BEAST 1.10 for ancestral sequence reconstruction?
901 views
Skip to first unread message
bioinform...@gmail.com
Apr 4, 2019, 1:49:44 AM4/4/19
to beast-users
Hi All!
I'm wondering if BEAST 1.10 is an appropriate choice for Bayesian ancestral state (sequence) reconstruction? From what I've been able to gather, it doesn't look like it, as it seems the ancestral state reconstruction was designed for phylogeographic analyses and not entire sequences...but I wanted to ask if it's possible and if so, how?
I'm interested in inferring ancestral sequences at specific nodes to be then able to determine clade-specific SNPs which I can then use to look for independent occurrences of common mutations. Would BEAST 1.10 be an appropriate choice for this? Or should I just use BEAST 1.10 to get my phylogeny and then use that for reconstruction in another Bayesian (or ML) program?
Any responses are greatly appreciated!
Thanks!
Red
Red
Andrew Rambaut
Apr 4, 2019, 2:01:35 AM4/4/19
to beast...@googlegroups.com
Hi Red,
The simple answer is yes - it will do ancestral state reconstruction for just about any data including nucleotides and amino acids. If you create a taxon set you can then specify the TMRCA of that taxon set as the node you wish to reconstruct for (it will create a log file where it will write those reconstructions). At the moment in BEAUti you can only specify one node but by modifying the XML, you could log other nodes.
Alternatively, you can reconstruct for all nodes but this will write the reconstructions into the trees file as annotations. For sequence data this will make the file size huge and you will need to come up with some way of parsing it out and processing it.
Hope this helps,
Andrew
--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To post to this group, send email to beast...@googlegroups.com.
Visit this group at https://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.
Brian Foley
Apr 4, 2019, 8:51:55 PM4/4/19
to beast-users
You should take a look at the NextStrain software and tutorials.
I found this discussion of RNA virus evolution:
Which shows that NextStrain can be used to track the evolution of codons in a tree.
You simply go to the inlfluenza tree in NextStrain, and there is a “Diversity” plot at the bottom of the page:
If you mouse over the bars for each amino acid (tall bar = high
diversity) to find amino acid 135 and click on it. It maps the various
amino acids at that site onto the tree:
It's a bit complex to install the auspice software that NextStrain runs on, but if you have a good data set it may be worth the effort.
Andrew Rambaut
Apr 5, 2019, 4:44:21 AM4/5/19
to beast...@googlegroups.com
Hi,
NextStrain uses IQTree to do its tree building and ancestral state reconstruction. It is probably much easier just to install that and use it directly.
The advantage of using BEAST is that you can sample a distribution of plausible reconstructions rather than a single ML point estimate so it allows you to construct statistical inferences.
If you are interested in reconstructing synonymous and non-synonymous mutations on the tree you can use the codon models but there is also a fast approximate method that uses the nucleotide models but imposes a codon structure on them.
Paper is here:
In BEAUti, you turn on codon partitioning (in Sites tab, "partition into codon positions = 1,2,3") then in States tab, turn on "Reconstruct synonymous/non-synonymous change counts"). You can reconstruct the states at the nodes as well, the advantage that these will be maintained as sense codons. If you want to reconstruct everything then you can do "Reconstruct complete change history on tree" - but a warning this produces very large trees as every change is written into them as annotations. I recommend not logging these too often.
Andrew
Reply all
Reply to author
Forward
0 new messages