ProteomeTools - Prosit iRT - Data (update) // missing c + splitting

The original Prosit iRT dataset missed ProteomeTools sequences containing cysteine.
We applied now the same criteria as in the original publicationData for learning was restricted to peptides identified with a minimum MaxQuant score of 70 and an iRT variance between different ProteomeTools runs of less than 1.

iRT values are the same as in the publication, the scaled values can differ as the normalization function is now applied on a different data range.
Please be aware that the original iRT model error can not be achieved as their was train/validation leakage. Therefore please don`t compare to the reported validation error of the weight file. The overall performance stays similar when using the new data.
Sry for the inconvenience.
The model trained on the original data will stay on the prosit website for fair comparison to the publication.