Algorithm | Help

The target identification algorithm publishes some newly developed target-centric models employing different types of molecular descriptions and machine learning algorithms; as well as a consensus strategy based on these models as a potential advancement above individual forecasts. The algorithms involved in this process are: decision tree (DT), gaussian naive bayes (GM), k-nearest neighbors (KNN), random forest (RF), and support vector machine (SVM) considering three groups of descriptors: morgan’s fingerprint (FGP), b) general molecular properties (DSC), and c) the fusion of both descriptors (FUS).

To use this tool, a compound needs to be submitted in SMILES format.

Then, the process is performed by the server with each algorithm (the applicability domain is also evaluated by each model), and the query molecule graph with an output table is shown after some time. The output target list includes the target ID and UniProt code with their individual binary predictions [0,1] and their class probability in parenthesis.