-RNA fingerprinting is a statistical framework for scRNA-seq data that maps transcriptional responses observed in new experiments to perturbation dictionaries. This takes place in two main steps. First, fingerprints (i.e. denoised representations of perturbation effects) are estimated for each reference perturbation in the dictionary using a multi-condition latent factor model. Second, queries are mapped to these fingerprints using a Bayesian regression framework. Queries can be mapped either at the level of individual cells or as samples (i.e. groups of cells with the same label). Further details on the statistical methodology and biological applications are in our preprint.
0 commit comments