Is there a way to make diamond return a higher identity subsequence/Diamond significantly undertrims candidate sequence.

[Thernn88]

Sometimes Diamond correctly finds a match, but extends the match far past where it should end.

Using the following command.

diamond blastx -d {diamond_db_path} -q {input_file.name} -o {out_path} --very-sensitive --masking 0 -e {args.evalue} --outfmt 6 qseqid sseqid qframe evalue bitscore qstart qend sstart send pident {quiet} --top {top_amount} --min-orf 20 --max-hsps 0 -p {num_threads}"

Example below. Diamond returns a match against these references for this sequence.

NODE_451470 28588:002869 -1 2.72e-23 88.6 287 66 3 81 64.6
NODE_451470 7227:002172 -1 2.72e-23 88.6 287 66 3 81 64.6
NODE_451470 7159:0010bd -1 6.23e-22 85.1 287 66 3 82 63.8
NODE_451470 7165:0005df -1 1.95e-20 81.3 287 66 3 82 58.8

You may notice that Pident is rather low.

What has happened is that Diamond has extended the correctly mapped section of our candidate sequence to include a poorly scoring segment that does not map.

In the below picture I've aligned the output coords of NODE_451470 to the references. NODE_451470 is the 1st sequence. You will see at the end there is a section that is pure junk.

This is quite a frustrating problem, so I appreciate any advice you have to give.

Edit: I am using Diamond v2.0.14.152

[Thernn88]

Using 2.1.6 returns a different (better) answer.

NODE_451470 7227:002172 -1 1.08e-22 87.0 287 141 3 54 84.6
NODE_451470 28588:002869 -1 3.05e-22 85.9 287 141 3 54 82.7
NODE_451470 7159:0010bd -1 6.96e-21 82.4 287 141 3 55 81.1
NODE_451470 7165:0005df -1 7.74e-20 79.7 287 141 3 55 75.5

What version was this changed? I go back through versions but it will take some time.

[bbuchfink]

I'm not sure what changed between the versions. If you send me the sequences I can look into it.