Skip to content

Latest commit

 

History

History
39 lines (29 loc) · 2.08 KB

missing.md

File metadata and controls

39 lines (29 loc) · 2.08 KB

What nr does not cover

The NCBI nr is every protein sequence that has been deposited publicly (GenBank translations, RefSeq, Swiss-Prot, PDB and smaller sources), with identical sequences merged into one entry. It is a list of what people have sequenced so far, not a census of life on Earth. Counts per part and group: statistics.md.

Coverage of the living world

About 2 million species are formally described. Only part of them have protein sequences in nr, and many of those only a handful of genes. Estimates of what exists (literature values, rounded):

estimated species described
eukaryotes ~8.7 million (Mora et al. 2011) ~1/5
insects ~5.5 million ~1 million
fungi 2-4 million ~150 000
protists largely unknown
prokaryotes millions up to ~1 trillion (Locey & Lennon 2016) most cannot be cultured

Metagenomic proteins (from environmental samples only) are largely not in nr; MGnify, for example, holds several billion such proteins, far more than nr.

Bias in the counts of this database

  • Gammaproteobacteria are the largest group (181 M sequences) because of tens of thousands of clinical isolate genomes of a few pathogens (E. coli, Klebsiella, Salmonella, Pseudomonas), not because they are the most diverse group.
  • Mammalia (15.5 M) are mostly human, mouse and farm animals: laboratory and medical research.
  • bacteria without a phylum (129 M) is the share from metagenome-assembled genomes (MAGs): organisms nobody has seen, known only from their DNA. It shows how much is still undiscovered.
  • All eukaryotes together (p1 + p2, ~172 M) are about 15 % of nr, although plants and animals make up almost all of Earth's biomass. Biomass and number of species have little to do with sequence counts.

In short: nr is dense around medicine, agriculture and model organisms and thin everywhere else. A search that finds nothing in nr does not mean the protein does not exist in nature.

Possible follow-up

Count the distinct species with sequences per group (a long query over blast_acc, best run at night).