On the day the lab pertained to phylogenetic inferences we arrived with an alignment of COI sequences including our own sushi DNA and 25 sequences form NCBI obtained through Geneious. The protocol instructed us to clean up all the sequences by editing and trimming the ends so all sequences presented the same number of base pairs. The results were observed and the amount of polymorphisms were determined. Also, since there was a lot of variation between three of my sequences and the rest of the sequences the three were deleted, but in the process the shark out-group was removed. This affected future steps in the protocol.

The next part of protocol was designed to help us understand choosing the best molecular evolution model. JModelTest2 was downloaded from ww.github.com. The most recently updated version was selected, “jModelTest2.1.10 Full release (tar.gz)”. Next form the downloaded file ‘JModelTest.jar’ was selected. The file required particular version of JAVA so I had to download extra software for my Mac to proceed with the methods. Once ‘JModelTest.jar’ was opened, in Geneious and we exported our alignment in Phylip format (relaxed). The .phy file was accessed through ‘JModelTest.jar’ analyzed and the likelihood scores were calculated using the program. The process took a couple minutes before we obtained results. The likelihood was developed from the comparison to 88 models that the system can access. Once the calculations were concluded we selected the best model based on two methods. The first method was the Akaike Information Criterion (AIC). Based on AIC the best fit model was TIM2+G. The second model, Bayesian Information Criterion (BIC), the best fit model was TrN+G. The AIC and BIC top choice models were not the same.

Following the evolutionary model step, we continued onto Bayesian inference. We ran MrBayes from Geneious which allowed us to run an analysis set to 10,100 posterior distributions. In this step the shark out-group could not be selected because it was deleted at the start of the protocol. The run produced graph with x-axis of the sequence and a y-axis of frequency. We also analyzed the trace table, taking into consideration the “fuzzy caterpillar” that should have been seen expressed by the data. What was obtained from this process was that our analysis was most likely too short which disrupted the posterior distribution. Next, we viewed the inferred tree model. The same process was repeated, but we changed the ‘chain length’ to 1,1000,000 and 100,000 for the burn-in.

Finally, we worked with the maximum likelihood to infer a phylogenetic tree. From Geneious RAxML was installed. We selected our evolutionary model and created a Consensus tree which was compared to the Bayesian analysis. Maximum likelihood was also calculated through PHYML. Because I deleted my shark sequence at the beginning of the process it was not included in the final tree that was developed.

Screen Shot of Likelihood analysis with PHYML: