search

GuoBioinfoLab / CATT

An ultra-sensitive and precise tool for characterizing T cell CDR3 sequences in TCR-seq and RNA-seq data.
http://bioinfo.life.hust.edu.cn/CATT/
20 stars 4 forks source link

Potential bug when using paired-end files #17

Open mapo9 opened 10 months ago

mapo9 commented 10 months ago

Hi, I found some weird behaviour when running paired-end data.

To test some stuff, I created some simulated datasets where I know all the parameters like repertoire size, size of each clonotype, V gene, J gene etc. I created the fastq's as paired-end seq files and ran catt with the following command catt --f1 test_R1.fastq --f2 test_R2.fastq -o test_out -t 20.

The unintended behaviour I experienced can nicely be seen in one of my samples with a repertoire of one clonotype with 10.000 clones. Catt returns 3 clones with exactly equal NNseq: AAseq,NNseq,Prob,Vregion,Jregion,Dregion,Frequency CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*02,TRBJ2-1*01,TRBD1*01,6727 CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*06,TRBJ2-1*01,TRBD1*01,6727 CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*03,TRBJ2-1*01,TRBD1*01,6546

When combining the "different" clonotypes into one the frequencies sum up to 20.000 clones instead of 10.000. So, it seems like catt is counting each clone twice

I thus merged the paired end files to a single file using pear and repeated the analysis. This returned the same results as the paired-end run, only the frequencies are different. AAseq,NNseq,Prob,Vregion,Jregion,Dregion,Frequency CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*06,TRBJ2-1*01,TRBD1*01,3379 CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*02,TRBJ2-1*01,TRBD1*01,3348 CSARGDRGLSYNEQFF,TGCAGTGCTCGGGGGGACAGGGGGCTATCCTACAATGAGCAGTTCTTC,0.00016,TRBV20-1*03,TRBJ2-1*01,TRBD1*01,3273 For the merged "single-end" files the clones sum up to the expected 10.000. What confuses me a little though that the counts aren't exactly half of the ones in the paired-end run

I guess that there must be some issue when counting the frequency for the paired-end samples.

Would be awesome if you could have a look! Thanks!

songofbin commented 2 weeks ago

If I understand correctly, the definition of a TCR clone in the CATT results takes into account the differences in V and J genes in addition to cdr3aa. The problem is that the delineation of V and J genes may sometimes be inaccurate, resulting in the separation of a same TCR.