Searching perturbation dictionaries with RNA fingerprints

By cataloguing transcriptional responses to a variety of biological and chemical stimuli, single-cell perturbation dictionaries are aiding the mechanistic interpretation of large single-cell RNA sequencing datasets. However, matching a specific cellular state to a reference perturbation remains challenging. Writing in Cell, Grabski and colleagues describe RNA fingerprinting, a statistical framework that maps transcriptional responses from new experiments to dictionaries of known perturbations.

RNA fingerprinting uses a two-stage statistical approach. First, denoised perturbation signatures, or ‘fingerprints’, are learned from a reference dictionary. Query cells are then matched to these fingerprints using a Bayesian probabilistic framework inspired by methods developed for genome-wide association studies. Using ground-truth datasets, the authors show that the method accurately matches known perturbations within and across cell types, scaled to genome-wide datasets and inferred combinatorial perturbations.

This is a preview of subscription content, access via your institution

Prices may be subject to local taxes which are calculated during checkout