Comparison between replicates

[ktroule]

Hi

I'm starting to play with mixcr to test I can run it without problems and I would like few things to know whether you can give me some hints.

I have some RNA samples across four different conditions (3 biological replicates for each condition, 12 in total). I have input each sample independently.
java -Xmx20g -jar /home//Programs/mixcr-3.0.4/mixcr.jar analyze shotgun "$data""${fastq["$index"]}" "${saveMixcr["$index"]}" -s MusMusculus --starting-material rna --receptor-type xcr --verbose

First thing I notice is that the number of lines in the All clonotype files does not match the number of lines of the rest of clonotypes files.

   491 Analysis_1.clonotypes.ALL.txt
   127 Analysis_1.clonotypes.IGH.txt
   297 Analysis_1.clonotypes.IGK.txt
    18 Analysis_1.clonotypes.IGL.txt
    23 Analysis_1.clonotypes.TRA.txt
    30 Analysis_1.clonotypes.TRB.txt
    23 Analysis_1.clonotypes.TRD.txt
     1 Analysis_1.clonotypes.TRG.txt

Can you suggest a way to compare differences in clonotypes between different conditions?

Thanks for your help.

[dbolotin]

The problem is in TRD/TRA files. Because of the TRA/TRD locus structure it is hard to tell TRA from TRD TCRs because they both uses the same V and J genes. The only reliable method is by looking at C gene, but it is possible only with RNA and with certain sample preparation protocols that leave enough number of C gene letters (outside the primer sequence, which might introduce chimeric artifacts). Or, in case of RNA-Seq, only for alignments covering C gene region.

This will be fixed in 3.1 version by outputting single TRAD file instead of separate TRA and TRD.

For the comparison method between conditions, I suggest you look into recent paper by Pogorelyy et al (@pogorely): https://elifesciences.org/articles/33050 But I'am afraid RNA-Seq data can't give you enough quantitative information to find the difference (unless it is RNA-Seq from sorted lymphocyte populations).

[ktroule]

Thanks, I'll check the paper.

Thanks for your help.

[ktroule]

To create the table that serves as input for the vdjRec tool identical CDR3 from different VJ combinations must be separated to perform the analysis. Wont you have a script to convert mixcr output to vdjRec input (as if not wrong both tools come from same lab).

Once more, thanks for your help,

[dbolotin]

Sorry, unfortunately we don't have such script. To separate clonotypes by V/J genes please use -OseparateByV=true (see here).