david-peede / mentee_research Goto Github PK

• Create a new directory sub directory in site_pattern_hmm called updated_summer_hmm.
• Annotate all of your old HMM code—don't need to do the figures just the actual HMM code.
• Download Juypter notebooks for desktop.

The bug that we saw this morning was a result of an input variable declaration including underscores to delineate thousandth places. For some reason, OSCAR doesn't recognize numbers with underscores.

Because I want the ease of just copying over updated versions of my files to OSCAR without having to worry about fixing it every time, I've removed these underscores on github so that instead of 50_000 and 20_000_000 we just have 50000 and 20000000.

As you remember, I calculated four performance indicators for each gamma matrix output of Baum-Welch:

• False Positive Rate
• False Negative Rate (Miss Rate):
• True Positive Rate (Sensitivity):
• True Negative Rate (Specificity):

I'd like to find and record which of the Baum-Welch iterations in a single rep perform the best, and what their indicators are. Should I prioritize one statistic (pick the version with the highest true positive rate)?. Or should I indicate the iteration that gives the best value for each metric independently?

• Make a directory kirz_tutorials
• Make a subdirectory vcf_processing
• Either copy over or download the chr1 TGP VCF file and the panel file from the shared data directory or from the online ftp
• Copy over the chr1 index file .tbi or index it using bgzip
• Filter the VCF file to only contain bi-allelic SNPs using bcftools (HINT: you don't need to download the program it is already a module on OSCAR)
• Filter the VCF file the same way but this time treat it as a text file and do it in python
• Make sure the VCF files are the same, if they're not figure out why
• Submit a pull request with all your code but exclude the VCF files