marbl
homoplastic vs unique SNPs
Also, the results we gathered from parsnp2 not only differ from the results gathered with parsnp1, but most of the SNPs also get flagged as homoplastic SNPs instead of unique ones. Maybe this issue is connected?
Kind regards,
Kadsae
Originally posted by @Kadsae in #169
@Kadsae,
It is not surprising that the results differ between the two Parsnp versions, as much has changed. Regarding the homoplastic vs unique, is this something that you believe to be a bug/error?
Based on simulated data (paper here), the SNPs from Parsnp2 were more informative than Parsnp1 when inferring phylogeny.
I would expect Parsnp2 to find most of the same SNPs as Parsnp1, in addition to others. Is it the case that Parsnp2 is "missing" many of the unique SNPs that Parsnp1 reports, or is Parsnp2 finding those as well as many new homoplastic SNPs?
-Bryce
Hi @bkille,
as mentioned in #169 parsnp v1.6.2 found significantly more SNPs than parsnp v2.1.1. It could be that the newest version of parsnp would find more SNPs - however, I could not find a workaround for the failing partition yet.
Nevertheless, when comparing the found SNPs of parsnp1 and parsnp2, many of those originally flagged as unique get flagged as homoplasies, although the number of reported SNPs becomes less.
We also experienced this kind of behavior last year when we were using a different tool called Maast. There, we also found less SNPs than with parsnp1 and many of those found were flagged as being homoplastic. Interestingly enough, with Maast we also dealt with the same problem discussed in #169.
Maybe there is a connection because of a similar approach by parsnp and maast (e.g. partitioning input to lower memory usage)?
Best, Kadsae
