A stemness score for single-cell RNA sequencing data using a protein-protein network as a scaffold

DANIELA SENRA; NARA GUISONI; LUIS DIAMBRA

virtual

Conferencia; 3nd WBDS-LA (2nd Women in Bioinformatics & Data Science LA); 2022

Women in Bioinformatics & Data Science LA

Single-cell RNA sequencing (scRNA-seq) is a powerful, high-resolution technique used tostudy cellular heterogeneity at the transcriptomic level. It has been successful in discoveringnovel cell types in cancer and analyzing lineages in the context of embryonic development.A common application of scRNA-seq data is trajectory inference. However, it is oftennecessary to previously identify the origin of the trajectories, namely the stem or progenitorcells. Typically, previously known stemness markers are used to determine the root cell, butit is not always feasible due to the high drop-out rate of the scRNA-seq technique. Moreover,stemness markers depend on the tissue and the developmental stage. A few algorithmshave also been proposed to determine stem cells, for example LandSCENT, anentropy-based algorithm. Another program developed to this end is CytoTRACE, whichhypothesizes that gene counts are generally correlated with the state of differentiation.Here, we developed a novel open-source computational tool to quantify pluripotency fromsingle cell transcriptomics data that does not need prior knowledge of stem markers. Theapproach uses the protein-protein interaction (PPI) network associated with thedifferentiation process as a scaffold and the gene expression matrix to calculate a score thatwe call differentiation activity. This score reflects how active the differentiation network is ineach cell. We benchmark the performance of our algorithm with LandSCENT andCytoTRACE, for four healthy human data sets: breast, colon, hematopoietic and lung. Weshow that our algorithm is more efficient than LandSCENT and requires less RAM memorythan the other programs. We also illustrate a complete workflow from the count matrix totrajectory inference using the breast data set.