Crash when using "mm" as species name

fabio-t commented 8 years ago

From within the mixcr installation directory, I ran importFromIMGT.sh and asked for "Mus Musculus" data, with names "mm:musmusculus".

When I try to run the alignment using -s mm, it crashes as below. If I use, instead, musmusculus, it works fine.

mixcr align -r log_align.txt -l TRA -s mm 150814_TCR_Valpha8_MID10_S7_L001_R1_001.fastq 150814_TCR_Valpha8_MID10_S7_L001_R2_001.fastq alignments.vdjca -f
Alignment: 3.3%
Exception in thread "main" java.lang.NullPointerException
    at com.milaboratory.mixcr.vdjaligners.VDJCAlignerPVFirst.createInitialHelper(VDJCAlignerPVFirst.java:127)
    at com.milaboratory.mixcr.vdjaligners.VDJCAlignerPVFirst.createInitialHelpers(VDJCAlignerPVFirst.java:122)
    at com.milaboratory.mixcr.vdjaligners.VDJCAlignerPVFirst.process(VDJCAlignerPVFirst.java:58)
    at com.milaboratory.mixcr.vdjaligners.VDJCAlignerWithMerge.process(VDJCAlignerWithMerge.java:86)
    at com.milaboratory.mixcr.vdjaligners.VDJCAlignerWithMerge.process(VDJCAlignerWithMerge.java:42)
    at cc.redberry.pipe.CUtils$9.process(CUtils.java:350)
    at cc.redberry.pipe.CUtils$9.process(CUtils.java:345)
    at cc.redberry.pipe.blocks.ParallelProcessor$Worker.run(ParallelProcessor.java:304)
    at java.lang.Thread.run(Thread.java:745)

dbolotin commented 8 years ago

Add --library local option to activate imported library of segments (your current command still uses built-it library, where mm is not defined, but musmusculus is).

Let me know whether it worked.

P.S. IMGT library for mouse TRA / TRD is badly formatted which results to wrong boundaries of CDR3. I'am currently releasing a version with fix for this problem (1.7.1). Hope it will be done in two-three hours.

fabio-t commented 8 years ago

Works great, thanks!

Although might I suggest some input validation on that option? Might be better to have an error telling me "species name unrecognised" or something similar than a NullPointerException :)

I will be eagerly waiting for the fix on mouse TRA; thanks for letting me know!

dbolotin commented 8 years ago

You are totally right, I will add human-readable error message here! Thanks for suggestion.

dbolotin commented 8 years ago

Related to #70.

dbolotin commented 8 years ago

Just finished: https://github.com/milaboratory/mixcr/releases/tag/v1.7.1

Please let me know if there will be any problems with it.

fabio-t commented 8 years ago

Doesn't seem to be working, the sequences have a wrong left boundary:

TACTACTGTGCTTTGAGTAGCAACTATCAGTTGATCTGG YYCALSSNYQLIW

I updated with brew, and it shows there version 1.7-2. However, mixcr version seems wrong:

$ mixcr --version
MiXCR v1.7 (built Tue Dec 29 10:58:03 CET 2015; rev=87f1b07; branch=release/v1.7)
Components: 
MiLib v1.2 (rev=4f56782; branch=release/v1.2)
MiTools v1.2 (rev=eb91603; branch=release/v1.2)

Might be that 1.7.1 didn't build properly?

dbolotin commented 8 years ago

Yap, thanks! I also just discovered it. Turned out that "1.7.1-1" < "1.7-2" for homebrew. Please, try to update mixcr one more time brew update && brew upgrade mixcr.

Also, to correct wrong boundaries you should reimport segments from IMGT.

fabio-t commented 8 years ago

The boundary works fine.. mostly :) I got a few strange sequences that I'm not sure make sense:

1       GAAGTACAGGGCAGAGTCTGACAGCTGCGCTGATCAGCGCAGCTGTGATACGCGTTTTCGTGCTTCTTTTGTGTGTGGTTGGGGGTGTGGGGCACTGGGCACCTTTTTTTTGGATGTGAAACACAGTTGACAGTCATCCCAAACATCCAGACCCCCGAACCTGCTGTGTACCAGTTAAAAGATCCTCGGTCTCAGG
ACATCCCCCTCTACTCGTCTGAACTCCAGTCACTTTG       EVQGRV*QLR*SAQL*YAFSCFFCVWLGVWGTGHLFFGC_NTVDSHPKHPDPRTCCVPVKRSSVSGHPPLLV*TPVTL
1       TGTGCTTTGAGTGAATACTTCCAGAACCCAGAACATGCTATGTACCAGGTAAAAGATCCTCGGCATCAGGCCAACACCCTCTACACACTTCCACTCCAGTCACATCACGCTCACGTTTGCCGTCTTCTACGCAACCGCCACCAGTTAAATAGCTTGCAAATTACGTGGCCTTATGGCTACAGTATGCCCATCACAA
TTAGCAACAATCACGACACTTT      CALSEYFQNPEHAMYQVKDPRHQANTLYTLPLQSHH_SRLPSSTQPPPVK*LANYVALWLQYAHHN*QQSRHF
1       TGTGCTTTGAGGGACGGGCAAGACCGGTAAACTGAGGTCTTGGCTGGTGAGAACCGATCGATGGCAACCACTCATCTAGAACACCCAGAATCCGCGGTGCCCGAGAAAAAATCTCCGCGGTCACAGCACCGCACCCTACACGCGTCTGCTCCCGTCACACTACCTTATCGCGTGCGCCCTCTTCCGTTTAAAAAAA
AAAAACCTACAAAACTCTCTCT      CALRDGQDR*TEVLAGENRSMATTHLEHPESAVPEK_ISAVTAPHPTRVCSRHTTLSRAPSSV*KKKTYKTLS
1       TGTGCTCTGAGTCCGGACTTGAATATATGAAACTGCCTTTGCAAGTTTGACTCAGGTGCTAGACTGCACTGACATCCACAACCCAAGAACTCCGAAGTACCAGTTAACCGGTCCACAGATTCTCGGCGGCACCCTCTCCCCCTCTACACGCCTGAAACATAGCCACTTCCCATTCCCGCTTCCCCTTTTCTCCTAA
AAAAAAAAAAACATAACCTCT       CALSPDLNI*NCLCKFDSGARLH*HPQPKNSEVPVN_GPQILGGTLSPSTRLKHSHFPFPLPLFS*KKKNITS
1       TGTGCTCTGAGTGCTAGCAAGATGGGCTATAAAGTTACTGTAGGGTGAAGCACAAACTTGGTGGTTGATCCAACCATCCAGCCCCCAGACCCTACTGTGTACGAGTTAGATGATAGTCCTTCTCATGCCGGCCCCCTCTACCTGTCTGAATTCCAGTCACATCACGCTCGTTTTTGTCGGCTTATGCTTGCATCAA
AAAATAAACTTTAACCTTTG        CALSASKMGYKVTVG*STNLVVDPTIQPPDPTVYELDDSPSHAGPLYLSEFQSHHARFCRLMLASKNKL*PL
1       TGTCTCTCATCACGATCTCGTATGCCGTCTTCTGCTTGAAATAAAAAACACCAAAAATAGCAACGACACCAAAAGCAGAATAACCAACAACATCAGAAATCTAGTTAGAAGAACATCCAAAATAAGAAAAGACACACCACTAATCCTGAGTACGCTGATAAAGTAATAAAATCTCACGACTACTGCTACAATAAAA
AACTAAAAATAGGACATT  CLSSRSRMPSSA*NKKHQK*QRHQKQNNQQHQKSS*_RTSKIRKDTPLILSTLIK**NLTTTATIKN*K*DI
1       TGTGCTTTGAGACATCCAGAACCCAGAACCTGCTGTGTACCAGTTAAAAGATCCTCGGTCTCAGGACAGCACCCTCTACACGTCTGAACTCCAGTCACATCACGATCTCGTATGCCGTCTTCTGCTTGAAAAAAAAAAAAAAGTATCATCCAATACTTACTACATCTTCTATCAAAATATGTTTACTATCACTTCC
CT  CALRHPEPRTCCVPVKRSSVSGQHPLHV*TPVTSRSRMPSSA*KKKKSIIQYLLHLLSKYVYYHFP
1       TGTGCTTTGAGACATCCAGAACCCAGAACCTGCTGTGTACCAGTTAAAAGATCCTCGGTCTCAGGACAGCACCCTCTACACGTCTGAACTCCAGTCACATCACGATCTCGTATGCCGTCTTCTGCTTGAAAAAAAAAACCCAAAACAATATTCCATACTCTAGTTCTCTCAACTCAAAAAAAAATCTATAAAAG
      CALRHPEPRTCCVPVKRSSVSGQHPLHV*TPV_ITISYAVFCLKKKTQNNIPYSSSLNSKKNL*K
1       GTAGTACAGGGCAGAGTCTGACAGCTGCGCTGATCAGCGCAGCTGTCAGACTCTGCCCTGGTCTTCTATCCTCTCAGCCCTCCTGACCCCATACCCTGCTTTACAACATTTATCAGATCCTCGGTCTCATGACAGCCCCCTCTACACTTCTGAACACCATTCACTTTG        VVQGRV*QLR*SAQLSDSAL
VFYPLSPPDPIPCFTTFIRSSVS*QPPLHF*TPFTL

and also a few (haven't counted them, but not many over the total) that don't respect the boundaries:

2       TATGCTTTGAGGATGTCTAATTACAACGTGCTTTACCTC YALRMSNYNVLYL
2       CGTGCTTTGAGTGAGATTATGCCCAGGGATCAACCTTC  RALSEI_AQGSTF

Maybe it's worth to open another issue? Or am I doing something wrong?

mixcr align -f --library local -l TRA -s mm r1.fastq r2.fastq  alignments.vdjca
mixcr assemble -f -ObadQualityThreshold=10 alignments.vdjca clones.clns
mixcr exportClones -count -sequence -aaFeature CDR3 -f clones.clns clones.txt

dbolotin commented 8 years ago

Yes, it's worth to open a dedicated issue for this.

Can you please describe the type of sample you are analysing.

Long-CDR3 clones definitely look like false-positives. If the sample is not highly enriched, MiXCR can produce some fraction of false-positives. If you don't have adapter sequences in your data, you can try to use parameters for RNA-Seq (just add -p rna-seq on alignment stage), they are specifically optimised to achieve zero false-positive rate, and actually even have slightly better yield (they are just not so universal as default params).
As for clones not starting from Cys / ending with Phe, they don't look like wrong CDR3 boundaries, but more like PCR or sequencing (as you lowered the quality threshold) errors in corresponding triplets. In the light of this second issue, can you please also describe in some details your sample preparation protocol, because both issues could arise from some artefacts on this stage (I've seen lots of very complex sample preparation issues).

fabio-t commented 8 years ago

Thanks for the reply, I will try the rna-seq specific parameter. I cannot provide right the sample preparation (not done by our lab) but I might come back here in the future with a new issue.

For now I'll be using MIXCOR with all my data. Thanks for this great software.

milaboratory / mixcr

Crash when using "mm" as species name #69