Background

 Single-cell clustering programs have been integral in revealing how gene expression data are distributed amongst clusters and which cellular features are differentially expressed or accessible between them. Traditionally, current processing programs make use of algorithms that categorize single cells into clusters based on principal component analysis and k-means clustering, but these systems provide limited identification capabilities and are unable to annotate individual cells within a cluster based on co-variances. To solve this issue, BCM Investigators have developed a novel single-cell annotation pipeline, single-cell rule association mining (SCRAM) capable of identifying co-occurring cell identities.

Technology Overview

This methodology utilizes an a priori algorithm combined with a market-basket analysis that can assign highly-refined cell identities to over 90% of cells present in human and mouse cancer scRNAseq and scATACseq datasets. Our method uses co-occurring gene markers for over 150 manually-curated cell types to annotate each individual cell. Additionally, our software applies copy number variation (CNV) calling is to further support identification of tumor cells and provides mutational profiling using single nucleotide variants (SNVs). The accuracy of this pipeline has been tested and validated using previously published datasets.

Stage of Development

While originally created and validated for glioblastoma studies, this methodology can easily be annotated for other tumor or organ types using similar datasets. This version of SCRAM has been compared to the output of other tools which were found to be more general and less specific to cell types. SCRAM offers accurate and specific single-cell data, which is a highly valuable and in-demand tool for clients seeking biotechnology services. SCRAM can be easily adoptable into current service offerings. Companies that provide scRNA (single cell RNA) sequencing analysis services would be able to incorporate SRCAM into their analysis tools to include single cell identification in the data supplied to their clients.

Benefits

The SCRAM program is ideal for identifying and classifying co-occurring genes and cell-type markers to assist in high-throughput screening of putative transdifferentiating cells within tumors and heterogeneous cell populations.

Able to extract co-occurring gene sets and co-occurring cell type markers, a current limitation of other differential gene expression analysis methods for single cell clusters

Accurately differentiates tumor from host cells

Can characterize each individual tumor cell using 3 modular components: (1) marker-expression modeling, (2) CNV-based scoring, (3) expressed SNV-based scoring.

Provides annotation of individual cells within clusters, including CNV and SNV

Easily expandable for use with any cancer

Applications

• Potential use for identifying candidate gene lists for drug screens
• Use for immunotherapy to identify tumor-specific antigens 

Opportunity

Available for non‑exclusive license