I have a dataset from a ddRAD library (Illumina HiSeq4K, 150PE). I'm playing around with STACKS, and noticed that it has some limitations when dealing with paired-end data from ddRAD datasets.
Namely, it treats them as separate/independent loci.
This seems like a problem from a population genomic point of view, since there is a base assumption (admittedly often violated) of independence between loci. If we have paired data (single-reads and paired-end ("PE") reads), we know that those two loci aren't at all likely to be independent.
What I have done so far is demultiplex individuals' data based on inline barcode sequence(s) and give a rough quality filter (sliding-window 15%, min quality score = 10). So I'm left with four files for each individual:
One file of SE reads and one file of PE reads, in-phase (where both reads from the same fragment/cluster were kept and not discarded).
One file of SE reads, whose PE counterpart has been discarded.
One file of PE reads, whose SE counterpart has been discarded.
Only one or the other read - SE or PE - needs to have a SNP. The other read can be discarded. Can STACKS keep track of this? Can it go through assembly and SNP-calling and keep track of header-titles, and use that information to figure out what those sequence pairs are?
Or is that information lost? (I suspect that for the remainders - the latter two files described - it would be quite difficult to recover that information...)
Alternatively, is it worth throwing caution to the wind and using the SE and PE reads and throwing all of those data together at the end? Is there another approach here?
Many thanks,
Sean
Namely, it treats them as separate/independent loci.
This seems like a problem from a population genomic point of view, since there is a base assumption (admittedly often violated) of independence between loci. If we have paired data (single-reads and paired-end ("PE") reads), we know that those two loci aren't at all likely to be independent.
What I have done so far is demultiplex individuals' data based on inline barcode sequence(s) and give a rough quality filter (sliding-window 15%, min quality score = 10). So I'm left with four files for each individual:
One file of SE reads and one file of PE reads, in-phase (where both reads from the same fragment/cluster were kept and not discarded).
One file of SE reads, whose PE counterpart has been discarded.
One file of PE reads, whose SE counterpart has been discarded.
Only one or the other read - SE or PE - needs to have a SNP. The other read can be discarded. Can STACKS keep track of this? Can it go through assembly and SNP-calling and keep track of header-titles, and use that information to figure out what those sequence pairs are?
Or is that information lost? (I suspect that for the remainders - the latter two files described - it would be quite difficult to recover that information...)
Alternatively, is it worth throwing caution to the wind and using the SE and PE reads and throwing all of those data together at the end? Is there another approach here?
Many thanks,
Sean
