🤖 AI Summary
This study addresses the challenge of reverse-identifying molecular targets and compounds from single-cell transcriptional perturbation responses. The authors propose the first multi-task Transformer-based retrieval model tailored for single-cell perturbation data, which jointly learns target prediction and molecular embedding within a fixed compound library. The model performs end-to-end inverse inference using differential expression profiles relative to cell-type-specific DMSO controls and incorporates a structure–transcriptome alignment constraint to enhance representational consistency. Evaluated on the Tahoe-100M dataset, the model achieves a target Recall@10 of 0.408 and a compound Hit@1 of 0.129, significantly outperforming baseline methods and demonstrating its effectiveness in retrieving known perturbation pairs.
📝 Abstract
Large-scale single-cell perturbation atlases make it possible to ask an inverse question: given an observed transcriptional response, which annotated targets and compounds in a fixed library are most consistent with that response? We present \model, a Transformer retrieval model for this closed-library setting. Each input is a cell-level perturbation signature formed by contrasting one treated cell with a cell-line-specific mean DMSO reference. The encoder maps the signature to a target-retrieval vector and a molecular-embedding vector, trained jointly with supervised target losses and structure--transcriptome alignment. We evaluate on Tahoe-100M conditions with mapped target annotations using a within-compound stratified 90/10 condition-pair split of 10,505 training and 1,168 validation drug--cell-line pairs. Because compounds and cell lines can occur in both partitions, the experiment measures held-out condition-pair retrieval rather than generalization to unseen compounds or cellular contexts. In a Monte Carlo evaluation over 38,400 sampled validation cells, \model\ achieved target Recall@10 of 0.408 and Recall@20 of 0.544, together with compound Hit@1 of 0.129, Hit@10 of 0.343, and mean reciprocal rank of 0.205 over a 379-compound bank. A separate diagnostic evaluation produced nearly identical values for the main model and large gains over a random-vector control and post-hoc bag-of-genes controls. These results demonstrate that a single multi-task model can recover both mapped target annotations and recorded compound identities from observed cell-level responses in the evaluated Tahoe-100M closed-library setting. Generalization to unseen compounds and cellular contexts remains to be established.