CAMML: Multi-Label Immune Cell-Typing and Stemness Analysis for Single-Cell RNA-sequencing.
Courtney Schiebout1, H Robert Frost
1Biomedical Data Science, Dartmouth College, Lebanon, NH 03766, USA, Courtney.T.Schiebout.GR@dartmouth.edu.
Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing
|December 10, 2021
Summary
We developed CAMML, a novel method for single-cell RNA sequencing (scRNA-seq) cell typing in complex tissues like the tumor-immune microenvironment (TME). CAMML accurately identifies single cell types and enables multi-label classification for cells with overlapping immune phenotypes.
Area of Science:
- Computational Biology
- Genomics
- Immunology
Background:
- Single-cell RNA sequencing (scRNA-seq) is crucial for understanding complex tissues, especially the tumor-immune microenvironment (TME).
- Challenges in scRNA-seq analysis include data sparsity, noise, and the continuous nature of immune cell types, hindering accurate cell identification.
- Existing single-label cell typing methods struggle with overlapping immune cell phenotypes in the TME.
Purpose of the Study:
- To develop a robust scRNA-seq cell-typing method capable of handling data sparsity, noise, and overlapping cell types.
- To introduce a novel multi-label classification approach for scRNA-seq data analysis.
- To improve the accuracy and adaptability of cell type inference in complex biological systems.
Main Methods:
- Developed Cell-typing using variance Adjusted Mahalanobis distances with Multi-Labeling (CAMML), a new scRNA-seq cell-typing method.
- CAMML utilizes cell type-specific weighted gene sets to score cells for multiple potential cell types.
- Implemented both single-label (highest score) and multi-label (score cut-off) classification strategies.
Main Results:
- CAMML demonstrates performance comparable to existing methods like SingleR and Garnett for single-label cell typing.
- The multi-label classification capability of CAMML provides significant advantages for identifying cells with mixed or undifferentiated phenotypes.
- CAMML proves robust and adaptable for integrating data across diverse studies, omics platforms, and species.
Conclusions:
- CAMML effectively addresses the challenges of scRNA-seq data sparsity, noise, and overlapping cell types, particularly in the TME.
- The method's multi-label classification offers a significant advancement over current state-of-the-art techniques for complex cellular populations.
- CAMML provides a versatile tool for enhancing scRNA-seq analysis and facilitating deeper biological insights.
