Read Length to Gigabase conversion

[yoyoming1001]

Really dumb question: how do you mathematically convert reads into gigabases and then calculate coverage?

if I have 600 million clusters sequenced at 2*100, then is it 600*200=120,000 gigabases sequenced? Or is that megabases?

Then do you divide 120,000/3,000 (3 billion bases in human genome) = 40. So 40x coverage?

I ask because I'm trying to figure out how many clusters/lane I need to get when sequencing a whole genome, and then how many lanes our lab would need to run. Would also be curious what people are doing typically nowadays.

[krobison]

2x100 yields 200 = 2x10^2
600M clusters = 6x10^8 clusters
so 1.2x10^11 bases
human genome approximately 3x10^9

1.2x10^11 / 3x10^9 = 0.4x10^2 or 4x10^1 so yes 40X coverage

[yoyoming1001]

Thanks krobison. Good to see the math confirmed. It's a little odd to me since I thought labs had to push cluster densities on the HiSeq a bit in order to get 30x coverage in three lanes, but that's only 150,000 clusters PF per lane, which isn't all that much nowadays. Am I missing something?

[kmcarr]

...but that's only 150,000 clusters PF per lane, which isn't all that much nowadays. Am I missing something?

Yes, three orders of magnitude. . (I know it was probably just a typo.)

You need 150 million (150,000,000) PF clusters @ 2x100bp per lane to get 30X (90Gbp) in three lanes.

Illumina specifications are for 187.5 million PF clusters per lane; in practice this can be pushed to 200-220 million PF clusters per lane with minimal loss of quality.

[yoyoming1001]

Haha, yeah, I knew it in my head but skipped over it when writing it down.

I wasn't sure if there were any bioinformatics factors that would decrease the usable reads and therefore the lab would need to push higher on average.

Thanks though! My understanding is a lot more clearer now.