|
Activity Number:
|
81
- Contributed Poster Presentations: Section on Statistics in Epidemiology
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
|
Abstract #313056
|
|
|
Title:
|
Automatic Cell Type Annotation of Single-Cell RNA Sequencing Data Using Deconvolution Methods
|
|
Author(s):
|
Amelia Schroeder* and Mingyao Li
|
|
Companies:
|
University of Pennsylvania School of Medicine and University of Pennsylvania
|
|
Keywords:
|
Single-cell RNA sequencing ;
Deconvolution;
Gene expression;
Cell type annotation;
Predictive modeling;
Cell identification
|
|
Abstract:
|
Single-cell RNA sequencing (scRNA-seq) is a powerful tool for understanding the cellular composition of complex tissues. A critical step in the analysis of scRNA-seq data is the identification and labeling of cell populations present in the data set. Existing annotation methods are time consuming, irreproducible, and only provide a cell label without quantifying the certainty of each prediction. Motivated by the concept of an automatic probabilistic cell type annotation method, we propose a novel application of bulk tissue deconvolution methods to automatically annotate cells by estimating the probability of each cell to a given cell type assignment. We utilized real data from human pancreatic tissue with known cell type labels to compare the prediction performance of two cell type annotation methods (i.e. CellAssign and scPred) to the performance of two deconvolution methods (i.e. Music and CiberSortX). We further evaluated the model performances by considering added noise and missing cell types in the reference data. Our results reveal that deconvolution methods can be effectively utilized to automatically annotate cells while providing probability estimates of each prediction.
|
Authors who are presenting talks have a * after their name.
