how to remove adaptor sequence which has more than 1 indexing barcode?

Question

[sckinta]

Illumina TruSeq® v1/v2/LT Sample Prep have 1 universal adaptor and 27 indexed adaptors. I have 45 samples pooled together to sequence. Each sample has been indexed with two barcodes. The Demultiplex_Stats.htm gives report like

1 10 tmk_lab TCCGCGAA-TATAGCCT 10 N DefaultProject 422 100.00 2,811,299 1.66 100.00 0.00 81.63 32.34
1 11 tmk_lab TCTCGCGC-TATAGCCT 11 N DefaultProject 483 100.00 3,218,494 1.91 100.00 0.00 81.01 32.19
1 12 tmk_lab AGCGATAG-TATAGCCT 12 N DefaultProject 817 100.00 5,443,850 3.22 100.00 0.00 81.86 32.37

In which second column means sample name while, the forth column means sequences of two barcodes.

According to the post, the barcode flanks indexed adaptor common sequences. Like

"GATCGGAAGAGCACACGTCTGAACTCCAGTCACTCCGCGAAATCTCGTATGCCGTCTTCTGCTTG"
"GATCGGAAGAGCACACGTCTGAACTCCAGTCACAGCGATAGATCTCGTATGCCGTCTTCTGCTTG"
 #                                ^------^

But here I have two barcode for each sample, Should I put two barcodes side-by-side and insert them together into indexed adaptor common sequences? like

>adaptor1
GATCGGAAGAGCACACGTCTGAACTCCAGTCACTCCGCGAATATAGCCTATCTCGTATGCCGTCTTCTGCTTG
#                                ^--------------^

Or insert each barcode into indexed adaptor common sequences and treat them as separated adaptors? like

>adaptor1
GATCGGAAGAGCACACGTCTGAACTCCAGTCACTCCGCGAAATCTCGTATGCCGTCTTCTGCTTG
#                                ^------^
>adaptor2
GATCGGAAGAGCACACGTCTGAACTCCAGTCACTATAGCCTATCTCGTATGCCGTCTTCTGCTTG
#                                ^------^

I personally think first one is right, but I have no idea whether they have flanking sequence between two barcodes.

Anyone has experiences with it?

Answer 1

[Brian Bushnell]

Normally, dual indexes have one barcode per end of the read, as in your second example. Don't modify the adapter sequence by squeezing two barcodes together inside a single adapter.

Ultimately, though, if you adapter-trim via kmer matching (with BBDuk), the bar code sequence ends up not being important as long as K is at most the length of the shorter common sequence (in this case 24).