hezhaobin / a014-sensu-stricto-divergence

estimate the divergence time between sensu stricto species

title	author	date
estimate coding sequence divergence between _sensu stricto_ species of yeast	Bin He	27 nov 2017

Goal

Estimate the level of synonymous and nonsynonymous divergence between sensu stricto species of yeasts. This was motivated by a discussion with Peter Andolfatto. I also wish to know this, namely how distantly related are the yeast species relative to each other.

Approach

1. The paper that described the genomic sequences of the sensu stricto group is Scannell et al 2011. In Figure 3 of that paper, they gave the "relaxed molecular clock estimates of divergence" based on a subset of 106 genes. However, I don't understand what the "substitution per site" stands for and thus unable to directly quote their estimate.

2. Alternatively, I decide to download the multiple alignment from the website and use PAML to estimate the level of divergence.

Notes

[2 dec 2017] troubleshoot paml in biopython

relative vs absolute file path

Problem

The Biopython PAML wrapper function cannot find the relevant alignment, tree, control or output files.

Solution

1. I found the Biopython wiki for PAML misleading. In particular, the "working_dir" option is designed to shorten the path of the files as input to paml, e.g. alignment, tree and control. After debugging the code (biopython's), I found that the wrapper function "codeml.py" actually expects absolute paths for those files.

2. The file format conversion that I got working yesterday produces the "phylip-intercleaved" format. This should work with PAML except that PAML expects a letter "I" on the first line for this format, which the Biopython function left out. I changed the output format to "phylip-sequential" and it worked.

Notes

In general, I've found working with Python and Biopython rather enjoyable, due to the ease of reading python code and the fact that the functions for biopython are quite transparent, which means I can debug them without much difficulty.

stop codons

• codeml will throw a warning when it detects a stop codon in the sequence. Even though this version (4.9e) doesn't abort the run, it does require a user input, which means I have to babysit the run.
  • I haven't figured out how to properly identify and remove columns in the alignment that contains at least one stop codon. What I did is to remove the last 18 columns (6 aa.) from each alignment to alleviate the problem. This means I still need to babysit the run.

unequal sequence length in some alignment (e.g. YLR038C)

• not clear why this would happen
• to not let it break the program, I put the AlignIO.read() function under a try operator, and use except ValueError: to capture the error. The function then returns with a "None" value, which I use in a test to decide if the result should be parsed or not.

alignment effectively zero after removing gapped columns (e.g. YFR034W-A)

• in this case, PAML will successfully complete, but the parsed result doesn't contain values, resulting in errors in the python code.
• I again used a try statement to try and catch the error
• After doing all the above, I could get the code to run for ~1000 genes at a time, before it would quit on some errors I didn't prepare for, and need to restart from where it was left off. I changed the name of the output and log files each time I restarted the program.
• finished ~3300 genes. didn't try to get everything.
• result in the analysis folder (see Rmd and html files)

[1 dec 2017] apply to many alignment

• Followed biopython tutorial to learn how to use python to read and convert alignment

  • it turned out that the AlignIO module can easily read in an alignment file and write it in PHYLIP format. that's exactly what I need.
  • I can also use split function to modify the sequence name. perfect.
  • also remember to slice the last three columns (stop codon) from the alignment

• Got the code working for

  1. get the file names for all *.codon.mfa in the data directory
  2. read in each file, slice out the last three nucleotides (stop codon) and edit the sequence name to retain only the species names
  3. tested PAML module in biopython, works great

• Parse the alignment

  1. process sequence names to include only 4 letter species abbreviation
  2. convert to PHYLIP intercleaved format
  3. rename to make it simpler to call in batch mode

[29 nov 2017] test PAML

• learned about the concepts used in the codon model, such as dS and S*dS, etc.
• tested PAML on one alignment (YBR093C)
  • need to format the alignment file by rewriting the species name
  • used ALIGN to convert the fasta format to PHYLIP (intercleaved)
  • wrote a tree file
  • edited a codonml.ctl
  • stored everything under analysis/test_paml folder
• based on this one gene, Scer and Spar are approx ~50% synonymous divergence.
• the next question is how to apply this to a large number of genes and extract the results?

[27 nov 2017] start the analysis

• Downloaded the multiple alignment
• Found Dujon 2006 to contain some useful information
• Talked to Patrick and decided to go with PAML. Will test it tomorrow.