Clip truseq adapter sequence from fastq

[uniportdb]

Here's my public history https://main.g2.bx.psu.edu/u/davidkim/h/ngsworkshop

I am trying to learn NGS using the public data https://www.ncbi.nlm.nih.gov/geo/que...i?acc=GSE39083

Fastqc on EBI_SRA__SRP014008_File__SRR518493.fastq.gz__1
found TruSeq adapters in the reads.
Sequence Count Percentage Possible Source
GATCGGAAGAGCACACGTCTGAACTCCAGTCACCCATCATAT 347588 3.340743581 TruSeq Adapter, Index 2 (97% over 37bp)
CGGCAGTCCACTCCGGTACGCTATCCCACTACTGCCTACCAC 159445 1.532460443 No Hit (possibly ncRNA? )

I did 'clip sequence' (with default setting and output only non-clipped sequences) and then I got only 83% of reads. But the fastqc showed only 3.3% were the truseq adapter sequences. Why did 'clip sequence' tool discard 17% of reads?

Clipping Adapter GATCGGAAGAGCACACGTCTGAACTCCAGTCACCCATCATAT
Input 10404510 reads.
Output 8683051 reads.
discarded 451106 too-short reads.
discarded 386260 adapter-only reads.
discarded 868903 clipped reads.

Thanks,

[fkrueger]

FastQC flags up sequences that make up more than 0.1% of your total library and it does so by looking for exact matches of the entire sequence. If you have adapter contamination at various positions in your read the total sequence is likely to be always different, or maybe just doesn't quite add up to 0.1%. The k-mer plot may sometimes help you identify adapter contamination that starts at various positions in the read. If you use tools like Cutadapt or Trim Galore you will get a histogram-type output that shows you exactly which part of the adapter was found and removed.