Predicted vs Observed (.pno)

The PNO file records the values output predictions from the trained model, and compares them to the input observed values. You can import the data they contain using FILE >> Import >> Compound Attributes... and navigating to the desired PNO file. It contains the following elements:

Program identification

Model Type: \<model type>

Neurons: \<number of neurons>

Inputs: \<number of descriptors>

Output shift: \<output shift>

Output scale: \<output scale>

Output labels: \<output categorical labels>

CFA parameters: \<confidence analysis parameters>

CFA avg & min: \<average & minimum confidence>

NAME SET PRED NORM OBSD NORM PRED ABS OBSD ABS

\<name> \<set> \<value> \<value> \<value> \<value>

…

Column meanings:

NAME: compound name

SET: the compound was put into training/verification (Trn/Ver) or test (Test) set

For regression models:

PRED NORM: normalized predicted output

OBSD NORM: normalized observed input

PRED ABS: absolute predicted output

OBSD ABS: absolute observed input

For classification models:

PRED: predicted output

OBSD: observed input

PVOTE: the number of positive votes received

Part of a .pno file is shown below (modified for clarity).

ADMET Modeler(TM) 4.1.0065 Predicted & Observed file 6/10/2010 10:56:37 AM

Model Type: ANNE Regression

Neurons: 2

Inputs: 8

Output shift: 0.374398460057748

Output scale: 9.62463907603465E-02

NAME SET PRED NORM OBSD NORM PREDABS OBSDBS

Data1 Test 0.606171394565825 0.526467757459095 2.40812078953892 1.58

Data2 Test 0.275156240931478 0.253128007699711 -1.03112665672194 -1.26

Data3 Trn/Ver 0.270319719956947 0.358036573628489 -1.08137810964732 -0.17

Data4 Test 0.638471833646776 0.639076034648701 2.74372235159 2.75

Data5 Trn/Ver 0.297606445894955 8.37343599615015E-02 -0.797869027151422 -3.02

Data6 Trn/Ver 0.116706975925972 2.61834824655249E-17 -2.67741452012916 -3.89

Data7 Trn/Ver 0.558984513551254 0.610202117420597 1.91784909579753 2.45

Data8 Test 0.638062487414218 0.838306063522618 2.73946924423372 4.82