This technique selected features randomly by choosing 10 samples from each dataset (sets A and E).
To test for potential conflicts among the four gene fragments, maximum likelihood analyses and Bayesian Inference were performed on each individual dataset set with the same setting as in the concatenated analysis (see below).
A set of 30 compounds were selected with inhibition constant as training set from 129 compounds of dataset set and subsequently the pharmacophore model was developed.
The dataset set for that survey was obtained from BPS-Statistics Indonesia.
Finally, the testing set was used to assess the performance of the developed model trained on the merged dataset (set 1 + set 2).
The non-cartilage arrays were a subset of the training dataset set aside for this validation only.
The training percentage of each dataset is set as 80%, a validation set occupies 10% of the total images, and the remainder is the test set.
Within each dataset we set any gene expression observation that is above the specified tolerance to be "1" or overexpressed, otherwise we set its value to "0" or underexpressed.
The results were collected for the 1000×100 generated dataset setting.
For the reusability assay, the first measurement in each dataset was set to 100%.
