2026-06-22

AI Summary

This study addresses the high cost of functional genomics experiments by proposing a novel method to predict transcriptional responses of genes not perturbed during training. The approach aggregates single-cell perturbation data into perturbation-level expression shifts and constructs a low-rank response basis from observed perturbations. It uniquely integrates multiple biological priors—including STRING protein–protein interactions, Gene Ontology annotations, network topology, and control expression statistics—into a graph convolutional network to predict coordinates of unseen genes within this basis. By combining low-rank matrix factorization with residualized expression modeling, the method achieves a cosine similarity of 0.592 on the K562 Perturb-seq benchmark, significantly outperforming existing approaches such as GEARS and demonstrating consistent superiority across multiple datasets and evaluation metrics.

Abstract

Predicting transcriptional responses to genetic perturbations could reduce the experimental burden of functional genomics, but extrapolation to genes that were never perturbed during training remains difficult. We present Stable-Shift, a structured method for estimating unseen-gene responses. Stable-Shift aggregates single-cell measurements into perturbation-level expression shifts, fits a low-rank response basis using training perturbations only, and predicts an unseen gene's coordinates in that basis from biological context. The context combines STRING interactions, network structure, control-cell expression statistics, and Gene Ontology annotations; the evaluated implementation uses graph convolution to integrate these inputs. On the supplied K562 Perturb-seq benchmark, Stable-Shift obtained 0.592 cosine similarity, compared with 0.569 for GEARS, together with higher Spearman correlation and top-gene precision among the evaluated methods. Its mean cosine similarity over five unseen-gene splits was 0.589 +/- 0.008. The same ordering was observed in the supplied graph-aware, residualized, gene-space, and Norman-dataset comparisons. These results support further study of biologically structured latent-response prediction, while the lower gene-space accuracy and sensitivity to sparse graph neighborhoods limit the scope of the present conclusions.

Problem

• transcriptional response
• unseen gene perturbations
• functional genomics
• extrapolation
• gene perturbation prediction

Innovation

• Stable-Shift
• transcriptional response prediction
• unseen gene perturbation
• biologically structured latent space
• graph convolution