Certain peak features, e.g. MeanIsotopeRatioConsistency, are only calculated if the signal exceeds intensity.threshold, and are imputed otherwise. However, if all values are missing (probably because intensity.threshold is substantially too high), the current imputation method sets the feature to -inf instead, which causes problems when training models.

We've seen errors such as:

Something is wrong; all the Accuracy metric values are missing:
    Accuracy       Kappa
 Min.   : NA   Min.   : NA
 1st Qu.: NA   1st Qu.: NA
 Median : NA   Median : NA
 Mean   :NaN   Mean   :NaN
 3rd Qu.: NA   3rd Qu.: NA
 Max.   : NA   Max.   : NA
 NA's   :3     NA's   :3

We can reproduce this using the supplied example data, and largely following the vignette examples (but with a higher value of intensity.threshold, e.g. 100000000.

While clearly the best solution is to set intensity.threshold to an appropriate value, the current behaviour leads to some hard-to-diagnose errors when running TargetedMSQC in a scripted environment. I've submitted an accompanying pull request that defaults these features to a non-informative value (I picked 1.0, but my understanding is that the exact value should not matter if it doesn't vary from peak to peak) instead of -inf.