Skip to content

Protein View

The protein view holds what the sequence database says of each protein a dataset names: its name, organism and sequence. Every other view keeps accessions only, so join this one on them. It is optional, and specification 1.3 added it.

File: {PREFIX}.protein.parquet. One row per accession.

Field Description Type Required
accession Protein accession as the other views name it string Yes
entry_name Entry name of the database, such as ALBU_HUMAN string, null No
description Protein name from the FASTA header string, null No
gene_name Gene symbol from the header's GN= string, null No
organism Organism from the header's OS= string, null No
taxonomy_id NCBI taxonomy id from the header's OX= int32, null No
length Sequence length in residues int32, null No
sequence Amino acid sequence string, null No

Producers

  • qpxc transform protein-properties --fasta and the --fasta of qpxc convert write it for the accessions of the pg and feature views that the FASTA has, with sequences. A value a producer recorded stays.
  • qpxc convert maxquant writes it from proteinGroups.txt: the Fasta headers and Sequence lengths of each Protein IDs member, where each cell has one token per member. It has no sequences.

A header not in UniProt form yields only its description. Decoys stay out, since a FASTA's decoy entries are skipped and MaxQuant's REV__ rows are dropped.

The FASTA itself

The provenance view's protein_properties_from_fasta step records the file's fasta, fasta_sha256 and, with --fasta-uri, the fasta_uri others can fetch it from. Use sequence when the file is not at hand.

Peptide positions

feature.pg_positions holds where a peptide sits in its proteins, and pepmap.pg_accessions also has per-protein start, end, pre and post. The start, end, pre and post of feature.pg_accessions are reserved: no converter fills them, so do not read positions there.