Have you tried FastQC? Adapter dimers might show up in the overrepresented sequences section of the report, and the overrepresented k-mer section may give you some additional clues.

In the FastQC source, FastQC/Contaminants/contaminant_list.txt has a convenient list of adapters that you can use as a starting point for further analysis.

I have a cheap little perl script called find_3p_adapter.pl I wrote for precisely this purpose. It works by taking in the sequence of a microRNA that you know must be present at pretty high levels in your sample, along with the raw untrimmed FASTQ file. For instance, in a human brain, miR-124 would be a good bet. The script will search for all occurrences of the miRNA query sequence, and track the 'suffiix' (or suffices) that comes after it. That will tell you the adapter sequence.

Here's an example with some plant data (so I've used miR156 as the query, b/c it's pretty abundant in most plant tissues):

Algonquin:raw michaelaxtell$ gzip -d -c Apr1A_R1.fastq.gz | ./find_3p_adapter.pl -m ugacagaagagagugagcac
./find_3p_adapter.pl version 0.2
Thu Aug  8 16:12:12 EDT 2013
directory: /Users/michaelaxtell/data/sRNAseq_data/Physcomitrella_patens/HiSeq2500_Apr22_2013/raw
Query sequence: TGACAGAAGAGAGTGAGCAC

    Searching...Done

20754 out of 23706986 reads matched query TGACAGAAGAGAGTGAGCAC (875 reads per million)

Here are the top four adapters found:
Sequence    Frequency
TGGAATTCTCGGGTGCCAAGGAACTCCAGT    17771    85.627 %
CTGGAATTCTCGGGTGCCAAGGAACTCCAG    937    4.515 %
ATGGAATTCTCGGGTGCCAAGGAACTCCAG    911    4.390 %
TTGGAATTCTCGGGTGCCAAGGAACTCCAG    358    1.725 %

So in the example above, its a good bet that the adapter starts with "TGGAATTCTCG"

You can just take a look at the SAM or FASTQ and see if you can identify a common sequence occurring in all the reads. The human brain is pretty great at pattern finding.

Print out reads for SAM:

samtools view -S file.sam | grep -o '[GATCN]\{30,300\}'

Print out reads for FASTQ:

grep -o '[GATCN]\{30,300\}' file.fastq

Thanks Daler, Matt Shirley and Arun for the help. I misunderstood the bit about FastQC (I thought my pipeline should produce QC that I can check). I downloaded the FastQC tool and run one of my fastq files and a screenprint of the result is here

My fastq sample sequenced with 36bp long reads, I want to trim the `3p end and I do not know which of the overrepresented sequences is the true adapter. Any suggestion ?

Thanks again.