Application to MZ and Ion mobility

[GeorgWa]

Hi,
Really cool project!

I wondered if the requirement for monotonic increase of RT and predicted RT during graph construction precludes the application to other non-monotonic regressionn tasks like MZ and ion mobility.

Best,
Georg

[JianSong2018]

Hi, GeorgWa

Thanks for your appreciation.

For non-monotonic regression, the edges of the graph will not be subject to any monotonicity constraints, resulting in a graph with more edges. At this point, whether the longest path represents the regression curve requires further exploration. Thanks for your insightful comment.

I noticed that AlphaDIA (well done!) uses a complex multi-kernel LOESS algorithm. Perhaps it could be replaced with calib-rt.

Thanks and best,
Jian Song

[GeorgWa]

Hi,

This makes sense, if you have ideas how to extend it I would be quite interested. AlphaDIAs LOESS regression also uses the scikit-learn like syntax, so it should be compatible. I'm quite busy at the moment but feel free to open a PR!

Best,
Georg