ChIP-seq of possibly high repeated sequences

[dawe]

Hi all, I have these ChIP-seq samples which may be enriched in high-repeated sequences... As many softwares for ChIP-seq rely on unique matches I wonder which could be a good strategy to analyze such samples....

Thanks

d

[mudshark]

Originally Posted by dawe

Hi all, I have these ChIP-seq samples which may be enriched in high-repeated sequences... As many softwares for ChIP-seq rely on unique matches I wonder which could be a good strategy to analyze such samples....

Thanks

d

hi d
if you are not explicitly interested in what binds the repeats, you just eliminate the non-unique sequences (and try to sequence sufficiently deep to get more than just repeptitive sequences).
if you are interested in targets that bind the repeats you will have problems to convincingly analyze the data (imho)
m

[dawe]

Originally Posted by mudshark

if you are interested in targets that bind the repeats you will have problems to convincingly analyze the data (imho)
m

That's exactly what I fear!
Thanks

d

[What_Da_Seq]

why don't you treat your sequences as de-novo assembly and try to align them with very very strict alignment parameters. This will hopefully result in many unique contigs that rely on small changes even within repeat regions. Then you just have to find the corresponding unique contigs between treatment and control and calculate the statistical difference.

[dawe]

Originally Posted by What_Da_Seq

why don't you treat your sequences as de-novo assembly and try to align them with very very strict alignment parameters. This will hopefully result in many unique contigs that rely on small changes even within repeat regions. Then you just have to find the corresponding unique contigs between treatment and control and calculate the statistical difference.

Mmm... that's an interesting approach... I'll try it for sure. Which assembler would you suggest for this task? I've tried once abyss but it didn't satisfy me. Velvet, maybe?

d

[Chipper]

What type of repeats do you expect to be enriched? Do you see any alignments at all in these regions? If your repeats are relatively short and you are planning to do more sequencing one way would be to select larger fragments and do additional shearing before library construction, this would perhaps give more unique alignments around the repeats.

[What_Da_Seq]

Sorry no experience with de-novo assembly only reference based assembly.

[What_Da_Seq]

Originally Posted by Chipper

... do additional shearing before library construction, this would perhaps give more unique alignments around the repeats.

I do not follow. Don't you want to do LESS shearing to get larger fragments?

[dawe]

Originally Posted by Chipper

What type of repeats do you expect to be enriched? Do you see any alignments at all in these regions? If your repeats are relatively short and you are planning to do more sequencing one way would be to select larger fragments and do additional shearing before library construction, this would perhaps give more unique alignments around the repeats.

The person who gave us the sample claims that there may be enrichment in rDNA sequences (mouse). I've tried to align reads on rDNA sequence only and I get ~0.15 % reads aligned (which seems to me pretty high for a 45kbp sequence) in IP. I've just excluded the matching reads to align them on the rest of the genome (just to see what's happening).
I've also noticed that rDNA sequence are not annotated on mouse genome (and they're expected to be ~400 copies scattered around).

d