Cluster Attention for Graph Machine Learning

arXiv:2604.07492v1 Announce Type: cross Abstract: Message Passing Neural Networks have recently become the most popular approach to graph machine learning tasks; however, their receptive field

For decades, molecular biologists have interpreted gene regulation through measurements of mean gene expression, because they could not resolve regulatory variation among individual cells. The advent of single-cell genomics has now made that variation measurable, revealing pervasive differences in gene expression among apparently similar cells. Whether this variation mainly reflects stochastic noise or an informative regulatory property remains unclear. Here we show that mean-corrected gene expression dispersion is a reproducible and biologically structured feature of gene regulation that reflects regulatory fidelity. In heterogeneous differentiated cardiac cultures, genes with low dispersion are shared across cell types, enriched for housekeeping functions, depleted for expression quantitative trait loci, and more highly connected in transcriptional and protein interaction networks. In a comparative single-cell system spanning human, chimpanzee, and allotetraploid cells, a substantial subset of interspecies differences in regulatory dispersion persists in a shared trans environment, indicating that gene expression fidelity is often regulated in cis. Our findings establish gene expression dispersion as a genetically encoded dimension of gene regulation that is distinct from mean expression, and places dispersion along a fidelity-plasticity axis with implications for development, disease, and threshold-dependent cellular phenotypes.

Subscribe for Updates

Copyright 2025 dijee Intelligence Ltd.   dijee Intelligence Ltd. is a private limited company registered in England and Wales at Media House, Sopers Road, Cuffley, Hertfordshire, EN6 4RY, UK registration number 16808844