Real performance of HiSeq 2000

[cip]

Hello, what is the 'real world' performance of HiSeq 2000 using 2x100 PE?

Illumina says 374 million/lane; so 374e6 x 2 x 100 = 74.80 Gb/lane.

According to Illumina: 80% percentage of Bases > Q30

74.80*0.80 =~ 60Gb.

However looking in web pages of sequencing centers the 'real value' ranges between 40-60Gb/lane.

Thank you very much.

[MLog]

When they say 374 million/lane, they mean 374 millions of 100bp reads, or 187 millions of 100+100 reads. This gives ~ 37 Gb. In our hands, we usually get 30-40 Gb, with Q30 ~90%.
It is possible to get a bit more, 40-45 Gb, but I would not recommend this.

[simonandrews]

We would normally hope to get 200 million reads from a lane these days. We've had quite a few runs from the Sanger centre lately which have been averaging nearer 100 million though. On a good day you can get 250+ million so the yield you get can vary by a factor of 3 depending on how optimally everything worked. Quality to the end of a 2x100bp run is pretty consistently good these days and we remove very little by quality trimming.

[cip]

Hi simonandrews, thank you.

When you say 200 million reads, do you mean paired reads (100 + 100) or single reads (100bps)?

Thanks.

[simonandrews]

200 million clusters, so 400 million reads from a 2x100 paired end run.

[GenoMax]

The numbers Simon is presenting are for well constructed libraries.

Every library run on a HiSeq does not yield this many reads per lane.

[cip]

Ok I think I understand you now. But please correct me if I am wrong. So if you have 400 million reads from a 2x100 paired end run (from 200 million clusters) ... You would get 40Gb from a lane?

However your output from a lane can range between 20-50 Gb (100-250 million reads) depending on how optimally everything worked.

[simonandrews]

The numbers Simon is presenting are for well constructed libraries.

We only make well constructed libraries

If only...

[GenoMax]

Ok I think I understand you now. But please correct me if I am wrong. So if you have 400 million reads from a 2x100 paired end run (from 200 million clusters) ... You would get 40Gb from a lane?

However your output from a lane can range between 20-50 Gb (100-250 million reads) depending on how optimally everything worked.

Output will range from 0 - 50 Gb (specially if you had not QC'ed your library before running it)

That aside the estimate you are using is reasonable.