RNASeq optimized analysis mode

[dbolotin]

Finally I end up with following modifications to parameters to overcome false-positives in RNASeq as well as other poorly enriched RNA-based samples (not sure about DNA, but this should also work):

• (most important) turned off floating bounds of V and J alignments:
  • -OvParameters.parameters.floatingLeftBound=false
  • -OjParameters.parameters.floatingRightBound=false
• new thresholds:
  • -OvParameters.parameters.absoluteMinScore=80 (was 40)
  • -OjParameters.parameters.absoluteMinScore=70 (was 40)
  • -OminSumScore=200 (was 120; see below)
• include C and optionally D genes score into sumScore

• more strict scoring for all alignments (v, j, c?):

  • -OxParameters.parameters.scoring.gapPenalty=-21
  • -OxParameters.parameters.scoring.subsMatrix='simple(match=5,mismatch=-12)'

  scoring strategy for D gene could be also modified in this direction

All this modifications applied at the same time require final testing.

[dbolotin]

For those who want to use this functionality before 1.5 release. Here is the command line with approximately the same functionality that will be in rnaSeq optimized mode in the next release:

$ mixcr align -OvParameters.parameters.floatingLeftBound=false \
  -OjParameters.parameters.floatingRightBound=false \
  -OvParameters.parameters.absoluteMinScore=80 \
  -OjParameters.parameters.absoluteMinScore=70 \
  -OminSumScore=200 \
  -OvParameters.parameters.scoring.gapPenalty=-21 \
  -OjParameters.parameters.scoring.gapPenalty=-21 \
  -OvParameters.parameters.scoring.subsMatrix=simple(match=5,mismatch=-12) \
  -OjParameters.parameters.scoring.subsMatrix=simple(match=5,mismatch=-12) \
  .....