Log File (.log)
The LOG file records the history of a saved ADMET Modeler project, including the settings used, the model-building steps performed by the program, and a summary of the results. More detailed results are recorded in each of the other output files. Reloading a project lets you access its .log file through the Performance Grid page in ADMET Modeler, so that you can review the progress of the previous run.
The sample log file shown below has been modified for clarity. Ellipses (...) indicate places where multiple rows have been omitted.
ADMET Modeler(TM) log file. Version 0.0.0020
6/2/2005 1:34:25 PM
# ================================================================
IDENTICAL AND LOW VARIANCE INPUTS:6/2/2005 1:34:25 PM
18 inputs had a CV% of less than 1%:
A_StatF CV% = 00.000
...
A_IoniHC CV% = 00.992
A_IoniPi CV% = 00.000
# =================================================================
UNDERREPRESENTED INPUTS:6/2/2005 1:34:26 PM
NONE
# =================================================================
HIGHLY CORRELATED INPUTS: 6/2/2005 1:34:26 PM
A_ConjPi & A_NeuPDS correlation = +0.992352982288175
A_NeutCh & A_NeutPo correlation = -0.999999713226267
...
A_NeutCh & A_IoniRE correlation = +0.999999222874679
Number of correlated inputs = 10
# =================================================================
19 CANDIDATE INPUTS (after removing low CV, underrepresented, and highly correlated):
A_Estate
A_Access
A_ConjAr
...
# =================================================================
PROGRAM CONTROL SETTINGS:
Random number seed: -1
Minimum descriptor representation: 4
Minimum input CV: 1
Maximum input correlation: 0.99
…
3D input names: RgGrav__3D RgGeom__3D RadMax__3D Prop1__3D
MIRxx___3D MIRyy___3D MIRzz___3DSM2xx___3D SM2yy___3D
BoxX____3D BoxY____3D BoxZ____3D DStokes_3D Prop2__3D
…
# ================================================================================
More than 16 unique values were found in the outputs.
This variable is unlikely to be solved by classification methods but is suited for regression.
Data points = 64
Kohonen map size = 8x8
Kohonen mapping start time = 6/2/2005 1:34:26 PM
Kohonen map finish time = 6/2/2005 1:34:26 PM
Kohonen cell counts:
03 02 00 01 00 00 00 00
04 01 00 06 03 00 06 00
00 00 00 00 03 02 00 02
00 00 02 00 00 04 00 02
01 00 00 02 00 00 00 02
01 01 00 00 02 00 05 00
00 00 02 01 03 00 00 01
02 00 00 00 00 00 00 00
Kohonen Map of Output Means
3.09E-01 6.56E-01 0.00E+00 9.47E-01 0.00E+00 0.00E+00 0.00E+00 0.00E+00
4.91E-01 4.06E-01 0.00E+00 9.68E-01 9.23E-01 0.00E+00 5.67E-01 0.00E+00
0.00E+00 0.00E+00 0.00E+00 0.00E+00 8.69E-01 9.60E-01 0.00E+00 3.65E-01
0.00E+00 0.00E+00 5.27E-01 0.00E+00 0.00E+00 9.65E-01 0.00E+00 3.74E-01
3.10E-01 0.00E+00 0.00E+00 5.19E-01 0.00E+00 0.00E+00 0.00E+00 3.05E-01
1.32E-01 3.20E-01 0.00E+00 0.00E+00 9.63E-02 0.00E+00 4.80E-01 0.00E+00
0.00E+00 0.00E+00 4.35E-01 5.17E-01 3.93E-01 0.00E+00 0.00E+00 8.86E-02
5.65E-01 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
Total cells with 0 data points = 38 (.% of data set)
Total cells with 1 data points = 7 (10.9% of data set)
Total cells with 2 data points = 10 (31.3% of data set)
Total cells with 3 data points = 4 (18.8% of data set)
Total cells with 4 data points = 2 (12.5% of data set)
Total cells with 5 data points = 1 (7.8% of data set)
Total cells with 6 data points = 2 (18.8% of data set)
Training set data points (26):
2 3 4 7 8 9 10 15 18 19 20 …
Verify set data points (19):
6 12 13 24 27 29 32 33 34 36 39 …
Test set data points (19):
0 1 5 11 14 16 17 21 22 26 41 …
# ================================================================================
SENSITIVITY ANALYSIS BY INPUT GRADIENT:6/2/2005 1:34:26 PM
training baseline ANN with all valid inputs and 1 neurons . . .
# ================================================================================
ANN ensemble 6/2/2005 1:34:28 PM
ANNs: 32, Neurons: 1, Inputs: 10, Weights = 13
Training set size = 26
Verification set size = 19
Inputs:
A_NeutAP Sensitivity = 22.201 Relative sensitivity = 1
A_NeutAE Sensitivity = 18.884 Relative sensitivity = 0.850
A_NeutAQ Sensitivity = 15.349 Relative sensitivity = 0.691
…
# --------------------------------------------------------------------------------
Training Verify Test
RMSE 3.64E-01 3.73E-01 3.80E-01
MAE 2.90E-01 2.91E-01 2.93E-01
RSqd 9.77E-01 9.76E-01 9.85E-01
SRCC 9.22E-01 9.18E-01 9.49E-01
QSQD 9.06E-01 9.39E-01 9.74E-01
Slope 1.01E+00 9.99E-01 9.52E-01
Intercept 1.12E-02 -1.71E-03 3.02E-01
Elapsed time to train this ensemble = 1.48200000000099 seconds.
... [ANN ensemble section repeated for each network architecture]
# ================================================================================
Modeler (tm) run completed at 6/2/2005 1:34:48 PM
Elapsed time = 0.372883333333281 minutes.
# ================================================================================
The sections of the log file are as follows:
Identical and low variance Inputs lists all descriptors which, for the current data file (.dat/.out) were found to have little or no variation across the data set. The descriptors in this list are therefore eliminated during preprocessing, and aren’t offered for selection as dependent or independent variables.
Underrepresented Inputs lists all descriptors which were found to have few valid values. If a descriptor is represented in fewer data points than the Minimum Representation setting in the Advanced Settings tab, then that descriptor is eliminated during preprocessing, and not used for model building.
Highly Correlated Inputs lists pairs of descriptors which were found to be identical or highly correlated. This usually indicates that two descriptors are directly related, and only one or the other is needed for model building. The first of the pair is eliminated during preprocessing, and the second is available for model building.
\<n> Candidate Inputs lists the descriptors which have passed preprocessing and are available for model building.
Program Control Settings lists control settings coming from the Settings pull-down menu, the Basic Settings tab, and the Advanced Settings tab.
Kohonen Map lists the results of Kohonen mapping. This section shows the degree of diversity or clustering of the input data. The Kohonen clustering results in the initial selection of the training, verification, and test sets. The compounds in the three sets are listed. The index of each compound is its position in the initial data file (.dat/.out).
Sensitivity Analysis indicates the results of sensitivity analysis you selected.
Models/Ensembles summarizes the performance of a trained ensemble, and is repeated for each ensemble.
Run Completed is the elapsed time for the run.