Parameters File (.prm)

The PRM file records a summary of all parameters necessary to reproduce the results of a modeling project, including a list of all of the descriptors were used as input. The .prm file is one of the files read in when you open a saved ADMET Modeler project, so that you can repeat the settings of the previous run.

Although many parameters are listed, the format of this file is straight-forward. Lines beginning with “#” are comments. Most parameters are listed as its name followed by its value, delimited by a TAB. Each parameter occupies one line.

A sample .prm file is shown below. It has been modified for clarity; indented ellipses (...) indicate places where multiple lines have been omitted.

# ADMET Modeler(TM) input parameter file. Version 1.4.3

# All keywords are followed by a tab and then the value(s) that the user wishes to assign

# to that parameter. All comment lines begin with the '#' symbol.

…

# Minimum percentage of the initial data set to be set aside as an external test set.

Minimum test set percent 10

…

# The integer value assigned to test_set_selection_method determines how Modeler

# will select the compounds that will be used in the test set. A default value of '0'

# causes the test set to be sampled from the clusters of a Kohonen SOM. A value of '1'

# causes the test set to be sampled at fixed intervals from a sorted data set. A value

# of '2' causes the test set to be sampled at random.

Test set selection method 0

…

# The MinInputCV is the lowest acceptable variability, defined as coefficient of variation

# (CV expressed in %), in a descriptor. Descriptors with less CV than this value are removed

# as candidates.

Minimum input CV 1

…

Maximum MLR inputs 91

MLR Inputs step 8

…

# The output_variable parameter is the name of the dependent variable that the data is to be

# modeled against.

Output variable ExlogP

…

# The input_variables parameter lists the names of the descriptors and other user-provided data

# columns that you wish to have Modeler evaluate for inclusion into its models. The variables

# listed here are merely candidates: only a useful subset of them will actually be used to train

# the final neural network ensemble.

Input variables

MWt

MolVol

VMcGowan

N_Atoms

N_Carbon

N_Ntrgen

N_Oxygen

N_Phsphr

…

Users are advised to copy and edit an existing file created by ADMET Modeler rather than trying to compose one from scratch.