GTestimate: Improving relative gene expression estimation in scRNA-seq using the Good-Turing estimator

Single-cell RNA-seq suffers from unwanted technical variation between cells, caused by its complex experiments and shallow sequencing depths. We present GTestimate a new normalization method based on the Good-Turing estimator, which improves upon conventional methods by accounting for unobserved genes. We validate GTestimate using new ultra-deep sequencing data, generated via a novel cell targeted PCR-amplification approach, and show substantial improvements in cell-cell distance estimation and downstream results. A human cerebral-organoid scRNA-seq library was split into two aliquotes and sequenced; once as a ‘typical’ run and once after cell targeted PCR-amplification (cta-seq). This second run enabled a ‘ultra-deep’ sequencing depth for a small set of targeted cells. *************************************************************** Submitter statement concerning raw data: Raw sequencing data could not be included in this submission due to patient privacy concerns. Raw-data will be made available through controlled access at the European Genome-Phenome Archive at a later date. ***************************************************************