Thumbnails:
List:
Year:
Category:
Session:
Poster:
Getting poster data...
Alex Diaz-Papkovich, Luke Anderson-Trocme, Simon Gravel (McGill University, Montreal, Canada.)
Genetic structure arises from technical, sampling, and demographic variation. Visualization is therefore a first step in most genomic analyses. A key challenge is how to effectively condense data from hundreds of thousands of dimensions to just two or three, as existing dimension reduction methods struggle with unbalanced sampling, dataset size, and reconciling local and global relationships between data. We investigate an approach that combines principal components analysis (PCA) with uniform manifold approximation and projection (UMAP) to illustrate population structure. We demonstrate that this approach intuitively clusters individuals who are more closely related while placing them in a global continuum of genetic variation. Using several datasets, we show that it reveals previously overlooked populations in the American Hispanic population as well as fine-scale relationships between geography, genotypes, and phenotypes in the UK population.