Using deep learning to quantify neuronal activation from single-cell and spatial transcriptomic data
Problem
Measuring neuronal activation from single-cell and spatial transcriptomic data was difficult; researchers needed a way to estimate activity-induced gene expression at cellular resolution.
Why it matters
Quantifying neuronal activation underpins understanding of memory, plasticity, and brain circuit function.
What the AI did
NEUROeSTIMator, a deep learning model, integrates transcriptomic signals to estimate neuronal activation, validated against Patch-seq electrophysiology and applied to published scRNA-seq and spatial transcriptomic studies.
Results
No numeric metrics stated; model shown associated with Patch-seq features, robust across species, cell type, and brain region, and accurately detected activity in published studies.
Impact
Enables estimation of learning-induced activity patterns across brain regions in mice, per the authors' demonstrations.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Ethan Bahl, Snehajyoti Chatterjee, Utsav Mukherjee, Muhammad Elsadany, Yann S. Vanrobaeys, Li‐Chun Lin, Miriam C. McDonough, Jon M. Resch et al. |
| Citations | 28 (OpenAlex, at import time) |
| References | 47 |
| Model or method | NEUROeSTIMator (deep learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-023-44503-5 |
| Open access | gold |
| Keywords | deep learning, single-cell RNA-seq, neuronal activation, spatial transcriptomics, Patch-seq |
Importance rating
real but narrower advance
Enables cellular-resolution measurement of neuronal activation, supporting brain circuit and memory research; a useful research tool but no scale of impact stated.