GenPerturb: sequence-grounded interpretation of perturbation transcriptomes using pretrained genomic models

Abstract Background Perturb-seq captures transcriptional responses to thousands of genetic and chemical perturbations, but does not directly resolve the cis-regulatory elements or transcription factor motifs underlying those responses. Existing approaches rely on indirect post hoc analyses or external epigenomic annotations, making it difficult to connect gene-level responses to specific regulatory elements. Results We present GenPerturb, a model-based interpretation workflow that fits condition-specific linear readouts of pretrained sequence embeddings to observed Perturb-seq pseudobulk expression. Because control and perturbation readouts operate on one gene-centred sequence representation, control-subtracted attribution ranks candidate regions and motif instances as computational hypotheses associated with each fitted contrast. The fitted readouts recover much of the observed expression variation between genes, whereas recovery of variation across perturbation conditions within each gene is more modest. On the single available erythroid Perturb-multiome benchmark, region rankings showed agreement with accessibility labels comparable to annotation-based scores, with slightly closer agreement in the 1–10 and 10–100 kb strata, while the perturbed transcription factor’s motif family was recovered at comparable frequencies from attribution-derived and chromatin-derived regions. Selected lineage and compound examples illustrate motif associations that can differ from those of the corresponding transcription factor mRNA. Conclusions GenPerturb transfers regulatory priors learned during external pretraining to observed RNA-only Perturb-seq profiles, producing ranked, locus-anchored cis-regulatory hypotheses without a matched chromatin assay in the target experiment. Scoped to observed responses rather than prediction of unseen perturbations or cell types, the workflow offers computational candidates for experimental assessment of enhancer and transcription-factor activity.