0

October 8 – Phylogenetic Inference

This week is the last part of the Sushi Test barcoding project. The first step was to clean up our alignments, which I did by editing the beginnings and the ends of the sequences to make them all line up. After this was done I looked through my alignment to spot the polymorphisms. Out of the first 20 column, 9 were polymorphic.

Next, we chose a model of molecular evolutionary using a program called jModelTest2. To use this we had to go to Geneious and export our alignment in relaxed Phylip format which ensures that if sequence names are truncated and identical then the full length names would be shown followed by a single space. After this I went back to jModelTest2 and opened the exported alignment. I went to ‘Analysis > Compute likelihood scores’ and clicked ‘Compute likelihood’. this calculated likelihoods for 88 models of molecular evolution.

** It didn’t work for my computer so Prof. Paul told me to use Mikayla’s

AIC chose GTR + I + G

BIC chose GTR + I + G

After this we started with our Bayesian Inference to create our phylogenetic trees. We ran it with a substitution model- GTR (because its what jModelTest2 chose for us), and chose the outgroup to be the shark species that we incorporated in our data. Since the best model showed the +I+G, I chose invgamma for my Rate Variation. Since this was just for lab to learn about the program we only ran it for a short time

  • Chain length: 1,100,000
  • Subsampling frequency: 200
  • Heated chains: 4
  • Burn- in length: 100,000
  • Heated Chain Temp: 0.2

Once it ran, I opened a parameter estimates tab. It showed a pretty empty graph which was expected since we ran the analysis for such a short time. Then I clicked on the Trace tab. This was also a bad graph, showing an upward curving line (again, expected). Lastly, I looked at the tree. I saw that it was probably missing clades. Once I ran the analysis again with a longer chain length and burn in some clades were recovered and there were new support values.

The next thing to do was look at maximum likelihood. To infer a phylogenetic tree using maximum likelihood, we installed a Plug In called RaxML which does a very fast maximum likelihood inference. We did a ‘Rapid bootstrap with rapid hill climbing’ run and made ‘RAxML bootstrapping tree’ then a ‘consensus tree using those.

Then, we used a different program using maximum likelihood called ‘PHYML’. I used HKY85 model and made my best tree with bootstrap support.

 

AT HOME –> I ran this again with the HKY85 model but ran it with 3 million generations with a subsample frequency of 500. Burn in was 300,000 and outgroup was the shark species.

 

 

 

orieney

Leave a Reply

Your email address will not be published. Required fields are marked *