Contrasting within-donor gene covariation identifies disease-associated cells and genes in case-control single-cell studies

IDC identifies disease-associated cells and genes in case-control single-cell studies by contrasting within-donor gene covariation, improving affected-cell detection and donor burden estimation under strong unrecorded batch effects.

• Why it matters: Disease signals can be diluted when all cells are pooled across patients and controls, and donor or batch shifts can be confounded with disease status. IDC addresses this by using within-donor cellular heterogeneity rather than batch labels.
• What they did: IDC contrasts within-donor gene-gene covariance between patients and controls, then uses positive-unlabeled learning to score each cell’s disease-like state and estimate donor disease burden. It also computes closed-form gene contribution and leave-gene-out cell scores for within-donor differential expression and cross-donor meta-analysis.
• The result: In simulations, IDC more accurately identifies affected cells and estimates donor burden than deep generative and label-refinement methods, and detects differentially expressed genes more sensitively than case-control pseudobulk while controlling false discoveries. In human Homo sapiens pancreatic islets and prefrontal cortex, it separates T2D from non-diabetic beta cells and late-stage AD from non-AD astrocytes and inhibitory neurons, with prediabetic and early-AD donors at intermediate burden levels.

AI summary of the abstract · may contain errors · not medical advice · read the paper.