I work on an ancient protein family that exists in both eukaryotes and prokaryotes and my interest is in elucidating its deeper branches. When I pull down the data from NCBI I get about 6000 sequences. If I use an algorithm to help me cluster similar proteins (e.g. 80% similar) it gets reduced, and if I do my clustering with 50% similarity I get about 550.
The question is, in your experience, which one would you choose:
- use the smaller dataset (or even smaller) and do the most sophisticated, yet computationally intensive, analyses you can hoping that if there is a signal, these analyses would pick it up and thus the deeper branches get better support.
- use a lot of data and hope that then the intermediate evolutionary steps could be inferred more easily and the deeper branches get more support?
I would choose one or two samples per family or genus. 500 tips is feasible though.
Yes, as Chris suggested, multiple similar sequences from the same genus will not really help your analysis of deep divergences and in fact, may skew your topology. A tree with fewer branches is also easier to present for publication.
The 2nd option is probably more likely to give a quality answer. If you have the time/computing power I think you should look at the larger dataset, too. If your clustering combines particularly disparate sequences you may be thinning out your taxa distribution and inviting more opportunity for long branch attraction. I would consider comparing the topologies and look for places where lots of relative diversity was collapsed in one area but not in other areas of the tree. If it is even throughout, the smaller dataset is probably much more likely to give a fair answer.
Matthews' answer has an important point that I want to highlight, pointing towards an intermediate approach. I might try first with the larger data set. If you find lots of near-duplicates in some clusters, thinning there taxon sampling may improve possibilities for more sophisticated methods without loosing phylogenetic signal for greater patterns. But, definitely, all more deviating taxa should be kept intact, otherwise the risk for severe bias due to long-branch attraction (or rather short branch explusion) is almost inevitable.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation