Entering edit mode
2.4 years ago
gubrins
▴
290
Heys,
I'm working with a genomic dataset and I am producing a phylogeny out of several sliding windows. I have two questions about this:
First: I compared the speed between raxml and iqtree, and the second is incredibly faster than the first one. Do you know how is that? Is iqtree more implemented for genomic data?
Second: I am obteining as many ML trees as sliding windows I have. How can I group and visualize all these trees in one? I was thinking in densitree, but it only accepts one file, not several. What do you suggest?
Thanks in advance!
Can I ask you what is the intention or the question behind using sliding window analysis in phylogenomics? Normally, one uses homologous sequences for making phylogenies, how would the sequences from the windows even be considered homologues?
Heys Michael, maybe I did not express myself clear enough. I made sliding windows of my samples based on a reference genome, so I am doing the phylogenies for each one of those sliding windows, which are homologous because it is the same genomic region for all the samples.
Hi, ok I think I get now what you are trying to do, but I don't quite understand what the advantage is over simply using all genes. How closely related are these sequences? To do the sliding windows analysis with a reference, I guess they must be quite similar. So then, there are only few variable sites, I guess? The invariant sites then would have no phylogenetic information and could be removed. Also, what do you do if some genomes had an insertion or rearrangement? The sequences would no longer be properly aligned. In consequence, I would stick with the classical approach and do a MSA of coding regions and then compute a phylogenetic tree on that.
The sequences are from the same species, with an outgroup of another species. I am interested in getting a genomic phylogeny, if I just do the phylogeny with genes I am rejecting out a lot of information and could be that the non-coding regions are showing me a pattern not shown by the coding ones. According to my knowledge, the invariant sites should be kept, because if not when raxml tries to estimate the number of invariant sites, it does not found any and I'm not sure how this could affect the model, what's your opinion on that? Last, I am not sure how the software is handling isnertion and rearrangements, that's a good point that I also asked myself and still did not find an answer.
Thank you very much, I will double check all the advices you gave me. However, once I already run some trials, do you know how could I visualize different runs of raxml at the same time, as densitree with beast? Thanks!
I think there is a function in RaXML to generate a consensus tree.