SaritaPoonia/unCTC: Characterising single circulating tumor cell transcriptomes

Unbiased identification and characterization of single CTC transcriptomes are aided by the unCTC process, which includes multiple techniques. Clustering of scRNA-seq profiles is a crucial part of this process. We describe a robust method for grouping single-cell transcriptomes in a meta-space that spans pathways and computes enrichment scores based on gene expression readouts. The log-transformed TPM (Transcripts Per Million) matrix/count associated with CTC transcriptomes is transformed into a matrix comprising pathway enrichment scores obtained using an R package, GSVA, for unsupervised clustering. Deep dictionary learning with k-means clustering cost (DDLK) is applied to these pathway scores. The K-means clustering cost is included in the deep dictionary learning (DDL) architecture by DDLK. With the use of Stouffer's Z-score, unCTC allows for the determination of a variety of canonical markers identifying malignant/epithelial/immune origins (Stouffer et al., 1949). In our opinion, gene-set-based techniques are effective in enhancing single-marker-based and inferred-CNV-based cell-group characterization.

install.packages("remotes")
remotes::install_github("SaritaPoonia/unCTC")