Molecular Data
Input data, predicted ADME properties and calculated molecular descriptors are conveniently displayed in a user-friendly Excel-like spreadsheet for inspection, one-click sorting, and edition. Molecular structures are depicted in the first column.

pK_a Prediction
Because the pK_a model predictions carry much more information than the regular ADMET models, its results are displayed in a separate window. Apparent pK_a, complete microspeciation including the fully deprotonated structure, and microstate percentages are displayed for each macroscopic protonation state. In addition, proton dissociation probabilities and microconstants for each microstate can be presented on demand.

No less important is the dependence of individual protonation micro- and macrostates on pH. This, plus the average number of protons and formal charge, are shown in the graphing window featuring the mouse-sensitive marker for instant numerical output.

Interpretation of Predictive Model Results
Nowadays, providing mere numerical estimates of ADMET properties is not enough. Much more valuable is answering the question of why a given molecule has a particular value of property X, what structural aspects influence X the most, and giving valuable synthetic hints on how to modify the molecule to improve its ADMET properties. The Descriptor Sensitivity Analysis of ADMET models tool is able to answer these questions through its intuitive interface to descriptor dependence gradients...

...and individual descriptor dependence graphs.

Descriptor Sensitivity Analysis is no longer limited to regression models. We have extended this approach to binary classification models based on Artificial Neural Networks. Descriptor sensitivity is defined here as the reciprocal of a minimal descriptor change to flip the prediction.

Detailed Structure Visualization
ADMET Predictor's Structure Visualization tool is a real treat for chemists who like analyzing their favorite molecules atom by atom. Many important atomic properties can be mapped onto structure depiction.

Input / Output
ADMET Predictor can read input files containing either 2D, or 3D molecular records, interactively or in batch, in one of the following formats:

• SMILES strings
• SDF (ISIS/Base™)
• RDF (ISIS/Base)
• MOL (ISIS/Draw™, ChemDraw™, etc.) or MAC (SYBYL™)

The program preserves a list of most recently open files

ADMET Predictor results can be saved as:

• GastroPlus™ import tables
• Tab-delimited text tables (for MS Excel™ import)
• A copy of the original input file with inserted ADMET properties
• A 2D SDF- or RDF-formatted file with inserted ADMET properties (for, e.g., an easy import into ISIS/Base)

Run Options
ADMET Predictor runs can be customized on-the-fly with an easy to use set of run options.

Property/Descriptor Histograms
The distribution of calculated values is conveniently displayed on the "Property/Descriptor Histograms" tab for properties with either continuous, ...

... or discrete spectrum:

Property/Descriptor Correlations
The "Property/Descriptor Correlations" tab displays correlation plots between any two numerical columns read or calculated by ADMET Predictor. Click of a button automatically calculates and displays statistics of a linear fit between the two variables.

 

A special Plot pKa button does an automatic pairing of observed vs. predicted pK_a (multiple values per molecule) allowing for easy and instantaneous evaluation of the pK_a model predictivity.

Because of its special nature, the display of ADMET Risk correlations has an added extra feature of color coding individual data points by ADMET Code representing satisfied ADMET Risk rules.

ADMET Risk Editor
ADMET Risk rules (an analog of Lipinski’s Rule of Five) are 100% customizable - existing rules can be edited or removed, new rules can be added, and program default rules may be specified.

4D Data Mining
The distribution of input molecules in the chemical space can be visualized with the aid of the 4D Data Mining module. Any one of the X, Y, Z variables can be assigned to either a molecular descriptor column, or a predicted property, or a local principal component pertaining to a given chemical dimension. The fourth variable, C, can be used to color data points.

Variables selected in the 4D Data Mining panel are displayed in a rotatable 3D chart. A normalized trend vector shows the magnitude and direction of the fastest increase of the coloring variable C.

Alternatively, the same 3D chart can be displayed with presentation quality in Miner3D™ graphics component built into ADMET Predictor.

Model Editor
Users can add up to 120 in-house predictive models to the program by simply appending records to the model table in Model Editor window.

On Line Help
The user manual, as well as on-line help files, contains a complete tutorial reducing the learning curve to a bare minimum and making expert users out of beginners in a very short time.