Prediction-pipeline of this Web-Server

ATC-statistic Target-statistic
This diagram illustrates the ATC prediction pipeline (left) and target prediction pipeline (right). The ATC prediction is carried out in three steps. In the first step the input compound is compared with the ATC dataset by the following methods namely 2D, fragment and 3D similarity searching. In the second step the ATC-class and the corresponding score is calculated for each method. The last step ensembles the predicted ATC-classes according to the score and predict the final ATC-class. Similarly, the target prediction is also carried out in three main steps. In the first step the input compound is compared based on structural similarity (2D). The second step analysed the statistical significance of the similarity score in comparison with precalculated statistics of the dataset. The last step computes the raw score for each target and finally the target is predicted with consideration of the weighted Z-score and E-value threshold.