todo.txt

1. Theory/simulations of robustness to ILS
   
   should probably run these without split populations?
   make sure parameters are giving expected sequence divergence
   - figure out what this percentage is from 100 genomes sequences/par reference:
     ~99.5% id w/ cer ref
     ~89% id w/ par ref
     (these numbers are calculated from only match/mismatch columns in alignments
     (gaps are ignored) so it's probably a bit of an overestimate if anything...
     unless shittiness of alignments outweighs that)

   goals of simulations: 
   establish tpr/fpr for reasonable parameters with HMM method
   pick reasonable parameters for HMM - average across different demographic models?
   show ILS is unlikely to be an issue (related to low fpr)

2. HMM algorithm -> show advantages over previous approaches; evaulate power/FPR; boundaries

3. Apply to data (maybe some new data?)
   - 100 genomes + SGRP cer
   - 37 par SGRP

4. Empirical patterns
   - how much introgression overall and per strain
   - relationship between introgression levels among strains and phylogeny/geographic
   - frequency dist
   - size differences and inferences of timing
   - functional variation/GO
   - if par analyzed, similarities or diff compared to cer
   - UNKNOWN

5. Experimental/function
   - GWAS




current goals for pipeline:
- alignments for beer strains, sgrp paradoxus strains, chinese strains (?)
- simulations for something that looks more like bayanus (for unknown category)
- introgression into paradoxus as well
- extract directories into separate file
- master run.sh that goes through all the steps without problems
- automatically read in hmm parameters, based on medians from simulations

for grant:
1. show ILS not likely to be a problem
2. description of HMM and validation of performance
3. application to real data
   - strain level summaries of introgressed sequence
   - gene level summaries
   - distribution of introgressed sequence size
   - bed file of coordinates for one strain
4. inferring timing/number of introgression events based on length distribution


figure out why in simulations we seem to be predicting introgressed sequence in cerevisiae reference