GPMDB evidence codes

GPMDB uses an evidence code system to rate the current observation status of individual protein sequences. The same system is used for gene loci, where applicable. The ratings are as follows:

The method for calculating these evidence codes in Perl can be found in the subroutine "gpmdbProteinEvidence" in gpmdb_rest.pl.

The evidence code associated with any protein accession number can be retrieved using the GPMDB REST interface with a URL call, for example, the evidence code for the mouse protein NSMUSP00000026459 can be obtained at:

http://gpmdb.thegpm.org/1/protein/evidence/acc=ENSMUSP00000026459

The current version of the algorithm for a protein is summarized below.

1. Retrieve the lowest E-value assigned to the protein of interest;
2. If that E-value is less than a threshold value (currently log(E) < -13.5), report EC = 4;
3. Retrieve all of the observed E-values for peptides assigned to the protein of interest;
4. If none of the peptides have an observation with an E-value less than 0.01, report EC = 1;
5. Generate a frequency histogram for each of the peptides, using bins based on the observations' log(E);
   • e.g., all appropriate observations are binned using the values floor(log(E)) = -2,-3,-4, ... -11;
6. Calculate the skew and excess kurtosis of the resulting histograms;
7. If the skew and excess kurtosis for any peptide are both ≤ 1.5, report EC = 4;
8. If either the skew or excess kurtosis for any peptide is ≤ 1.5, report EC = 3;
9. Report EC = 2.