Compounds evaluated
5Risk signals detected
11Out of domain issues
3Report status
COMPLETEDPredictions by Organ
The label compares the predicted probability with each model’s threshold. The domain status indicates whether the prediction is sufficiently reliable.
| Compound | Heart | Liver | Nervous System | Kidney |
|---|---|---|---|---|
|
Allopurinol
CHEMBL1467
|
Risk 0.88
Within domain
|
Risk 0.92
Within domain
|
Risk 0.70
Within domain
|
No risk 0.53
Within domain
|
|
Bisphenol A
CHEMBL418971
|
Risk 0.87
Within domain
|
Risk 0.87
Within domain
|
Risk 0.87
Within domain
|
No risk 0.29
Within domain
|
|
Triclocarban
CHEMBL1076347
|
Risk 0.78
Within domain
|
Risk 0.97
Within domain
|
Risk 0.79
Within domain
|
Risk 0.63
Within domain
|
|
(S)-3-Acetylamino-2,2-difluoro-4-phenyl-butyric acid methyl
ester
CHEMBL300764
|
Risk 0.86
Out of domain
low reliability
|
No risk 0.33
Out of domain
low reliability
|
No risk 0.30
Out of domain
low reliability
|
No risk 0.29
Within domain
|
Hydroxyitraconazole was excluded because it has no associated protein targets.
Applicability Domain
The Applicability Domain (AD) evaluates whether a query compound is similar enough to the model's training data to ensure a reliable prediction. If a compound lies beyond the dataset's coverage, the model’s prediction carries higher uncertainty. To evaluate this, two plots are shown below:
Heart analysis
PCA Projection Map
Projection of the model’s latent space
Limit Distribution
Distance to the statistical threshold
Liver analysis
PCA Projection Map
Projection of the model’s latent space
Limit Distribution
Distance to the statistical threshold
Nervous System analysis
PCA Projection Map
Projection of the model’s latent space
Limit Distribution
Distance to the statistical threshold
Kidney analysis
PCA Projection Map
Projection of the model’s latent space
Limit Distribution
Distance to the statistical threshold