scPred: Cell type prediction at single-cell resolution

Abstract

10Single-cell RNA sequencing has enabled the characterization of highly specific cell types in many human tissues, as well as both primary and stem cell-derived cell lines. An important facet of these studies is the ability to identify the transcriptional signatures that define a cell type or state. In theory, this information can be used to classify an unknown cell based on its transcriptional profile; and clearly, the ability to accurately predict a cell type and any 15 pathologic-related state will play a critical role in the early diagnosis of disease and decisions around the personalized treatment for patients. Here we present a new generalizable method (scPred ) for prediction of cell type(s), using a combination of unbiased feature selection from a reduced-dimension space, and machine-learning classification. scPred solves several problems associated with the identification of individual gene feature selection, and is able to capture 20 subtle effects of many genes, increasing the overall variance explained by the model, and correspondingly improving the prediction accuracy. We apply scPred to scRNA-seq data from pancreatic tissue, colorectal tumor biopsies, and circulating dendritic cells, and show that scPred is able to classify cell subtypes with an accuracy of 96.1-99.2% . Furthermore, we demonstrate that the feature selection step of scPred is able to discriminate from both tran-25 scriptional variation between single cell RNA-sequencing protocols, and between laboratory batch effects, and still predict cell subtype with an accuracy greater the 96%. Collectively, our results demonstrate the utility of scPred as a single cell prediction method that can be used for a wide variety of applications. The generalized method is implemented in software available here: https://github.com/IMB-Computational-Genomics-Lab/scPred/ 30

Single cell genomic characterization reveals the cellular reprogramming of the gastric tumor microenvironment

“…As an independent determination of gastric tumor versus normal epithelial cells, we employed a completely different method that uses a supervised machine learning algorithm. Called scPred, this approach eliminates the statistical inconsistencies seen with other single cell methods and thus provides highly accurate cell type assignment (10). We applied scPred on gastric epithelial cells derived from normal and tumor tissue.…”

Evaluation of methods to assign cell type labels to cell clusters from single-cell RNAsequencing data

“…In contrast, cell type labeling of cell clusters is still conducted manually by most researchers. This is in part due to a scarcity of reference cell type gene expression signatures and also because most methods to address this challenge are only available via web servers with limited number of cell types (Crow et al, 2018;Alquicira-Hernandez et al, 2018;Alavi et al, 2018), making it difficult for users to adapt them for their needs. In this study we used three scRNA-seq datasets to benchmark four methods that can address these challenges.…”

“…Methods explicitly developed to assign cell type labels to cell clusters of scRNA-seq data have been reported (Crow et al, 2018;Alquicira-Hernandez et al, 2018;Alavi et al, 2018). However, to our knowledge they are in beta, or implemented as web-servers to process cell types for which we could not find reference cell type annotations ( Figure 1F) that we would require to include in our evaluation.…”

MarkerCapsule: Explainable Single Cell Typing using Capsule Networks

“…For scPred, we use the R package as provided in the original paper 12 . We use function 'eigenDecompose' and 'getFeatureSpace' with default parameters (#PC n=10) on the integrated data and use the transformed…”