Dear experts,
I have 300 samples and I would like to conduct ddRADseq for population genomic study (RE: Sbf1 and MseI).
The genome size of the samples is roughly 27G.
I would like to use Hiseq 2500 for sequencing.
I would like to ask "how many lanes should I adapt for this exp (if 60x coverage is require)?
Dear Bing-Hong Huang
There are about 220~250M clusters per lane in HiSeq High-Throughput Mode, 150~180M clusters per lane in Rapid mode.
I thought you probably need estimation below:
- 27G * 0.001(After ddRAD digestion, I don't actually know the ratio) * 300(samples) * 60(coverage) / 200M (clusters per lane) in single end mode.
Theoretically, one cluster only sequences one DNA fragments.
So, you probably need 2430 lanes.
I think the 60x coverage is too high for ddRAD-seq if you want to get SNP data.
If there's something wrong, please correct me.
It's really hard to tell how much lane you need.
To make a good estimate, you will need an idea of the number of rad loci that you expect from the enzymes that you will be using (i.e., basically distribution of the cut sites across the genome; need genome).
Then, during the RAD, the size range of the rad loci that you want will also affect it. For example, a size range of 450 to 650 bp will probably capture more rad loci than 450 to 550 bp. Therefore, the amount of sequencing needed to obtain a certain coverage will vary.
What we do is run a pilot sequencing. Say if you can share a lane with some else, it will be nice to test your rad design with a number of samples. This will give you an idea of the number of loci that you can recover given a size selection window, and the average depth that you get given a certain amount sequencing.
But if you have a whole genome reference, you can do an in silico digestion and get an information about the rad loci that you will get.
In order to calculate the expected coverage, you need to know how many loci you expect to recover. You can either do this via simulation (somewhat reliable if using a reference genome for your species or a closely related one), or by quantifying the proportion of DNA found in your intended size selection range using samples digested and run on some form of quantitative fragment analyser (e.g. TapeStation)
If you choose to go the simulation route, I've made a pretty simple Perl script which you can use (https://github.com/tkchafin/fragmatic). Let me know if you need any help.
If you choose to quantify via the Tapestation, you can calculate the expected number of loci as (I think):
(2*(haploid genome size) * (proportion of DNA mass in size selection range) * (proportion of sequence-able fragments)) / (mean DNA fragment size in size range)
Where the 'size range' refers to your intended size selection (NOT adjusted for adapter lengths) and 'proportion of sequence-able fragments' refers to the proportion of DNA fragments which are terminated on exactly one end each by SbfI and MseI. SbfI-SbfI and MseI-MseI fragments will not sequence, as these will have the same adaptor on both ends. You can estimate the sequence-able proportion using the same in silico method above.
Note that quantifying using real digests can be very misleading if your sample already has some degredation- be sure to use samples for which your extractions are very high molecular weight, and use many replicate samples to get an idea of the variance.
Once you know the number of loci which will be targeted, you can calculate expected coverage as:
Exp. reads per lane / (number of samples * number of loci)
If this coverage is too low, you can adjust in several ways:
1. Multiplex less samples per lane
2. Narrow your size selection range
3. Choose rarer-cutting enzymes (i.e. replace MseI with a 6-base cutter)
Good luck! :)
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size