Deep learning of gene relationships from single cell time-course expression data
Problem
Inferring regulatory and signaling relationships between genes from single-cell time-course RNA-Seq data lacked methods, as existing tools were built for bulk data.
Why it matters
Mapping gene interactions is fundamental to understanding cell behaviour and disease mechanisms.
What the AI did
TDL encodes scRNA-Seq data as 3D tensors and trains convolutional and recurrent neural networks to predict gene-gene interactions.
Results
Accurately identified causal and regulatory interactions across five scRNA-Seq datasets; quantitative metrics not stated
Impact
Peer-reviewed method enabling interaction prediction and gene function assignment from single-cell data.
Record
| Venue | Briefings in Bioinformatics |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Ye Yuan, Ziv Bar‐Joseph |
| Citations | 38 (OpenAlex, at import time) |
| References | 39 |
| Model or method | TDL (CNN and RNN on 3D tensors) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/bib/bbab142 |
| Open access | green |
| Keywords | single-cell RNA-Seq, gene regulation, deep learning, time-course, neural networks |
Importance rating
real but narrower advance
Gene-gene interaction inference from single-cell time-course data is fundamental to understanding cell behavior and disease, but abstract states no scale of impact or affected population.