What it is
UMAP (Uniform Manifold Approximation and Projection) is a nonlinear dimensional-reduction method used in this study to identify distinct antimicrobial-resistance (AMR) ecotypes. 1
- Canonical name: UMAP. 1
- Known aliases: Uniform Manifold Approximation and Projection. 1
- Stable external identifier: None was reported in the source documents. 1,2
- Related method: dbscan was applied after UMAP for density-based clustering. 1
Applications and findings
In the AMR ecotype analysis, UMAP included 974 species with at least 15 genomes suitable for clustering. 1 Of those 974 species, 190 (19.5%) formed at least two distinct AMR ecotypes. 1 The median silhouette score for the ecotype clusters was 0.620. 1
Case-study UMAP plots for klebsiella-pneumoniae, staphylococcus-aureus, and salmonella-enterica showed visible environmental structuring, although the available statistical tests were underpowered to establish a general environment–ecotype association. 1 Environment–ecotype testing was limited because 52.7% of genomes had no classifiable isolation_source, and only 2 species had sufficient within-species environmental diversity for chi-squared testing after strict expected-frequency criteria. 1 Escherichia coli was not included in the case studies because it exceeded the 500-genome computational cap. 1 UMAP outputs included 176,177 ecotype assignments and 974 ecotype summaries. 1
The genome-embedding explorer refines this AMR-focused use: UMAP projected 64-dimensional AlphaEarth satellite-derived embeddings for 83,287 genomes into two dimensions, and dbscan identified 320 clusters. 2 Many clusters were dominated by a single harmonized environment category, while phylum coloring also showed taxonomic structure partly confounded with environment. 2 This supports using UMAP to inspect environmental and taxonomic structure, but its nonlinear projections depend on parameters including n_neighbors and min_dist and may not represent the true high-dimensional topology. 2
The method and its results contribute to Within-Species AMR Strain Variation and AlphaEarth Embeddings, Geography & Environment Explorer, informing the environmental-resistome synthesis and the environment-embedding-geography synthesis. 1,2