Abstract:Uniform manifold approximation and projection (UMAP) has been rapidly adopted by the population genetics community to study population structure. It has become common in visualizing the ancestral composition of human genetic datasets, as well as searching for unique clusters of data, and for identifying geographic patterns. Here we give an overview of applications of UMAP in population genetics, provide recommendations for best practices, and offer insights on optimal uses for the technique.
“…Our results demonstrating the advantages of semi-supervised reduction suggest such methods could be utilized to refine upper bounds on the dimension of transcriptomes, such as the dimensions necessary to separate cell type designations [56]. Moreover, these questions and results are relevant to multi-faceted datasets outside of single-cell genomics, such as in phylogenetics or population genetics, where UMAP/t-SNE are used to explore structure of genetic interactions and evolutionary relationships [57]. Our findings should also be of interest beyond the biological sciences, e.g.…”
Dimensionality reduction is standard practice for filtering noise and identifying relevant dimensions in large-scale data analyses. In biology, single-cell expression studies almost always begin with reduction to two or three dimensions to produce 'all-in-one' visuals of the data that are amenable to the human eye, and these are subsequently used for qualitative and quantitative analysis of cell relationships. However, there is little theoretical support for this practice. We examine the theoretical and practical implications of low-dimensional embedding of single-cell data, and find extensive distortions incurred on the global and local properties of biological patterns relative to the high-dimensional, ambient space. In lieu of this, we propose semi-supervised dimension reduction to higher dimension, and show that such targeted reduction guided by the metadata associated with single-cell experiments provides useful latent space representations for hypothesis-driven biological discovery.
“…Our results demonstrating the advantages of semi-supervised reduction suggest such methods could be utilized to refine upper bounds on the dimension of transcriptomes, such as the dimensions necessary to separate cell type designations [56]. Moreover, these questions and results are relevant to multi-faceted datasets outside of single-cell genomics, such as in phylogenetics or population genetics, where UMAP/t-SNE are used to explore structure of genetic interactions and evolutionary relationships [57]. Our findings should also be of interest beyond the biological sciences, e.g.…”
Dimensionality reduction is standard practice for filtering noise and identifying relevant dimensions in large-scale data analyses. In biology, single-cell expression studies almost always begin with reduction to two or three dimensions to produce 'all-in-one' visuals of the data that are amenable to the human eye, and these are subsequently used for qualitative and quantitative analysis of cell relationships. However, there is little theoretical support for this practice. We examine the theoretical and practical implications of low-dimensional embedding of single-cell data, and find extensive distortions incurred on the global and local properties of biological patterns relative to the high-dimensional, ambient space. In lieu of this, we propose semi-supervised dimension reduction to higher dimension, and show that such targeted reduction guided by the metadata associated with single-cell experiments provides useful latent space representations for hypothesis-driven biological discovery.
“…Fig. 2 demonstrates uniform manifold approximate and projection (UMAP) GSE154245 data set for quality control (QC) ( Diaz-Papkovich et al, 2021 ). Fig.…”
“…Secondly, remaining samples were projected onto a principal component analysis (PCA), calculated with an in-house European reference panel to calculate the 20 first principal components for each population. UMAP 54 was used to reduce the coordinates of test samples on 20 principal components to two dimensions. Additional European samples not in the original reference set were also projected onto the PCA and UMAP components to identify ancestries, and samples with Swedish ancestry were identified.…”
Section: Genotyping and Association Analysismentioning
Understanding how hematopoietic stem and progenitor cells (HSPCs) are regulated is of central importance for the development of new therapies for blood disorders and stem cell transplantation. To date, HSPC regulation has been extensively studied in vitro and in animal models, but less is known about the mechanisms in vivo in humans. Here, in a genome-wide association study on 13,167 individuals, we identify 9 significant and 2 suggestive DNA sequence variants that influence HSPC (CD34+) levels in human blood. The identified loci associate with blood disorders, harbor known and novel HSPC genes, and affect gene expression in HSPCs. Interestingly, our strongest association maps to the PPM1H gene, encoding an evolutionarily conserved serine/threonine phosphatase never previously implicated in stem cell biology. PPM1H is expressed in HSPCs, and the allele that confers higher blood CD34+ cell levels downregulates PPM1H. By functional fine-mapping, we find that this downregulation is caused by the variant rs772557-A, which abrogates a MYB transcription factor binding site in PPM1H intron 1 that is active in specific HSPC subpopulations, including hematopoietic stem cells, and interacts with the promoter by chromatin looping. Furthermore, rs772557-A selectively increases HSPC subpopulations in which the MYB site is active, and PPM1H shRNA-knockdown increases CD34+ and CD34+90+ cell proportions in umbilical cord blood assays. Our findings represent the first large-scale association study on a stem cell trait, illuminating HSPC regulation in vivo in humans, and identifying PPM1H as a novel inhibition target that can potentially be utilized clinically to facilitate stem cell harvesting for transplantation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
