ADMETlab 2.0 provides a convenient and easy-to-use interface for users. Two services, Evaluation and Screening, are designed to support single-molecule and batch evaluation, whose input parameters and output information will be elaborated respectively.
In the Evaluation pattern, two molecular submission approaches are provided by pasting the SMILES string or drawing the chemical structure with the help of JMSE molecule editor (17 (link)). Once a user submits the job, the webserver will automatically standardize the input SMILES strings and compute all the endpoints. The prediction results are mainly displayed in the tabular format in the browser, with the 2D molecular structure and a radar plot summarizing the physicochemical quality of the compound. For those endpoints predicted by the regression models, such as Caco-2 permeability, plasma protein binding, etc., concrete predictive values are provided. For the endpoints predicted by the classification models, such as Pgp-inhibitor, hERG Blocker, etc., the prediction probability values are transformed into six symbols: 0-0.1(−−−), 0.1-0.3(−−), 0.3-0.5(−), 0.5-0.7(+), 0.7-0.9(++), and 0.9-1.0(+++). Usually, the token ‘+++’ or ‘++’ represents the molecule is more likely to be toxic or defective, while ‘−−−’ or ‘−−’ represents nontoxic or appropriate. Here, we do not recommend trusting predictions symbolled by ‘+’ or ‘−' (probably values in 0.3-0.7), and corresponding molecules require further assessment. The substructural rules available in the webserver, such as PAINS, SureChEMBL Rule, etc., were implemented using the SMARTS recognition capability of RDKit function. And the calculation of physicochemical and medicinal chemistry endpoints was based on the python library Scopy (18 ), following the parameters reported in corresponding original papers strictly. If the number of alerts is not zero, users can click the DETAIL button to check the undesirable substructures in the molecule. Finally, the full result file can be downloaded from the website in CSV or PDF format.
In the Screening pattern, two molecular submission approaches are provided by entering a list of SMILES strings or uploading a SDF or TXT formatted file. It should be noted that the file should only contain molecules without column headers and molecular indexes, otherwise the server may declare invalid input type. After all the predictions are completed, the results for each input molecule will be presented on a separate row, containing the assigned index, SMILES string, 2D molecular structure, and a View button. The prediction details can be accessed by clicking the View button of the corresponding molecule that links to the single-molecule evaluation page. These results can also be downloaded as a CSV-formatted file to the user's computer, where concrete probably values of classification endpoints are provided to enable the users to define their own thresholds to filter out deficient compounds with different levels of reliability. A typical ADMETlab 2.0 task for 1000 molecules requires ∼84 s, but it may also depend on the complexity of molecules.
In the Evaluation pattern, two molecular submission approaches are provided by pasting the SMILES string or drawing the chemical structure with the help of JMSE molecule editor (17 (link)). Once a user submits the job, the webserver will automatically standardize the input SMILES strings and compute all the endpoints. The prediction results are mainly displayed in the tabular format in the browser, with the 2D molecular structure and a radar plot summarizing the physicochemical quality of the compound. For those endpoints predicted by the regression models, such as Caco-2 permeability, plasma protein binding, etc., concrete predictive values are provided. For the endpoints predicted by the classification models, such as Pgp-inhibitor, hERG Blocker, etc., the prediction probability values are transformed into six symbols: 0-0.1(−−−), 0.1-0.3(−−), 0.3-0.5(−), 0.5-0.7(+), 0.7-0.9(++), and 0.9-1.0(+++). Usually, the token ‘+++’ or ‘++’ represents the molecule is more likely to be toxic or defective, while ‘−−−’ or ‘−−’ represents nontoxic or appropriate. Here, we do not recommend trusting predictions symbolled by ‘+’ or ‘−' (probably values in 0.3-0.7), and corresponding molecules require further assessment. The substructural rules available in the webserver, such as PAINS, SureChEMBL Rule, etc., were implemented using the SMARTS recognition capability of RDKit function. And the calculation of physicochemical and medicinal chemistry endpoints was based on the python library Scopy (18 ), following the parameters reported in corresponding original papers strictly. If the number of alerts is not zero, users can click the DETAIL button to check the undesirable substructures in the molecule. Finally, the full result file can be downloaded from the website in CSV or PDF format.
In the Screening pattern, two molecular submission approaches are provided by entering a list of SMILES strings or uploading a SDF or TXT formatted file. It should be noted that the file should only contain molecules without column headers and molecular indexes, otherwise the server may declare invalid input type. After all the predictions are completed, the results for each input molecule will be presented on a separate row, containing the assigned index, SMILES string, 2D molecular structure, and a View button. The prediction details can be accessed by clicking the View button of the corresponding molecule that links to the single-molecule evaluation page. These results can also be downloaded as a CSV-formatted file to the user's computer, where concrete probably values of classification endpoints are provided to enable the users to define their own thresholds to filter out deficient compounds with different levels of reliability. A typical ADMETlab 2.0 task for 1000 molecules requires ∼84 s, but it may also depend on the complexity of molecules.