highly significantP-value <0.001
When compared to the results of all types of randomization, models trained to the real data and PKN1 were highly significant (P-value <0.001, Table S1 ), indicating that the family of trained cFL models fit the data better than expected by random chance.