The human retina is a complex neural tissue that detects light and sends visual information to the brain. However, the molecular and cellular processes that underlie aging primate retina remain unclear. Here, we provide a comprehensive transcriptomic atlas based on 119,520 single cells of the foveal and peripheral retina of humans and macaques covering different ages. The molecular features of retinal cells differed between the two species, suggesting the distinct regional and species specializations of the human and macaque retinae. In addition, human retinal aging occurred in a region- and cell-type- specific manner. Aging of human retina exhibited a foveal to peripheral gradient. MYO9A− rods and a horizontal cell subtype were greatly reduced in aging retina, indicating their vulnerability to aging. Moreover, we generated a dataset showing the cell-type- and region- specific gene expression associated with 55 types of human retinal disease, which provides a foundation to understand the molecular and cellular mechanisms underlying human retinal diseases. Together, these datasets are valuable for understanding the molecular characteristics of primate retina, as well as the molecular regulation of aging progression and related diseases.
Elucidating how microscale molecular architecture shapes macroscale brain network organization requires integrative frameworks that link transcriptional profiles, structural connectivity, and functional dynamics across scales. However, existing evidence largely relies on cross-modal data acquired from different individuals, limiting causal and mechanistic inference. To overcome this, we performed concurrent awake optogenetic functional MRI (opto-fMRI) and spatial transcriptomics (opto-ST) within the same mice, enabling coupled analysis of neural activity and gene expression in fronto-thalamic and hippo-thalamic networks. Optogenetic stimulation of the medial prefrontal cortex (mPFC) or subiculum (SUB) evoked distinct thalamic BOLD response patterns. Multivariate spatial modeling revealed that energy metabolism and core biochemical processes underlie these GLM-derived activation patterns—a finding further supported by data-driven PCA, which identified enrichments in similar biological processes. These results suggest that conventional linear models predominantly capture linear gene–process relationships, highlighting the need for nonlinear computational approaches in fMRI analysis. Overall, our multi-omics framework reveals that genes regulating metabolic pathways underlie region-specific neurovascular coupling, thereby bridging local transcriptomics and large-scale functional imaging signatures.