Incorrect protein sequences being retrieved for some accessions

widdowquinn commented 2 years ago

Summary:

Input protein sequences deriving from a known organism (e.g. human) are retrieiving nucleotide sequences from a different organism (e.g. bos taurus).

Description:

The input sequence

>CAD6020544.1/6-36 amtB [Escherichia coli] GN=CAD6020544.1
------DKADNAFMMICTALVLFMTIPGIALFYGGLI

does not give an output nucleotide sequence, as the wrong originating sequence is identified in the Elink linker step.

Reproducible Steps:

With the above sequence as input, run ncfp as normal:

Current Output:

[INFO] [ncbi_cds_from_protein.scripts.ncfp]: Sequence CAD6020544.1/6-36 matches GenBank entry X60065.1
[INFO] [ncbi_cds_from_protein.scripts.ncfp]: Searching for CDS: CAD6020544.1
[INFO] [ncbi_cds_from_protein.scripts.ncfp]: Could not identify CDS feature for CAD6020544.1/6-36

Expected Output:

A nucleotide coding sequence corresponding to the input protein, in the output directory.

`ncfp` Version:

commit 694d806

Python Version:

3.9

Operating System:

macOS

widdowquinn commented 2 years ago

This issue was brought to my attention by @tharis.

widdowquinn commented 2 years ago

First inspection indicates that:

the input protein recovers the correct source nucleotide sequence at the Elink step
the input protein recovers additional, incorrect sequences at the Elink step
where the incorrect recovered sequence is shorter than the correct sequence, it is preferrred over the correct sequence (for reasons of preserving bandwidth)
the incorrect sequence is recovered only when the /[start]-[end] Stockholm format domain is present in the input sequence accession/ID

widdowquinn / ncfp