Modeling single nucleus microglia across species identifies immune pathways and therapeutic candidates in Alzheimer’s disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disease characterized by memory loss and behavioral changes. A pivotal influence on AD pathology is the dysregulation of microglia in the brain. Despite promising findings in mouse models, there are limitations to the translatable biological information across species due to differences in the physiology, timeline of disease, and human heterogeneity. To address these interspecies discrepancies, we developed a novel implementation of the Translatable Components Regression (TransComp-R) framework, which integrated microglial single-nucleus transcriptomic data to identify biological pathways in mice AD models predictive of human AD. We compared model variations with sparse and traditional principal component analysis (PCA), finding that standard PCA encoded more interpretable mouse PCs compared to sPCA despite limited differences in technical performance. Mouse PCs significantly differentiated human AD from control microglial cells in the BA41/42, BA6/8, hippocampus and entorhinal cortex brain regions. However, these PCs had limited separation of human AD from control microglia in the prefrontal cortex. Additionally, we identified gene signatures from FDA-approved drugs that correlated with significant mouse component loadings, including valproic-acid and calcifediol. This computational framework may support the discovery of cross-species disease similarities, including the identification of candidate pharmacological solutions that may translate across species.