algo: -e, update help and tests

algo/tests/should-vdj-to-tap.py

@@ -20,8 +20,8 @@ from subprocess import Popen, PIPE, STDOUT
 import os
 import argparse
 
-VIDJIL_FINE = '{directory}/vidjil -p 1e-6 -c segment -i -g {directory}/germline %s > %s'
-VIDJIL_KMER = '{directory}/vidjil -p 1e-6 -b out -c windows -uU -i -g {directory}/germline %s > /dev/null ; cat out/out.segmented.vdj.fa out/out.unsegmented.vdj.fa > %s'
+VIDJIL_FINE = '{directory}/vidjil -e 1e-6 -c segment -i -g {directory}/germline %s > %s'
+VIDJIL_KMER = '{directory}/vidjil -e 1e-6 -b out -c windows -uU -i -g {directory}/germline %s > /dev/null ; cat out/out.segmented.vdj.fa out/out.unsegmented.vdj.fa > %s'
 
 parser = argparse.ArgumentParser(formatter_class=argparse.RawTextHelpFormatter)
 parser.add_argument('--program', '-p', default=VIDJIL_FINE, help='program to launch on each file (%(default)s)')

doc/algo.org

@@ -159,7 +159,7 @@ Window prediction
   -m <int>      minimal admissible delta between last V and first J k-mer (default: -10) (default with -D: 0)
   -M <int>      maximal admissible delta between last V and first J k-mer (default: 20) (default with -D: 80)
   -w <int>      w-mer size used for the length of the extracted window (default: 40)(default with -D: 60)
-  -p <float>    e-value like for determining if a window must be segmented or no (default: 100)
+  -e <float>    maximal e-value for determining if a segmentation can be trusted (default: 'all', no limit)
 #+END_EXAMPLE
 
 The =-s=, =-k=, =-m= and =-M= options are the options of the seed-based heuristic. A detailed
@@ -179,12 +179,11 @@ Setting =-w= to 30 for VJ and 50 for VDJ recombinations may "segment" (analyze)
 few more reads, but may in some rare cases falsely cluster reads from
 different clones.  Setting =-w= to lower values is not recommended.
 
-The =-p= option is used to determine what is the maximal e-value accepted for
-segmenting a sequence.  If a segmentation has a higher e-value, it will not be
-segmented.  The default value is 100 but it is *not* the recommended
-value. The purpose is to keep Vidjil with the same behaviour. The
-recommendation would be to use a value of 1e-6 (or lower with huge datasets).
-
+The =-e= option sets the maximal e-value accepted for segmenting a sequence.
+If a segmentation has a higher e-value, it will not be segmented.
+The default value is 'all', but the recommended value is to use something
+like 1e-6 for datasets with a billion of reads.
+Further help on this point will come in next releases.
 
 ** Threshold on clone output
 
@@ -264,7 +263,7 @@ browser.
 Setting the =-n= option triggers an additional automatic
 clustering using DBSCAN algorithm (Ester and al., 1996). 
 
-The =-e= option allows to specify a file for manually clustering two windows
+The =-E= option allows to specify a file for manually clustering two windows
 considered as similar. Such a file may be automatically produced by vidjil
 (out/edges), depending on the option provided. Only the two first columns 
 (separed by one space) are important to vidjil, they only consist of the