mate pair sequencing

[Chien-Yuan Chen]

Does Anybody has the experience of mate pair sequence? Except the insert size is different from pair end sequence, and what is the benefit with this method?

[seqgirl123]

Never tried mate pair myself. Only used PE so far.


Straight from Illumina: http://investor.illumina.com/phoenix...574&highlight=

Check this website that talks of mate pair (MP) libraries: http://seq.molbiol.ru/#i_c_fr

This older thread might also help clarify some things: http://seqanswers.com/forums/showthread.php?t=503

Hope that helps.

[Torst]

Illumina's regular "paired end" kits can do an insert size around 500 bp.

Their first "mate pair" kits use a different technique, and can do much longer insert sizes. I understand 4 kbp kits are available, and 10 kbp kits will be available soon.

The BENEFIT of longer insert size comes when trying to assemble sequences. It is IMPOSSIBLE to resolve a repeated sequence of length N unless you have paired end reads with an insert size of at least size N. As many common repeats in genomes are usually much larger than 500 bp, the "mate pair" kits are crucial to resolving these repeats and finishing/closing genomes (or at least drastically reducing the number of contigs).

[bair]

Is it correct that: PE is used for SNP/indel detection and ME for structural variation. Both can be used for de novo assembly?

[Torst]

Is it correct that: PE is used for SNP/indel detection and ME for structural variation. Both can be used for de novo assembly?

Yes it is true, but not quite the full truth.

ANY reads can be used for SNP/indel coverage - that includes short reads, long reads, paired-end reads, single-end reads, mate-pair reads - it doesn't matter, as long as there is sufficient coverage and/or quality to infer a statistically signficant result.

For structural variation, yes of course you need to be able to examine large scale sequence relationships, and PE or SE reads will not do this beyond ~500 bp say. Large insert libraries such as MP and 454FLX are needed here. Or other strategies such as Opgen's "optical mapping" maps.

De novo assembly can use any reads you give it, including SE, PE, and MP. The ability of different software to handle mixes of these data types varies however.

[bair]

Thank you Torst,

In order to get "robust" SNPs/indel, it might be better to use PE / ME to confirm SNPs by two directions, but I did not see SNPs detection programs doing things like this way. Maybe I'm wrong.

[kmcarr]

Is it correct that: PE is used for SNP/indel detection and ME for structural variation. Both can be used for de novo assembly?

Mate pair reads are essential (or at least extremely helpful) in de novo projects for orienting contigs into scaffolds. However you shouldn't use mate pair libraries as your source for general coverage. Illumina recommends reads no longer than 36nt when sequencing mate pair libraries; this is to reduce the probability of a read crossing the junction point where the two distant ends of the original DNA fragment were joined during circularization. To generate the bulk of your coverage it is best to use 2x100 (or 2x120) paired end reads. Secondly, the method used to generate mate paired libraries requires a lot of DNA and ends up "throwing away" most of it. This has implications for library diversity.

[Torst]

In order to get "robust" SNPs/indel, it might be better to use PE / ME to confirm SNPs by two directions, but I did not see SNPs detection programs doing things like this way. Maybe I'm wrong.

You are still thinking in old "Sanger sequencing" terms. High throughput sequencing produces coverage depths of 50 to 500 fold! About half the reads will be "forward" and half "reverse". The statistical significance of SNPs at this coverage is high.

[Torst]

Mate pair reads are essential (or at least extremely helpful) in de novo projects for orienting contigs into scaffolds. However you shouldn't use mate pair libraries as your source for general coverage. Illumina recommends reads no longer than 36nt when sequencing mate pair libraries; this is to reduce the probability of a read crossing the junction point where the two distant ends of the original DNA fragment were joined during circularization. To generate the bulk of your coverage it is best to use 2x100 (or 2x120) paired end reads. Secondly, the method used to generate mate paired libraries requires a lot of DNA and ends up "throwing away" most of it. This has implications for library diversity.

I agree with this mostly. The MP protocol has issue with (1) chimeras, (2) pollution with fragments not including the junction, which end up like PE reads, and (3) the 10x amount of DNA required. If you can do a MP and a PE run as kmcarr describes, you get the benefit of large scale scaffold information from the MP, and the finer depth and detail of the PE.