Downloading Human And Other Completely Sequenced Proteomes To Search Homologs

Question

[Pappu]

I want to download the human and other completely sequenced proteomes in order to search for homologs. A uniprot search results in ~136500 sequences in case of human:

http://www.uniprot.org/uniprot/?query=taxonomy%3A9606&sort=score

Searching for a protein sequence among these sequences yields too many homologs in human which is impossible. CD-HIT filtering by 90% sequence identity does not not reduce the number of hits much. The reviewed ~20000 entries in case of human do not include all the human proteins. I am wondering if Ensembl would be a better choice.

Answer 1

[Elisabeth Gasteiger]

See also this FAQ: What is the human complete proteome? http://www.uniprot.org/faq/48

Answer 2

[hpmcwill]

See the UniProt complete and reference proteome sets for a more appropriate set for this kind of search. While UniProtKB contains 136,536 entries describing human proteins, the corresponding reference proteome set contains 68,756 entries (see http://www.uniprot.org/taxonomy/9606).

[Pappu]

I am aware of that. As far as I know human has <30k protein sequences excluding alternative splicing. Ensembl seem to have ~100k human CDS.

Answer 3

[Biojl]

You can download that data from Ensembl. Take into account the transcript_biotype or gene_biotype tag. For human if you select only gene_biotype=protein_coding you'll end up with 22.836 transcripts in version 75 (biomart).