I recently ran a Bayesian inference analysis using MrBayes on the CIPRES Science Gateway, but the analysis ran out of time (168 hours). Currently, the analysis is set to 4 runs, 6 chains, 0.7 temperature, 200 million generations on 604 sequences. This is only for the COI region. Is there a way to decrease the time taken for the analysis, for instance, by adding more runs/chains? On the side, I did run the analysis at 50 million generations before, but the split frequency did not stabilize at or close to 0.1. The analysis also has several duplicate entries for some species (roughly 30+ sequences per species) because i was not sure if only 1-2 sequences per species would be enough.
200 mln generation on 604 sequences are not a good option for MrBayes. Maybe try to do some sampling on representative sequences or use other phylogenomic programs (e.g. fasttree might be your friend, or PhyML).
As you mention duplications (and maybe very similar seqs) - they are bad for your tree, as MrBayes 'can't confidently decide' the very last and detailed nodes - how they should be split, therefore maybe run sth like CD-HIT / CD-HIT-EST with suitable identity threshold (e.g. cd-hit-est -c 1.0 input output) and only then use your sequences with the MrBayes.
I know the question is pretty old now, but hope the answer might be of use for others wondering.
- Badges
- Science topic
- Similar topics
- Correction