mfSuSiE enables multi-cell-type fine-mapping and multi-omic integration of chromatin accessibility QTLs in aging brain.

Molecular quantitative trait locus (QTL) studies increasingly profile chromatin accessibility, histone modifications, DNA methylation, RNA modifications such as N6-methyladenosine (m6A), and transcription across multiple cell types using high-throughput sequencing, generating dense base-pair-resolved measurements. The conventional approach of testing each variant against each molecular feature independently suffers from severe multiple testing burden and ignores linkage disequilibrium and spatial correlation. Existing fine-mapping methods only partially address these challenges and are sub-optimal for analyzing such datasets: multivariate approaches such as mvSuSiE jointly analyze multiple molecular contexts but are designed for a single trait value per context and cannot accommodate thousands of base-resolution measurements per context, while functional approaches such as fSuSiE model spatial structure across thousands of measurements but analyze each context separately. Here, we introduce mfSuSiE, which integrates multivariate analysis with wavelet-based functional regression to jointly fine-map thousands of base-resolution traits across multiple cell types. In simulations, mfSuSiE identified causal variants and affected molecular features more accurately than fSuSiE, while mvSuSiE cannot be applied to this type of data. Applied to single-nucleus chromatin accessibility data from six brain cell types from postmortem aging human brains, mfSuSiE substantially increased discovery and resolution, with substantial power gains for cell types with limited samples. Multi-cell-type analysis revealed extensive sharing of regulatory effects on chromatin accessibility (caQTL). Importantly, mfSuSiE produces Bayesian inference compatible with the SuSiE framework, enabling systematic multi-omic integration. Applied to Alzheimers disease loci, we integrated caQTL with expression QTLs, epigenomic QTLs, and GWAS, observing regulatory patterns suggesting complex mechanisms at loci including EARS2, CHRNE, SCIMP, and RABEP1.