I am curious as to what range of pIC50 values are generated by the assay system in use. Are predictions that are either below and above that range are scored in a special way (ie clipping to assay range)?
Created by Gregory Koytiger gregkoytiger Please note that the test data are pKd values not pIC50.
The assay used for generating the test data (for both Round 1 and Round 2) is the DiscoverX KinomeScan. You can find more information here: https://www.discoverx.com/technologies-platforms/competitive-binding-technology/kinomescan-technology-platform
It doesn't help - because when you are running the data using Kinomescan, depending on the concentrations tested they will report certain IC50 values as being greater than a thresholded minimum (eg. >10uM). If those are then being evaluated as being exactly 10uM and a team's prediction is 30uM, it would be unfairly penalized. Especially for Round 2, the more explicit the challenge organizers can be about how the data is being generated (number of concentrations, doses) and what is the dynamic range of the reported IC50 values, the better it would be for the participants One can see the typical range of pIC50 or pKd values by extracting the corresponding assay values from DTC or ChEMBL.
I hope this helps. With regards to the assay system, that's a good question and one that I don't have an immediate answer for. I'll see if I can find anything out but perhaps @anna.cichonska or @aittokal know?
For scoring, there is no clipping. As long as the predicted pKd values are numeric and greater than 0, the predictions will be scored as-is. For non-numeric or values <0, the prediction file will fail to validate and you'll get an error instead of a score.