which is better, paired or single end

[analyst]

Can anyone please answer these

For chip-seq and RNA-seq, which is better PE or SE

In one lane in illumina, one gets ~15 GB sequence which is 5X for human. Is this enough for chip-seq and RNA seq, or should one run multiple lanes.

Is it good idea to multiplex in the above experiment if budget is constraint.

Thanks in advance for you valuable time. It will be a big help really.

[edawad]

PE is almost always better for RNA-seq--you gain more information about splice junctions etc. It doesn't hurt for ChIP-Seq and may help you better identify enriched binding sites in repetitive regions (although some mappers may not be able to handle PE tags).

For most ChIP-Seq applications 10-15 million reads is enough (unless it's a histone or highly ubiquitous TF). You can computationally determine your ChIP-Seq coverage/saturation using a program like MACS --diag option. I can't say much for RNA-Seq but somewhere out there I've seen a table with suggested sequence coverage. RNA-Seq probably requires more reads than ChIP-Seq, moreso if you plan on getting quantitative information out of it.

PE for RNA Seq
SE for ChIP Seq. There really isn't the need for Paired reads.

the throughput really varies on the experiment. Histone modifications or TF analysis?

[sciencewu]

SE is better for you .

[bioinfosm]

SE is faster and cheaper... PE on the other hand is more data, and theoretically more efficient, provided you use the appropriate methods to make use of paired information

[yxibcm]

Actually, I would suggest using PE for ChIP-seq too. For SE reads, existing ChIP-seq software, such as MACS, shift and extend the reads to build the whole genome profile. The shift and extend distance was a fixed value estimated from the double peak pattern or provided by command line parameters. This might be inaccurate if the wrong shift/extend distance were used, and might cause the peak appear as doublets. So the peak heights are sensitive to the shift/extend values. PE sequencing provides fragment size information, therefore build whole genome profile is straightforward, no shift or extend involved. PE is also not that expensive compared with SE.

Originally Posted by SeqAA

PE for RNA Seq
SE for ChIP Seq. There really isn't the need for Paired reads.

the throughput really varies on the experiment. Histone modifications or TF analysis?

[ETHANol]

Are there peak callers that accept paired-end data?

[ETHANol]

A little searching answers that question:

http://www.biomedcentral.com/1471-2105/11/81
http://www.ebi.ac.uk/~swilder/SWEMBL/
http://liulab.dfci.harvard.edu/MACS/README.html

And maybe some others.

[mudshark]

second that. at least it would be helpful to have one corresponding PE run to see how uneven fragment size distribution along the genome is. in addition the fragment size knowledge might provide important information on local chromatin structure. have a look at: www.ncbi.nlm.nih.gov/pubmed?term=21131275

Originally Posted by yxibcm

Actually, I would suggest using PE for ChIP-seq too. For SE reads, existing ChIP-seq software, such as MACS, shift and extend the reads to build the whole genome profile. The shift and extend distance was a fixed value estimated from the double peak pattern or provided by command line parameters. This might be inaccurate if the wrong shift/extend distance were used, and might cause the peak appear as doublets. So the peak heights are sensitive to the shift/extend values. PE sequencing provides fragment size information, therefore build whole genome profile is straightforward, no shift or extend involved. PE is also not that expensive compared with SE.