fastqc shows strange GC content

[frymor]

Hi,

we have a 36-samples ChIP-Seq experiment with a very strange GC-content behavior (s. images)
The samples are from yeast. This strange behavior comes in the IP samples, some of them shows (the first image is from such a file) in the fastqc report an over-representative accumulation of certain sequences, such as

GGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG
TGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG	
TTGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG
AGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG

But some of them don't ( the second image is from a file with no(!) over-rep. sequences at all.
Other than that the report looks quite good with no abnormalities (except the GC content).

I was thinking that this curve might be due to the over-rep. polyG sequences, but what about the samples where there is no apparent accumulation of such sequences. Is there a possible biological reason for this thing?

The problem occurs only in the IP samples. Is it possible that lack of biological material for the sequencer for this samples causes the machine to put a G (or identify a G ) intread of the real nt?
we are using a nextseq500 with a high kit from Illumina

thanks
Assa

[frymor]

I just wanted to add that this case is even stranger, as i have this sample, where the GC content is classified as good, even though it is completely shifted to the left.

Q: How does fastqc calculate the theoretical GC content?

[nucacidhunter]

The plots that you are referring should show normal distribution around the GC content of genome in a random library. But IP is not expected to be random so there would be a bias. Extreme GC at the start of reads could also be due to library construction method where some non-template bases are added.

Stretches of G in NextSeq data indicates that there was not any signal in those cycles which could be due to short inserts and adapter dimers. F

From FastQC manual:

“This module measures the GC content across the whole length of each sequence in a file and compares it to a modelled normal distribution of GC content.”

“An unusually shaped distribution could indicate a contaminated library or some other kinds of biased subset. A normal distribution which is shifted indicates some systematic bias which is independent of base position. If there is a systematic bias which creates a shifted normal distribution then this won't be flagged as an error by the module since it doesn't know what your genome's GC content should be.”

[kmcarr]

Hi,

we have a 36-samples ChIP-Seq experiment with a very strange GC-content behavior (s. images)
The samples are from yeast. This strange behavior comes in the IP samples, some of them shows (the first image is from such a file) in the fastqc report an over-representative accumulation of certain sequences, such as

GGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG
TGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG	
TTGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG
AGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG

But some of them don't ( the second image is from a file with no(!) over-rep. sequences at all.
Other than that the report looks quite good with no abnormalities (except the GC content).

I was thinking that this curve might be due to the over-rep. polyG sequences, but what about the samples where there is no apparent accumulation of such sequences. Is there a possible biological reason for this thing?

The problem occurs only in the IP samples. Is it possible that lack of biological material for the sequencer for this samples causes the machine to put a G (or identify a G ) intread of the real nt?
we are using a nextseq500 with a high kit from Illumina

thanks
Assa

frymor,

You correctly identified the cause in your last statement above. The Illumina NextSeq500 (and NovaSeq) use 2 Color Chemistry for tagging and identifying bases. Using this system G's are defined by the absence of a fluorescent signal. This means that anything which can cause a cluster to stop producing a signal with result in a polyG output. A much better and more thorough explanation can be found in this QC Fail post.

[nucacidhunter]

Stretches of G in NextSeq indicates lack of signal in those cycles and it could be result of short inserts and adapter-dimer.

In a shot gun library GC content peak will be representative of genome GC but IP pull down is not random and some bias is expected depending on what sequences are pulled down and enriched.

From FastQC manual:

"This module measures the GC content across the whole length of each sequence in a file and compares it to a modelled normal distribution of GC content."

"An unusually shaped distribution could indicate a contaminated library or some other kinds of biased subset. A normal distribution which is shifted indicates some systematic bias which is independent of base position. If there is a systematic bias which creates a shifted normal distribution then this won't be flagged as an error by the module since it doesn't know what your genome's GC content should be."