The metabolic universe of 4,659 genome-scale models
Two pangenome collections — 2,313 Escherichia coli and 2,346 Lactobacillaceae (26 species, 13 genera) — reconstructed as constraint-based metabolic models and reduced to a shared reaction/metabolite vocabulary you can compare, cluster and embed entirely in your browser.
Metabolic landscape new
Reaction frequency spectrum new
Collection at a glance
Metabolic landscape embedding
Classical multidimensional scaling (PCoA) of reaction-content distances. Each point is positioned so that metabolic dissimilarity ≈ 2-D distance. Zoom, pan and lasso to explore; hover any point for its identity.
Panreactome openness
Is the metabolic panreactome open (new reactions keep appearing as you add genomes) or closed? Two classic lenses: the reaction-frequency spectrum, and rarefaction/accumulation curves with a Heaps'-law openness exponent.
Reaction frequency spectrum
Rarefaction / accumulation new
Species × species metabolic distance
A clustered matrix of all 27 organisms compared by the cosine distance of their mean reaction-presence vectors. Complements the dendrogram: blocks of low distance are metabolically coherent clades.
Species tree
Toggle between NCBI-style taxonomy and a dendrogram built purely from metabolic (reaction) content. The content tree recovers the genera and places E. coli as an outgroup — click any species to load its GEMs into the clustermap.
Genome properties & model richness
How the raw genomes translate into models. GC content, genome length and CDS count set the scale of each reconstruction — here they meet the reaction, metabolite and gene counts of the models built from them, split by collection.
GC content vs genome length
Genome length vs model size
Distribution of model size by collection
Geography & ecology of the collection
Where these strains were isolated, and from what. Sampling geography and ecological niche shape any pangenome — a collection is only as broad as the places and habitats it was drawn from, so read these as the sampling frame, not the true biology.
Isolation geography
Top isolation sources
Top host organisms
Compare specific models
Add two or more models to contrast their reaction / metabolite repertoires — pairwise Jaccard, an UpSet-style intersection plot, and a differential feature table.
Compare groups of GEMs
Define two cohorts by species or metadata, then contrast their repertoires — distributions, overlap, a differential-prevalence table, and a statistical volcano plot.
Presence / absence clustermap upgraded
A seaborn-style clustermap: rows (models) and columns (features) are hierarchically clustered, with dendrograms on both margins and a collection/species annotation strip. Filled cell = feature present.