insert size bias

[superpyrin]

We have 3 different lengths of inserts: ~700, ~500 and ~275 bp and we consider mixing them in one lane in HiSeq. Is it known, or is it possible to estimate what is the bias in favor of shorter inserts? Is the 700 bp insert going to be completely overwhelmed by shorter ones?

[pmiguel]

My advice:
You are better off putting the 700's in another lane. But if that isn't a possibility then, do qPCR to determine their concentrations. Then based on those results make a pool that has 2x more 700 than 500 and 0.5x 275 than 500.

57:29:14 ratios for 700:500:275 insert sizes.

I'm mainly going by intuition here, though. I might be over-correcting or under-correcting...

--
Phillip

[superpyrin]

Thank you. Could you elaborate or refer me to literature on what is the reasoning behind your suggestion? Is there some model that provides predictions how the bias is correlated with the insert length?

[pmiguel]

Thank you. Could you elaborate or refer me to literature on what is the reasoning behind your suggestion? Is there some model that provides predictions how the bias is correlated with the insert length?

Like I wrote above, this is my intuition based on our experience with the instrument.

--
Phillip

[GenoMax]

I second Phillip's intuition. Our experience is similar. In practice smaller inserts always out-compete larger ones when it comes to clustering.

[MU Core]

superpyrin, the bias toward the small insert size has been our own personal experience as well. I've observed a 20% skewing of the read count toward the smaller fragment size (350 vs. 550) when both library were loaded at equal ratios. You'll want to perform a pooling approach as pmiguel suggest if pooling all three libraries together.