The past few years have seen a hundred-fold explosion in the number of genetic variants associated with brain diseases. The vast majority are common, single-nucleotide variants (i.e. one-“letter” changes) for well-studied diseases. We are expanding this landscape by leveraging population-scale biobanks like the UK Biobank, All of Us and FinnGen, alongside disease-specific cohorts like the Alzheimer’s Disease Sequencing Project. We focus on understudied classes of variants like rare variants, structural variants, and repeats. We also explore understudied diseases like fibromyalgia and other “functional disorders”, recently highlighted by a top medical journal as one of medicine’s biggest failures.
Associating a variant with disease is only the first step towards understanding its function. Nearby variants co-occur across people (since chromosomes are inherited in blocks) so many nearby variants may be associated with a disease even if only one is causal. Variants may affect the expression of far-away genes, so it is often unclear which gene(s) a variant causes disease through, and we and others have shown that common ways of inferring this have high false positive rates. Finally, brains contain many types of cells, and different variants may cause disease through different cell types. We are developing statistical and machine learning models to infer causal variants, genes, and cell types for brain diseases, leveraging how variants relate to genes and cell types (single-cell omics) and how genes relate to each other (biological networks).
One of the most exciting new genetic technologies is single-cell omics, which measures biological variables like gene expression in individual cells. This lets us find molecular signatures of a disease in individual brain cell types. In collaboration with Vilas Menon’s lab at Columbia, we are co-leading the largest-ever single-cell meta-analysis of Alzheimer’s disease, encompassing over 3 million cells across 7 studies. We plan to cross-reference the molecular signatures gleaned from studies like this with single-cell CRISPR screens (e.g. Perturb-seq) in brain disease models, to find genes that induce an opposite molecular signature to the disease when perturbed. This approach represents a generalizable strategy for drug discovery.