How to use Deep-PK

Deep-PK is a robust machine learning method based on diverse molecular descriptors and graph neural networks. Deep-PK contains 73 pharmacokinetic and toxicity properties and provides other userful information such as toxicophores, toxicity, drug-likeness and substructure analysis.
A. Users can submit a job through the prediction page on "Prediction" or locally through "API".
B. Users can also access this help page on "Help".
C. All data used in Deep-PK experiments are downloadable by clicking on "Data" at the top menu. All of them are properly explained on "Theory"

A. Users can either submit an SDF file (1), a SMILES file (2) or an individual SMILES string (3). For the SMILES string, users can also draw their molecule using JSME molecular editor and retrieve its SMILES by clickling on "Get SMILES and Fill It In".
B. Optionally, users can receive a notice the job status and linke by providing an email address at the submission form.
C. Lastly, users are required to choose a run mode among Absorption, Distribution, Metabolism, Excretion, Toxicity, ADMET, and Example.

A. Based on the run mode, users can select the class of pharmacokinetic properties to filter the results in a table.
B. A Focus page shows more details about the molecule and its predictions.
C. Users can further conduct "substructure analysis" and "pharmacokinetic optimisation" through their respective pages.
D. The result table can be downloaded as CSV format.

A. On the Optimisation page, users can compare their query molecule(s) with the ones that were optimised by Deep-PK in terms of the ADMET predicted properties..

A. Users can download Deep-PK datasets used for training/validation/and blind-test on "Data" page.