Deep learning–based cell composition analysis from tissue expression profiles
Problem
Inferring the cellular composition of bulk tissue from gene expression profiles was difficult and sensitive to noise and bias.
Why it matters
Cell composition underlies molecular understanding of healthy and diseased tissues from routine bulk expression data.
What the AI did
Scaden, a deep neural network trained on single-cell RNA-seq data, infers cell proportions from bulk expression without complex preprocessing or feature selection.
Results
Stated to outperform existing deconvolution algorithms in precision and robustness; no numeric metrics given.
Impact
A freely available tool deconvolving bulk RNA-seq and microarray data across human and mouse tissues.
Record
| Venue | Science Advances |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Kevin Menden, Matteo Marouf, Sergio Oller, Anupriya Dalmia, Daniel Sumner Magruder, Karin Kloiber, Peter Heutink, Stefan Bonn |
| Citations | 262 (OpenAlex, at import time) |
| References | 43 |
| Model or method | Scaden (deep neural network) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1126/sciadv.aba2619 |
| Open access | gold |
| Keywords | deep learning, cell deconvolution, RNA-seq, single-cell, tissue composition |
Importance rating
real but narrower advance
Abstract claims robust deconvolution deepens molecular and cellular understanding of developmental and disease processes; broadly useful analysis type, but no disease or population scale stated.