GitLab

model_loader.js : fct strAnalysis - Use the value of manualchanged and insert clone values of Locus/segment if needed

clone_test.js : New Qunit tests (manual modification button, already changed icon, modification lists)

clone.js : button+fct to show/hide germline lists; icone to show if a clone has already been modify (+ param 'manuallyModify')

This reverts commit c31f2ef0.

One of the sequences in Stanford_S22 with a common k-mer between V and
D segments (with the non-default seed #####-#####).

There are common k-mers (with the non-default seed #####-#####) between IGHV and IGHD.
Interestingly, the number of detected junction is the same in the new version, but there are
more segmented reads, as the large '-w 100' option needs a fairly positioned window.

The strand detection can work even without any V/J k-mers, as when there are D k-mers
or k-mers from other locus (with the '-1' option).
These cases were previously reported as UNSEG_TOO_FEW_V, and now we correclty report these cases.

Bug detected thanks to trd-dd2-dd3-complete.should_get

The D sequences are now inserted in the index, but some unit tests
were previously launched by giving fake D repertoires. This is fixed.

There are sometimes common k-mers between D and V/J genes (this was at least
the case for IGH with -t 0), and this may give badly positioned windows.
Now all D k-mers are stored in the index, marking these k-mers as AMBIGUOUS.
Note that it is the first time that the D germline is actually used in the KmerSegmenter.

See d70796dc. We now test the clusters in the .vidjil output.

6577fd78 by @magiraud fixed the UNSEG_AMBIGUOUS, hence the test passes

We must clean the repertory first since Vidjil had previously be compiled for
tests (-O0 -g) and we don't want a release executable with such options.

The name of the executable depends on the architecture and on the release name.

The -m option is now only useful for FineSegmentation, not for the main heuristic.

All sequences are now segmented. Put -k 16 so that some sequences are unsegmented again.
Additionnally this tests whether the MapKmer works

More sequences are segmented in Standford_S22

delta_max is now useless. It was use to make sure we didn't segment
something not robust enough. Now we use the e-value, which is a much
trustable value.

delta_min is useless to the KmerSegmenter (we can't have overlaps
with the first_max, last_max algorithms).
But it can still be useful to FineSegmenter, that's why we keep it.

In the output the * could be expanded

When there are labeled windows, some kept windows may have, on purpose,
fewer reads than the '-r' threshold.
Previously an exception was launched in these cases. This is fixed.

Vidjil Team authored May 26, 2015 and Mathieu Giraud committed May 26, 2015

Started by a mail from @mikael-s to P. Wu, and edited and completed by @magiraud.

Since 40c17d5e and eafb88cb (before the 2015.05 release !), UNSEG_NOISY is not used.

It fails when the python version is incorrect.