unCTC: Unbiased characterisation of single circulating tumor cell transcriptomes II

The identification and characterisation of Circulating Tumour Cells (CTCs) is important to get insights into the biology of metastatic cancers, monitoring disease progression, and potential use in liquid biopsy-based personalised cancer treatment. The major limitation of CTC isolation is due to their heterogeneous nature, altered phenotype from primary tumour and availability in limited numbers. In the past years, several techniques have been developed to detect CTCs from peripheral blood, based on canonical markers. These methods are, however, prone to miss a larger set of CTCs due to variable or no expression of these markers. Furthermore, CTC enrichment processes are not free from White Blood Cell (WBC) contamination. Single cell RNA sequencing (scRNA-Seq) of CTCs provides a wealth of information about their tumors of origin as well as their fate. The first and most important roadblock encountered in CTC scRNA-Seq data analysis is confirmation of CTC capture. We present unCTC, an R software for unsupervised and semi-supervised characterisation of CTC transcriptomes, in contrast with WBCs. unCTC features various standard and novel computational/statistical modules for clustering, Copy Number Variation (CNV) inference, and marker based verification of the malignant phenotypes. Notably, we propose Deep Dictionary Learning using K-means clustering cost (DDLK) that robustly clusters scRNA-Seq profiles of CTCs and WBC contaminates to characterise heterogeneity among the concerned cell population. Interestingly, DDLK performs better as gene expression data is transformed into pathway enrichment profiles. To demonstrate the utility of unCTC, we produce scRNA-Seq profiles of breast CTCs enriched by the integrated ClearCell® FX/PolarisTM workflow that works on the principles of size selection and negative enrichment for CD45, a pan leukocyte marker. Overall design: 96 CD45+ live cells were captured using Fluidgm C1 single cell system and processed for whole transcriptome library preparation.