Integrative multi-omics QTL colocalization maps regulatory architecture in aging human brain.

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Multi-trait QTL (xQTL) colocalization identifies shared genetic etiology across molecular data and diseases.
• Integrating large-scale multi-omics data for xQTL regulation insights is limited by current methods' scalability and efficiency.

Purpose of the Study:

• To develop a scalable and efficient multi-task learning method, ColocBoost, for multi-trait QTL colocalization.
• To enhance the detection of shared genetic signals and causal variants by accounting for multiple causal variants and adaptively coupling traits.

Main Methods:

• ColocBoost utilizes a specialized gradient boosting framework for multi-task learning.
• The method is applied genome-wide to 17 gene-level xQTL datasets from aging brain cortex (ROSMAP), covering multiple cell types, brain regions, and molecular modalities (expression, splicing, protein abundance).

Main Results:

• ColocBoost identified 16,503 distinct colocalization events, showing a 10.7-fold enrichment for heritability across 57 complex diseases/traits.
• It demonstrated strong concordance with CRISPR-validated element-gene pairs and identified 2.5-fold more Alzheimer's disease (AD) colocalized loci, explaining twice the AD heritability.
• The method enhanced detection of gene-distal colocalizations, aligning with known enhancer-gene links and providing functional support for genes like BLNK and CTSH in AD pathogenesis.

Conclusions:

• ColocBoost offers a scalable and efficient approach for multi-trait QTL colocalization, advancing the integration of multi-omics data.
• The method significantly improves the identification of shared genetic signals and causal variants, particularly for complex diseases like AD.
• ColocBoost provides novel functional insights into disease susceptibility loci and underlying regulatory mechanisms.