question about Trimmomatic

[dejavu2010]

I have a PE 100 reads from hiseq 2000, 100PE reads =166,867,542 PE reads. when i opened it by fastqc, it only shows total sequence:16000000. Anyway, I just wanted to eliminate adapter sequences by Trimmomatic, so I use:

java -classpath ~/Software/Trimmomatic-0.17/trimmomatic-0.17.jar org.usadellab.trimmomatic.TrimmomaticPE -phred33 1145LJZ0007_read1.txt 1145LJZ0007_read2.txt reads_fwd_paired.fq.gz reads_fwd_UNpaired.fq.gz reads_rev_paired.fq.gz reads_rev_UNpaired.fq.gz ILLUMINACLIP:illuminaContaminants.fa:2:40:15

I ended up with 15845809. I checked reads_fwd_paired.fq.gz, the last line is HWI-ST829:1260CWTABXX:5:1104:6650:64208 1:N:0:..., which means the last read from tile 1104. This means you just had 4 tiles processed. Usually there are 48 tiles per lane. Therefore only 10% reads got processed. I wonder if Trimmomatic has 16000000 read limit, which CASAVA has it for ELAND. If that is the case, where does this read limit comes from and how to bypass this problem. Thanks.

[arvid]

I can see more than 16 Mio reads in FastQC and filter more than that with Trimmomatic. Are you completely sure that your input file isn't truncated? What does

grep -c "^@HWI" 1145LJZ0007_read1.txt

tell you?

[tonybolger]

I have a PE 100 reads from hiseq 2000, 100PE reads =166,867,542 PE reads. when i opened it by fastqc, it only shows total sequence:16000000.

I agree with arvid - the original file is probably truncated for some reason. Neither tool is limited to 16m reads.

Incidentally, the new version of trimmomatic gives you the input / output counts when it runs successfully, which will help indicate if it is dying early for some reason.

[dejavu2010]

I found the actual cause for that problem. Since when we demultiplex sample, we limit reads number 16000000, and gunzip. Later on we concatenate them. Since Trimmomatic is a JAVA program which, I guess, uses java based gzip decompressor. The gzip decompressor in java only reads the first gzip header in a file and stops extracting data when the block reported there is complete, even if there is another header later in the file.

[dejavu2010]

1145LJZ0007_read1.txt is not original. The original is .txt.gz. sorry for confusion.