PEPs are constant one for certain inputs

[ezander]

For certain inputs of scores and targets the peps computed by getQvaluesFromScores are all one. This is due to the monotonization step at the end of the function (line 81, probs = monotonize(probs)).

I've created a fork at https://github.com/ezander/triqler and added two files (data.py and issue.py) in https://github.com/ezander/triqler/tree/issue/pep_constant_one/tests/issue that demonstrate the issue.

[RalfG]

We have also seen this happening when the highest scoring PSM is a decoy. As a workaround (but obviously not a fix), such decoy PSMs could be removed from the dataset.