Multi-omics Differential Inference for Functional Interpretation (MoDIFI): A Statistical Framework to Prioritize Cell Lines for Neurodevelopmental Variants

Noncoding variants contribute to neurodevelopmental disorders (NDDs), but their regulatory effects are often cell-type specific, making it difficult to choose an in vitro model for high-throughput assays such as massively parallel reporter assays. We asked: given a set of noncoding variants, which cell line and regulatory regions are most likely to reveal measurable allele-specific effects? We generated matched multi-omics profiles across commonly used NDD in vitro models: human neuronal lines (i.e., IMR-32, SH-SY5Y, SK-N-SH), mouse neuronal lines (i.e., HT-22, Neuro-2a), and a non-neuronal line (i.e., HEK-293), using RNA-seq, ATAC-seq, and Hi-C under consistent conditions. To integrate these orthogonal data types, we developed MoDIFI (Multi-omics Differential Inference for Functional Interpretation), a Bayesian framework that quantifies cell-line-specific regulatory activity by computing posterior inclusion probabilities (PIPs) for differential gene-loop interactions. MoDIFI identifies regulatory regions supported by coordinated 3D contacts, accessibility, and transcriptional output, producing cell-line-resolved regulatory maps that highlight both shared synaptic programs and context-dependent mechanisms. These results provide a practical strategy for prioritizing the most informative cell lines and candidate regulatory elements for targeted functional testing of NDD-relevant noncoding variation.