Characterization of Tumor Stromal Cells at the Single-Cell Level Using Machine Learning

Share this content:
Researchers describe SuperCT, a predictive framework for the identification of tumor stromal cell type based on scRNA-seq data.
Researchers describe SuperCT, a predictive framework for the identification of tumor stromal cell type based on scRNA-seq data.

Following single-cell RNA sequencing (scRNA-seq) of tumor stromal cells, an artificial intelligence-based tool called SuperCT was able to correctly tease out many of the different cell types that were present within the samples, according to a study presented at the 2018 AACR Pancreatic Cancer: Advances in Science and Clinical Care conference in Boston, Massachusetts. This method was found to be as reliable as manual labeling approaches using canonical markers, according to the researchers.

Previous attempts to characterize distinct cell types following scRNA-seq have occurred, but those analytic pipelines use “unsupervised” clustering algorithms to group and visualize cells, so they can be less precise in identifying differences at the single-cell level.

Researchers described a new method of cell typing through the use of a proprietary, supervised framework. In a study, the investigators first trained an artificial-neural-network model using several well-annotated public datasets, such as the mouse cell atlas (MCA) dataset and other similar sets that were generated through the 10xGenomics Chromium platform.

Then, using the trained classifier framework, SuperCT, researchers were able to identify 37 cell types from healthy tissues; the types that were identified through machine learning were 87% concordant with the original cell labels. When they looked at the cells that were discordant compared with the labels — or those cells that were not identified by SuperCT — they noticed that their framework produced “more convincing marker signals” than those that were identified by hand.

The predictions were accurate even when the team analyzed new cell types (in tandem with the cells' corresponding training datasets), which demonstrated the tool's reliability and utility, they wrote.

When the team analyzed the scRNA-seq datasets from human and mouse pancreatic tumor tissues using SuperCT, they were able to assign 18 cell types, "including a population that is likely involved in epithelial-mesenchymal transition." They argued that knowing the composition and state of existing tumors may be important for the development of new therapeutic approaches in cancer — and SuperCT could help facilitate this knowledge.

Reference

  1. Lin W, Xie P, Gao M, et al. SuperCT, a supervised machine learning method to characterize cell types and states within solid tumor tissues. Presented at: AACR Pancreatic Cancer: Advances in Science and Clinical Care; Boston, Massachusetts, September 21-24, 2018. Abstract A079.

Related Resources

You must be a registered member of Cancer Therapy Advisor to post a comment.

Sign Up for Free e-newsletters



Regimen and Drug Listings

GET FULL LISTINGS OF TREATMENT Regimens and Drug INFORMATION

Bone Cancer Regimens Drugs
Brain Cancer Regimens Drugs
Breast Cancer Regimens Drugs
Endocrine Cancer Regimens Drugs
Gastrointestinal Cancer Regimens Drugs
Gynecologic Cancer Regimens Drugs
Head and Neck Cancer Regimens Drugs
Hematologic Cancer Regimens Drugs
Lung Cancer Regimens Drugs
Other Cancers Regimens
Prostate Cancer Regimens Drugs
Rare Cancers Regimens
Renal Cell Carcinoma Regimens Drugs
Skin Cancer Regimens Drugs
Urologic Cancers Regimens Drugs