Deep exponential families for single-cell data analysis


Loading...

Date

2022-09-18

Publication Type

Working Paper

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Single-cell gene expression data characterizes the complex heterogeneity of living systems. This heterogeneity is composed of various cells with diverse cell states driven by different sets of genes. Cell states can often belong to different categories according to biological hierarchies. Hierarchical clustering therefore not only improves functional interpretation, but can also be leveraged to ensure that gene signatures are less influenced by noise and batch effects. We present single-cell Deep Exponential Families (scDEF), a two-level Bayesian matrix factorization model which extracts hierarchical gene signatures from single-cell RNA-sequencing data. scDEF can additionally make use of prior information to guide the hierarchy. It can be used for dimensionality reduction, gene signature identification, and batch integration. We validate scDEF with simulations and multiple annotated real data sets, and show that scDEF recovers meaningful hierarchies in single- and multiple-batch scenarios.

Publication status

published

Editor

Book title

Journal / series

Volume

Pages / Article No.

Publisher

Cold Spring Harbor Laboratory

Event

Edition / version

1v

Methods

Software

Geographic location

Date collected

Date created

Subject

Organisational unit

03790 - Beerenwinkel, Niko / Beerenwinkel, Niko check_circle

Notes

Funding

Related publications and datasets