variable sized amplicons -bias?

[BAME]

Hi
I am doing some amplicon sequencing on the Miseq (300b.p pair end runs) and wish to run pools of mixed size amplicons (from 7 different genes) that vary in length from approx 379-750 b.p (as below) on the same run.

I understand that there is a bias towards smaller amplicons preferentially binding to the flow cell. If so how much bias towards the smaller amplicon (379 bp) is there likely to be and should I take account of it when pooling the DNA from the 7 different amplicons (or can I just pool them together in equal concentrations)? Id like to get roughly equal coverage of the different amplicons if possible.

amplicon lengths b.p
gene 1) 379
gene 2) 561
gene 3) 642
gene 4) 648
gene 5) 671
gene 6) 683
gene 7) 750

cheers for any advice

[Jessica D.]

You definitely won't want to pool together in equal concentrations, otherwise you'll find that your larger fragments will be proportionately underrepresented.

The concentration is partly a function of fragment length, so a 2nM library at 250bp is going to have a higher absolute number of fragments than a 2nM library at 550bp. My suggestion for normalizing if you need to pool all those sizes together would be to calculate your number of molecules present (based on concentrations and fragment length) and then try to normalize by number of molecules.

That should get you close but my experience- at least partly due to stochastic variation in template hyb and bridge amplification- mean that unfortunately, it's not quite as simple as just doubling the amount of input for a fragment that's twice the length.

[austinso]

Concentration in terms of molarity is a direct measure of the number of molecules per unit volume, so is independent of fragment length.

Concentration in terms of mass, on the other hand, is a function of fragment length. i.e. a molecule of 500 bases is heavier than a molecule of 100 bases.

Binding to the flow cell is some function of fragment length (charge density and stiffness), but it is hard to predict.

I'd say that you should just mix equal amounts, and you get what you get as far as coverage. If they are indexed separately, then you could come up with a correction function for normalization based on expected length.

However, by far the greatest contributor to variability in coverage will be the (in)accuracy of your method of determining concentration.

[Jessica D.]

Oh, right. I mean, you size correct the qPCR data when you're analyzing, so the molarity measurements must be independent of fragment length. I guess that wasn't explained to me in the right way, then. Would you blame all the variation on template hybridization efficiency as a function of charge density then, or is(are) there some other factor(s) at work?

What I have noticed when trying to pool libraries of disparate sizes is that there are pretty drastic differences and larger fragments tend to be underrepresented in the final data set. If I was generally loading 12.5pM on a flow cell, I might bump a slightly larger library to 14pM, assuming the bp difference was relatively minor (like 375bp vs 250bp). For a size range of 350 to 750, it's harder to say how much to change the loading concentration. My FAS had explained to me a long time ago that clustering wasn't a linear function so it's always been a little trial and error.

Since most of BAME's fragments are larger, it might make more sense for them to load all the larger amplicons at one concentration and take the one library at 379bp down one or two pM, just to try and even things out. This is what's worked for me in the past, but of course, YMMV.