Confused about read duplicates after adapter ligation

[hujian5241]

Recently I am working on developing new methods for NGS PE library preparation. I am confused about PCR duplicates. It is obvious that duplicates will arise during PCR amplification dealing with low DNA input. How ever, when I think about PCR-free library preparation, one double-stranded DNA molecule is ligated 2 Y adapters at both left and right side. This results in 2 different single strand products:
1. 5'-P5 - plus strand insert ssDNA -P7'-3'
2. 3'-P7'- minus strand insert ssDNA -P5-5'

These 2 single strand products are actually duplicates since the insert ssDNA are fully complementary to each other and they could be both ligated to flow cell. This means even for PCR-free library, 50% of the reads are duplicates, theoretically both strand from one ds-DNA could be sequenced. however, when I dealt with fastq file after sequencing, the percentage of duplicates was much lower than 50%, as for PCR-free library, there was nearly no duplicates. Can anybody help me about this ?

[pmiguel]

One read pair will be the reverse-complement of the other read-pair. So they will not be considered "duplicates".
Also, even if your software did consider these to be "duplicates" not all amplicon strands cluster. Various flowcells have different efficiencies.

If you cluster about 100ul of a 20pM library in one lane of a HiSeq 2500 you would get about 150-220 million pass filter clusters. To a first approximation "pM" is millions of molecules/ul. So about 2 billion molecules of your library would go into a HiSeq 2500 lane. So only about 20% actually cluster. So your chance of clustering both strands from a given amplicon would be 4%.

I think most other Illumina instruments cluster at a lower efficiency than than the HiSeq 2500.

--
Phillip