scDNM-VAE enables directly inspectable deep clustering of single-cell RNA-seq data through signed dendritic gating
Abstract
Deep clustering models for single-cell RNA sequencing often assign cells through latent or centroid-based mechanisms that are difficult to inspect. We introduce scDNM-VAE (single-cell Dendritic Neuron Model Variational Autoencoder), a deep clustering framework that combines a variational autoencoder with a dendritic neuron-inspired head. Cluster assignments are governed by learnable signed synaptic weights and thresholds: the weight sign determines the direction of a gate's response to a latent coordinate, its magnitude controls steepness, and the weight-threshold pair determines the transition location. The trained clustering function can therefore be inspected directly without fitting a post-hoc explanation model. We benchmark scDNM-VAE on four datasets spanning immune, cortical, cardiac, and hematopoietic cells against scVI followed by KMeans and an MLP-DEC ablation. scDNM-VAE performs better than scVI on PBMC3k, comparably on the Human Heart Cell Atlas and Paul15, and worse on Zeisel, while producing biologically coherent marker-gene signatures. Ablating each cluster's three highest-magnitude synaptic dimensions causes numerically greater reassignment than random-dimension ablation across all datasets, but the margins are modest and negligible on Zeisel. These results show that signed dendritic gating supports competitive clustering with a parameter-inspectable decision function, while indicating that decision-relevant information is distributed across the latent space.
Related
- Droplet-array culture enables single-cell isolation of bacteria directly from blood
- BDCA2-dependent phagocytosis enables strain-specific activation of plasmacytoid dendritic cells by lactic acid bacteria
- Unsupervised Clustering Enables Cohort-Specific Variant Recalibration in Crops
- Agricultural practices shape beetle communities in field margins directly and through indirect effects driven by plants
- Tolerogenic dendritic cells attenuate collagen-induced arthritis through a mucosal immunity–gut microbiota–butyrate axis
BibTeX
@misc{agraz2026scdnmvae,
title = {{scDNM-VAE enables directly inspectable deep clustering of single-cell RNA-seq data through signed dendritic gating}},
author = {Melih Agraz and Deniz Karapinar and Aysel Topsir and Qianying Cao and Erol Egrioglu and Gaurav Choudhary},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2608.17228},
eprint = {2608.17228},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.17228},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - scDNM-VAE enables directly inspectable deep clustering of single-cell RNA-seq data through signed dendritic gating
AU - Melih Agraz
AU - Deniz Karapinar
AU - Aysel Topsir
AU - Qianying Cao
AU - Erol Egrioglu
AU - Gaurav Choudhary
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2608.17228
UR - https://arxiv.org/abs/2608.17228
AB - Deep clustering models for single-cell RNA sequencing often assign cells through latent or centroid-based mechanisms that are difficult to inspect. We introduce scDNM-VAE (single-cell Dendritic Neuron Model Variational Autoencoder), a deep clustering framework that combines a variational autoencoder with a dendritic neuron-inspired head. Cluster assignments are governed by learnable signed synaptic weights and thresholds: the weight sign determines the direction of a gate's response to a latent coordinate, its magnitude controls steepness, and the weight-threshold pair determines the transition location. The trained clustering function can therefore be inspected directly without fitting a post-hoc explanation model. We benchmark scDNM-VAE on four datasets spanning immune, cortical, cardiac, and hematopoietic cells against scVI followed by KMeans and an MLP-DEC ablation. scDNM-VAE performs better than scVI on PBMC3k, comparably on the Human Heart Cell Atlas and Paul15, and worse on Zeisel, while producing biologically coherent marker-gene signatures. Ablating each cluster's three highest-magnitude synaptic dimensions causes numerically greater reassignment than random-dimension ablation across all datasets, but the margins are modest and negligible on Zeisel. These results show that signed dendritic gating supports competitive clustering with a parameter-inspectable decision function, while indicating that decision-relevant information is distributed across the latent space.
ER -