plot vafs from primary and recurrence against each other
Filter out mutations that are non-diploid (use CNV filter) (edited)
Implement at 25X threshold for local read depth (ref_count + alt_count), which should have a 10% error in VAF per the calculation on page 2 of Javad’s paper
Javad said that it would NOT be ideal to use the VAFs of all tumors in a group UNLESS you were to assume they all had the same mutation rate (they do not)
See the R package https://github.com/marcjwilliams1/neutralitytestr, following discussions with Javad: