Program operation is controlled by configuration
file cfg that is formed by the interface Java applet. Data file (see download filed on the applet) containing the descriptor and activity values
and model file (can be only avaialble if mone saved it from previous analysis) are used to provide data and models.
Text calculation results are returned in file comp and the data for illustrative plots are output to file graphics.
The format of these files is specified below. Sample input and output files
are provided with the software.
Configuration file consists of a number of "KEYWORD = VALUE" pairs given in separate lines.
This file is available on page with calculated results.
The possible keywords and values of parameters are listed in the table below.
|
Keyword
|
Type
|
Default
|
Description
|
|
INPUTS
|
INTEGER
|
*
|
Number of independent
variable (descriptor) columns. The next data column is used for
activity/property values, if necessary [MIN=1 MAX=32767]
|
|
ROWS
|
INTEGER
|
*
|
Number of compound
rows [MIN=1 MAX=32767] |
|
NAMES
|
INTEGER
|
0
|
Compound names are listed
in the first (additional) column: 0-no 1-yes
|
|
VARIABLE_NAMES
|
INTEGER
|
0
|
Variable (descriptor)
names are listed in the first (additional) row: 0-no 1-yes
|
|
MODE
|
INTEGER
|
0
|
Mode: 0-calibrate (build
model); 1-predict; 2-validate prediction. Calibrate and Validate
modes require the additional data column for activity/property values
|
|
MAX_PLS_FACTORS
|
INTEGER
|
10
|
Maximum number of extracted PLS factors [MIN=1 MAX=50]
|
|
RESIDUAL_VARIANCE |
FLOAT
|
0.001
|
Minimum relative residual
variance of PLS factors [MIN=0 MAX=1]
|
|
KEEP_PLS_FACTORS
|
INTEGER
|
0
|
Number of PLS factors
to keep in saved models; 0-choose model with highest predictivity (Q2) [MIN=0
MAX=20]
|
|
CONSTANT
|
INTEGER
|
2
|
Remove low-variable
descriptors that have the same (constant) value for most of the
compounds in the training set (all but the number specified by this
parameter); 0-do not remove [MIN=0 MAX=5]
|
|
GENETIC
|
INTEGER
|
1
|
Genetic selection of descriptors
(predictivity - guided optimization of descriptor subset by
means of the genetic algorithm): 0-no 1-yes
|
*Mandatory parameters.
