GitLab

Following the fix on min_cover_representative.
The representative on small datasets may be longer.

Thus we need to update those. There also some side effects:
- V genes are less ambiguous (hence less warnings)
- Consensus is longer (hence less warnings too)
- Positions changes of end of V, J, CDR3…

Checks previous commit

See #3792.

See #3785.

Use the new option names.
Remove also '--trim 0' options used in some tests.

See #3295.

This should ease parsing the output.

see also should#31

close #3772.

was done twice in 40aef950 and 556d8b0b, but git merge
did not understood what we meant there ;-)

see #3773

The dataset was generated to have a minority of sequences with a TTT insertion
which would make it the sequence of choice with the default ReadScore (based
on length and quality).

On the contrary with a random sampler we should not be impacted by this
insertion as it happens in the minority of sequences.

Here is the command that generated the dataset:
for i in $(seq 1 3500); do
  if [ $((RANDOM%3)) -ne 0 ]; then
    echo ">seq$i";
    echo ctacctactactgtgccttgtgggaggtgatagtagtgattggatcaag;
  else
    echo ">seq$i";
    echo cTTTtacctactactgtgccttgtgggaggtgatagtagtgattggatcaag;
  fi;
done > test-random-consensus.fa

See 1a7bb7f8 for reason why.

Fix (in part?) #3771

Some old slaves may work better with less germlines.
At least this separates independant tests.

See #3652.

With Aho we have slightly longer affectations (on the whole read instead of
read - k + 1). Those k-mers are unaffected.

Just a random k-mer in the J may allow the sequence to be segmented (because
there is few sequence). Therefore we slightly lower the e-value

See #2595

With Aho-Corasick we now segment them all!

It is segmented even without any downstream sequence. The e-value is now
better due to the Aho-Corasick implementation.

However on should-vdj the locus found is unexpected rather than TRD+
(this is not the case for the should-get as we don't search for unexpected by
default)

see #3762

Closes #3729.

See #1233.

Close #3572 and #3573.

Closes #3575.