basecall.py a reference example from the jmr twitter dataset updated to include a 4-dye variable rate model
Usage:
basic run:
python3 ./basecall.py -ie 0.09 -cf 0.065 -dr 0.02 -plots
output: cycles: 40 basecalls: CTGACTGACTAACGTCAGTCCATGCACTCACGTTTAACTG cumulative error: 0.704400 also creates a couple of plots to show original input signals per cycle, predicted signals, error, signal loss
grid-search:
python3 ./basecall.py -ie 0.09 -cf 0.065 -dr 0.02 -plots -grid
outputs base calls for each cf/ie/dr combination, tracks total error per combination, and selects the lowest error rate
4-dye run:
python3 ./basecall.py -ie 0.09 -cf 0.065 -dr 0.02 -plots -model 4dye
note that the 4 dye version is hard-coding a 'G' at 1.5 higher cf rate as an example the rates should all be relative to some baseline, and the grid search is not designed to work with this version