Nextera vs Illumina Library Prep

[maggie]

Does anyone have any information about performance of Nextera multiplex library prep versus the Illumina method for sequencing on the Illumina platform?

[Turnerac0987]

That's actually a question I've been wondering too. Did you ever get any info on that? Can anyone else give any insight?

[husamia]

Let me just make sure I am on the right page before giving my opinion. your referring to Nextera vs. TruSeq prep kit, correct?
if so, I think there is difference since it involves the adding of adapters. I cant give more details since I am still investigating. But in general I can say nextera is more random than trueseq when looking at the alignment. In terms of overall coverage I don't see significant difference.

[kmkocot]

Are you sequencing genomic DNA?

[maggie]

yes, sequencing genomic dna and/or hybrid captured genomic dna. i'm wondering if there is bias in nextera vs truseq. otherwise, why not use nextera?

[steinmann]

But in general I can say nextera is more random than trueseq

By that you mean that the library is more complex and you have more unique fragments?

[koadman]

we have been using epicentre's nextera pretty extensively in our lab. these are the pre-Illumina takeover kits from epicentre, so I can't comment on whether Illumina's revised kits behave differently. Without a doubt there is a greater degree of fragmentation site bias than typically observed with mechanical shearing. We see this in our own data, but you can also see this in the supplementary material of Adey et al 2010 . Look for the logo plots of nucleotide content near the fragmentation site. That paper also presents numbers for library complexity which at first seem to indicate that transposon-catalyzed libraries have complexity nearly as high as mechanically sheared libraries. However, they compute library complexity as a function of unique start sites for both reads in a pair in libraries with a broad fragment size distribution. This approach allows nextera libraries to appear to have much higher complexity than if the complexity were calculated using start sites of individual reads (not paired).

That said, there do seem to be ways to boost the complexity in tnp-catalyzed libraries. We have found that if the tagmentation is carried out with an excess of transposase relative to target DNA, we obtain a fragment size distribution skewed toward very small fragments. This result was pretty well characterized by others. However, if we then size select to keep only the larger fragment range, the resulting libraries have higher complexity. My intuition as to why this works is that the preferred target sites for tnp are more likely to be in small fragments, and cutting those fragments out helps to normalize the distribution of target sites. Hopefully we'll get this organized into a peer reviewed publication sometime soon but for now it's just anecdote.

If husamia's comment about nextera being more random than truseq is correct, then illumina/epicentre must have made some dramatic improvements in their enzyme chemistry. it would be great to see a head-to-head comparison of each method.