Total reads in Illumina run

[nbr]

Hi folks,

I am new to high-throughput sequencing and my lab has generated ChIP-seq data using Illumina.
Does anybody know why would the total number of reads vary from sample to sample? Our first runs yielded about 17M reads. The last time, 2 samples had a total of 25M reads (one was input/control and another a TF).
Supposedly, the DNA concentrations are normalized for sequencing.
However, we do know that the 2 samples with more reads had initially much more DNA, so it is very tempting to believe that the normalization done was not that good.
Does anybody know of other reasons why one would obtain more reads in a lane?
Thanks in advance!

[james hadfield]

Hi,
The normalisation done before sequencing; i.e. quantitate each library and put it on a flowcell lane, is not completely straightforward. Cluster density is determined by the concentration of the library and the dilution performed on it to bring it down to around 10pM. If anything is not quite right then the actual concentration you end up with, and the clusters you see, will be different. This is why there is variation from lane to lane. On top of that there is also variation between flowcells in the amount of oligo available for sample hybridisation. If we run the same diluted PhiX on two different flow cells we get different cluster densities.

If you got over 15M reads then whoever ran your samples did a good, certainly adequate, job. Getting 25M reads from a lane is great, I wish we could get that all the time as well. 24M is our current best.

James.

[nbr]

Thank you for the reply, James, it was very informative.

Although we got many more reads in those 2 samples, we obtained 25% and 36% (input) of uniquely aligned reads. The other 2 samples that had total number of reads around 10M had 38% (TF) and 40% (histone marker) uniquely aligned reads.

The sample that yielded 25% seems to be a failed IP, or more likely, a bad antibody.
I wonder if this would be an explanation for the lower % of alignable reads and if not, what is possible to do to obtain higher numbers of uniquely aligned reads (specially on the IP end versus the sequencing end, to which protocol we do not have access).
Is it just a sample quality issue, or library construction can also be a major difference?

Thank you very much!

[der_eiskern]

Hi,
The normalisation done before sequencing; i.e. quantitate each library and put it on a flowcell lane, is not completely straightforward. Cluster density is determined by the concentration of the library and the dilution performed on it to bring it down to around 10pM. If anything is not quite right then the actual concentration you end up with, and the clusters you see, will be different. This is why there is variation from lane to lane. On top of that there is also variation between flowcells in the amount of oligo available for sample hybridisation. If we run the same diluted PhiX on two different flow cells we get different cluster densities.

If you got over 15M reads then whoever ran your samples did a good, certainly adequate, job. Getting 25M reads from a lane is great, I wish we could get that all the time as well. 24M is our current best.

James.

James, is your 24M current best from a single lane before or after passing filter? Just want to make sure I'm not comparing apples and oranges (i got between 20 and 27M before filtering with on average 70% of the reads passing filter.--14M-18M PF)