Batch Correction via Cross-Modal Anchoring

Anchor-guided batch correction across seven TCGA cohorts removes almost all platform variance while preserving biology, outperforming ComBat in biological preservation and CpG recovery under confounding.

• Why it matters: Unimodal batch correction struggles to disentangle batch effects from true biology without explicit covariates, limiting reliable integration of molecular cohorts.
• What they did: The method uses a comparatively stable paired omic layer to correct a batch-affected assay in a label-free way, returning full feature-level profiles; it was benchmarked with controlled confounding, role-reversed transcriptomic workflows, and simulations.
• The result: In an imbalanced breast cancer benchmark, ComBat yielded AUC 0.37 for subtype retrieval and 73.2% differential methylation recovery, while anchor-guided correction achieved AUC 0.72 and recovered 82.8% of reference CpG signals.

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