Create Bayesian Model

Description
Builds a user model using Bayesian categorization.

Uses Bayesian categorization to build a model that distinguishes between "good" and "bad" ligands. "Good" ligands are defined by having a value of "1" for the property set by Property for Active.

A user model is created, named according to Model Name. The model can be used to predict how "good" ligands are using Calculate Molecular Properties.

Information
Name
Create Bayesian Model
Status
Success
User
Administrator
DS Version
3.5.0.12159
PP Version
8.5.0.200
DS Client Version
3.5.0.12158
Server Name
localhost (Windows64)
Server Ports
9944 (9943)
Start Time
01/04/21 09:52:27
Finish Time
01/04/21 09:53:05
Execution Time
00:00:38
Summary
ROC score is 0.843 (leave-one-out).
Best cutoff for this model is -1.074.
See ModelDescription.html for more detailed information about this model.
Test set validation: ROC score = 0.852226666795649
Model Rating: Accuracy 0.852: Good
Confusion Matrix: True Positives = 523, False Negatives = 165, False Positives = 121, True Negatives = 480
5-Fold Cross-Validation Result
Model Name ROC Score ROC Rating True Positive False Negative False Positive True Negative Sensitivity Specificity Concordance
111BayesianTempModel-10 0.835 Good 2350 404 168 2237 0.853 0.930 0.889
Validation Result Using External Test Set test1-1289.sd
Model Name ROC Score ROC Rating True Positive False Negative False Positive True Negative Sensitivity Specificity Concordance
111BayesianTempModel-10 0.852 Good 523 165 121 480 0.760 0.799 0.778
Parameters
Protocol Settings
 
Input Ligands
 
Input Test Ligands
 
Property for Active
Class
 
Model Name
111BayesianTempModel-10
Independent Properties
ECFP_14,Apol,Num_H_Donors,Num_Rings,Wiener
Calculable Properties
ECFP_14,Apol,Num_H_Donors,Num_Rings,Wiener
User Properties
Cross Validation
True
Folds
5
Learn Options
Validate Models
Remove Uninformative Bins
Equipopulate Bins
Model Domain Fingerprint
FCFP_2
Additional Properties
Advanced
Number of Bins
10

© 2021 Accelrys Software Inc.