uniprot-pipeline

Imports external db ids, protein objects and sequences from UniProtKB.

PROTEIN LOADER

• load protein_to_gene associations (into RGD_ASSOCIATIONS table)
• load secondary uniprot xdb ids (into RGD_ACC_XDB table); secondary accession ids are imported as xdb ids with XDB_KEY=60
• load protein sequences (into RGD_SEQUENCES table);
  • the canonical protein sequences (available in the incoming data) are loaded with seq_type 'uniprot_seq'
  • the old protein sequences are created with seq_type 'old_uniprot_seq'; they are created when the incoming sequence differs from the sequence in the database; sequences of type 'old_uniprot_seq' combined are considered the 'consensus sequence history'

FILE PARSER

• processing of AC lines:
  • primary uniprot acc id is the first acc id on the first AC line
  • when a record had multiple AC lines, the other AC lines must be ignored;
• processing os OS lines: (species)
  • logic must be aware that OS value can span multiple lines
• processing of Gene3D accession ids:
  • they should be loaded just as they are; however in the past, Gene3D acc ids were in the format 'G3DSA:XX.XX.XX.XX'

added processing of chinchilla, bonobo, dog and squirrel

fixed matching logic to exclude self-matching (matching by UniProtKB ids brought in by UniProtKB pipeline in the past)

deletion of UniProt id results in insertion of an 'old_protein_id' alias

MATCHING

1. matching algorithm change:

• incoming data is matched to RGD data in the following order (as below); matching is terminated as soon there is some matching data
• OLD logic: RGD, GeneId, HGNC, MGI, UniProt, RefSeq Nucl, Ensembl
• NEW logic: GeneId, HGNC, MGI, RefSeq Nucl, UniProt, Ensembl, UniProt GeneName

2. improved handling of multis:

• where one UniProtKB id matches multiple genes in RGD previous logic was not properly handling of up to 0.1% of uniprot ids

3. discontinued ids_merged.log

• ids_inserted and ids_deleted logs until June 22, 2015 contain a lot of duplication: due to code flaw some entries are unnecessarily inserted, and during next pipeline run they are deleted and so on; therefore as of June 22, 2015 these logs are discontinued; inserted.log and deleted.log are used instead

PROTEIN DOMAINS

protein domains are identified by name and loaded into the database;

if a protein has a given domain appearing multiple times, the occurrence number is added to the domain name, f.e. P58365 protein has Cadherin domain appearing 27 times! so the domain names are like this: 'Cadherin 1', 'Cadherin 2', ... 'Cadherin 27' we want to keep a clean domain name and therefore we strip the last part and keep only the domain name f.e. 'Cadherin'