Importance of narrow size selection?

[sotoole]

Hello Everyone,

This is my first post in this forum. Hooray!! Anyways, on to my question:

My lab is just beginning to get into RNA sequencing and we are currently working on our library preparation technique for sequencing using the illumina platform (GA2). In my travels through the literature on this topic I have noticed that several papers have suggested that during the post-amplification size selection of the library preparation, the band size should be relatively narrow. In other words if you are sequencing with 300 bp fragments it is better to choose a range of 300bp +/-50bp as opposed to +/-100 bp. Why is this? Is it okay to size select for the full range of fragment sizes that the illumina platform is capable of (I believe this is 100 bp to 600 bp)? If the answer is no then what are the disadvantages to doing this?

Any insight would be greatly appreciated.

Cheers,
Eager Scientist

[james hadfield]

There are two major reasons I can think of why tight size selection is important. One technical, one bioinformatic.

1: The bridge amplification is more efficient with shorter fragments, if your library is a mix you might end up biasing towards those short fragments.

2: If you are running Paired End sequencing to look at splicing then you need to use the size distribution of your fragments to call when exons are skipped.

[Thorondor]

and it is crucial for most assembly tools (especially if you want to do scaffolding) to set the parameter "insert length" so if the insert lengths vary a lot bioinformatics won't be happy. ;-)

[HESmith]

Another technical reason: the tight distribution minimizes size-dependent bias during library amplification.

[sotoole]

I would just like to thank everyone for the answers they gave to my question. I considered these replies very useful.

[captainentropy]

+1 for james hadfield's answer.

Also, according to Illumina, due to the limitations in the image analysis software, the longer the fragments the more irregular the "spots" (in four colors representing A, T, C, and G) are on the image. The more irregular they are the more likely the software will ignore them.

That said I make libraries from chromatin sheared to ~200-600 bps, and the results are awesome. But that's for ChIP-seq, not RNA-seq.