Identification of cell types and cellular dynamics genetically associated with brain disorders and cognitive traits

Human brain development plays a fundamental role in the development of brain disorders and cognitive ability, yet the process remains poorly understood. Recent studies have proposed that a more comprehensive understanding of brain development and its relationship to complex traits can be achieved by interpreting genetic signals from genome-wide association studies (GWAS) in cell type resolution. However, the causal cell type in which genetic variants and genes affect trait variation is not known, and there is no gold standard to compare the performance of different methods. To address this, we first established a set of putatively causal cell type and trait/disease pairs as the ground truth based on general knowledge and empirical evidence from prior studies. We curated GWAS data from a broad range of traits and diseases and used single cell RNA-seq data and candidate cell type specific cis-regulatory element (cCRE) annotations from both murine and human datasets. We comprehensively assessed the performance of different cell-type specific metrics and different statistical methods that integrate GWAS, single cell transcriptomic and epigenomics data. The best method in the benchmark analysis was then applied to 11 brain disorders and cognitive traits, using scRNA-seq data from 154,748 single nuclei in human prefrontal cortex (PFC) across 6 developmental stages (data from Herring et al 2022 Cell). Single-cell disease relevance score (scDRS) was used to detect temporal effects on the cellular heterogeneity in association of disease. In the method benchmark analysis, we found that the performance of different cell-type specific metrics and integrative methods for GWAS, scRNA-seq and cCRE annotation data could be sensitive to the underlying biological mechanisms and trait genetic architecture. Using a Cauchy combination method to combine results across methods maximized power and well controlled false positive rate for identifying associations between cell types and diseases. In the analysis of PFC data, we identified cell types significantly associated with brain disorders/traits, among which a large proportion exhibited significant cellular heterogeneity signals in samples across states of development. Subsequent analysis found cell types in which the cumulative expression of disease-associated genes increased or decreased with the observed developmental stage or inferred cell age from genes with differential expression over development. Our study provides a benchmark between different methods for integrative analysis of GWAS,scRNA-seq data and cCRE annotations. Our application of the best method on the PFC data enhances our understanding of brain disease aetiology in relation to the dynamics of brain development, and the results suggest that the disease-associated genes may be consistently upward or downward regulated before disease onset.