scROSHI: robust supervised hierarchical identification of single cells

Abstract

Identifying cell types based on expression profiles is a pillar of single cell analysis. Existing machine-learning methods identify predictive features from annotated training data, which are often not available in early-stage studies. This can lead to overfitting and inferior performance when applied to new data. To address these challenges we present scROSHI, which utilizes previously obtained cell type-specific gene lists and does not require training or the existence of annotated data. By respecting the hierarchical nature of cell type relationships and assigning cells consecutively to more specialized identities, excellent prediction performance is achieved. In a benchmark based on publicly available PBMC data sets, scROSHI outperforms competing methods when training data are limited or the diversity between experiments is large.

Detection of isoforms and genomic alterations by high-throughput full-length single-cell RNA sequencing in ovarian cancer

“…Subsequently, counts were normalized with sctransform 96 , regressing out cell cycle effects, library size, and sample effects as nonregularized dependent variables. Similar cells were grouped based on unsupervised clustering using Phenograph 97 , and automated cell-type classification was performed independently for each cell 98 using gene lists defining highly expressed genes in different cell types. Major cell-type marker lists were developed in-house based on unpublished datasets (manuscripts in preparation), including the Tumor Profiler Study 99 , using the Seurat FindMarkers method 100 .…”

De novo detection of somatic variants in long-read single-cell RNA sequencing data

“…Cells were annotated with scROSHI (Prummer et al 2023) using ovarian cancer marker gene lists. Marker genes are available at https://github.com/ETH-NEXUS/scAmpi_single_cell_RNA/blob/master/required_files/ovarian/celltype_list_ovarian.gm).…”