ctDNA feature derivation for predictive modeling in IMpower150 training/test and OAK validation data
The ctDNA analysis plan for the machine learning model was finalized before the development of the model. ctDNA levels were quantified using 23 different metrics measured for each time point , and the change in ctDNA relative to baseline was quantified using 55 different metrics for each on-treatment time point .
. Individual features were scaled by the IQR of that feature before running the machine learning model., along with the rank concordance of each metric with landmark OS and PFS for each visit.
genentech That's fantastic news and a great application for technology.