Is There Any Advantage Of Paired End Sequencing For Chip-Seq?

Question

[Fidel]

Hi,

I hope that some of you may have a recommendation regarding the use of paired-end sequencing over single-end for ChIP-seq. In principle we expect only a minor improvement of using paired-end sequencing for ChIP-seq analysis over single-end. Since paired-end sequencing is more expensive and the processing of the data takes longer time we will like to confidently decide when is worth to use paired-end and when is not.

In the past, we noticed that in the case of ChIPs targeting long repetitive regions the use of paired-end sequencing helped significantly to identify enrichments where single reads will not map uniquely. Off course we only got enrichments at the flanks of the repetitive regions.

We are about to run some tests to compare paired-end vs. single-end for more 'standard' cases but it never hurts to address the collective knowledge for advice.

Thanks,

Answer 1

[Istvan Albert]

As our understanding of the genome advances we start to appreciate just how important the accuracy of the measurements is.

The extra cost associated with paired end sequencing is negligible compared to the overall cost of sample preparation and analysis. And one never knows what the data holds and what unexpected measures will be required to settle a point, some that may be impossible to obtain on single ended data.

I always recommend to use paired end sequencing unless one can clearly quantify the disadvantages and those are overwhelming. You can always analyze these in single end mode but not vice-versa.

Answer 2

[Sukhi Singh]

I also think paired end in not that advantageous considering the resources (time, money) involved in terms of ChIP-Seq. The majority of the noise in the ChIP-Seq depends on the protocol used and the specific steps like

• cross-linking and elution times
• use of salmon sperm
• double cross-linking (first linking protein and protein, then protein and DNA)

Biological replicates could also work (we tried but variations in the protocol, could never confirm the same results), which can be tested by the use of tools like IDR: Reproducibility and automatic thresholding of ChIP-seq data

Btw, there are very good points at the Useq webpage about ChIP recommendations under the

Benchtop ChIP Recommendations

Another discussion post at Seqanswers

Answer 3

[Ian]

One thing paired-end reads adds to ChIP-seq analysis is a more accurate estimation of mean fragment length. MACS 1.4.2 is not very good at predicting this, although I believe MACS 2.0.10 is better as it uses cross-correlation. As Istvan says paired end is a good option IF the money is available. For ChIP-seq I don't think the advantages balance the extra cost.

MaSC was recently released that more accurately estimates mean fragment length for genomes where mappability scores are available. You can start of by reading my mini-review Masc - A Tool For Calculating Mean Fragment Size For Chip-Seq Peak Calling Analysis if you are interested.