eToxPred
commit deff5beEstimate toxicity risk and synthetic accessibility scores from SMILES for compound prioritization.
Estimate toxicity risk and synthetic accessibility scores from SMILES for compound prioritization.
eToxPred is a machine learning tool for predicting the toxicity and synthetic accessibility of small molecules from their chemical structures. Developed at Louisiana State University by Limeng Pu, Michal Brylinski, and colleagues, eToxPred filters out potentially toxic or difficult-to-synthesize compounds early in the drug discovery process.
The tool provides two complementary scores:
Together, these scores help prioritize which drug candidates are worth pursuing.
As such, we recommend using eToxPred for the following screening applications:
ProteinIQ provides a web-based interface for running eToxPred without command-line installation or Python environment configuration. Enter SMILES strings and receive toxicity and synthetic accessibility predictions.
| Input | Description |
|---|---|
Molecule | SMILES strings for compounds to analyze. Enter one SMILES per line, or use tab-separated format with compound names: CC(=O)Oc1ccccc1C(=O)O aspirin. Name-first rows such as aspirin CC(=O)Oc1ccccc1C(=O)O are also accepted. Supports file upload (.smi, .smiles, .txt, .csv) or PubChem batch fetching. |
The output is a spreadsheet with toxicity and synthetic accessibility predictions for each compound.
| Column | Description |
|---|---|
Compound ID | Name provided in input or auto-generated identifier (Compound_1, Compound_2, etc.). |
SMILES | The input SMILES string for reference. |
Toxicity Score | Probability of toxicity (0–1). Higher values indicate greater toxicity risk. |
SA Score | Transformed synthetic accessibility score (0–1). Higher values indicate easier synthesis. |
The Tox-score represents the probability that a compound exhibits general toxicity based on structural similarity to known toxic and non-toxic compounds.
| Tox-score | Risk level | Recommendation |
|---|---|---|
| 0.0–0.3 | Low | Proceed with standard testing |
| 0.3–0.5 | Moderate | Investigate structural features |
| 0.5–0.7 | Elevated | Consider structural modifications |
| 0.7–1.0 | High | Likely requires redesign |
The optimal discrimination threshold is 0.58, which most effectively separates toxic from non-toxic compounds in validation studies. FDA-approved drugs have a median Tox-score of approximately 0.34, while known toxins from the T3DB database typically score above 0.6.
The SA score is eToxPred’s transformed synthetic accessibility estimate. Higher values indicate compounds predicted to be easier to synthesize; lower values indicate greater synthetic difficulty. The source does not define universal decision bands for this transformed value, so use it primarily to rank compounds analyzed with the same model and confirm promising candidates with synthesis planning or expert review.
eToxPred combines toxicity prediction from an Extremely Randomized Trees (Extra Trees) classifier with synthetic accessibility scoring based on RDKit molecular fragments and complexity penalties.
The toxicity model was trained on 4,550 compounds: 1,515 FDA-approved drugs representing the non-toxic class and 3,035 compounds from TOXNET representing the toxic class. Independent validation used 3,682 compounds from KEGG-Drug (non-toxic) and 1,283 compounds from T3DB (toxic).
Each molecule is converted to a 1024-bit Morgan fingerprint with radius 2 using RDKit. These binary fingerprints encode the presence or absence of structural fragments, capturing the chemical features relevant to toxicity.
The current trained Extra Trees model uses the published eToxPred model archive. Key hyperparameters reported with the model:
The ensemble votes on classification, with the final Tox-score representing the proportion of trees predicting toxicity. This approach handles noisy biological data well and resists overfitting.
The SA score combines historical synthetic knowledge with complexity penalties. The current eToxPred implementation calculates an Ertl-Schuffenhauer-style raw score and transforms it with:
This transformation means higher displayed values correspond to easier predicted synthesis.
The fragment score compares molecular substructures against fragments frequently found in known synthesized compounds. Common fragments score higher (easier to make); unusual fragments score lower.
The complexity penalty accounts for structural features that complicate synthesis:
eToxPred was validated on independent test sets not used during training.
| Metric | Value |
|---|---|
| Accuracy | 72.1% |
| Sensitivity (true positive rate) | 63.1% |
| Specificity | 75.2% |
| Matthews Correlation Coefficient | 0.35 |
| ROC AUC | 0.82 |
The model was also evaluated on datasets for specific toxicity types:
| Endpoint | AUC | Accuracy |
|---|---|---|
| Acute oral toxicity | 0.80 | 85.4% |
| Cardiotoxicity | 0.80 | 79.8% |
| Endocrine disruption | 0.75 | 74.4% |
| Carcinogenicity | 0.72 | 72.2% |
The current SA score calculation follows the Ertl-Schuffenhauer synthetic accessibility approach used by the eToxPred package, then transforms the score so higher 0–1 values indicate easier synthesis.
Several tools on ProteinIQ address overlapping aspects of compound evaluation:
eToxPred provides general toxicity screening with synthetic accessibility in a single analysis. The machine learning model captures patterns across diverse toxic compounds but does not distinguish between specific toxicity mechanisms.
ADMET-AI uses graph neural networks to predict 41 specific ADMET endpoints, including hERG inhibition (cardiotoxicity), hepatotoxicity, CYP interactions, and plasma protein binding. For endpoint-specific toxicity predictions, ADMET-AI offers more detailed information.
Toxicity Prediction uses rule-based structural alerts (PAINS, Brenk filters) rather than machine learning. This approach identifies specific problematic substructures like reactive groups or known interference patterns. The two approaches are complementary—eToxPred captures general toxicity patterns while structural alerts identify specific problematic features.
Lipinski's Rule of 5 evaluates oral bioavailability potential using simple physicochemical rules (molecular weight, LogP, hydrogen bond donors/acceptors). This rule-based approach is interpretable but does not predict toxicity.
QEPPi scores drug-likeness specifically for protein-protein interaction inhibitors, which require different physicochemical properties than conventional drugs.
A typical drug discovery screening workflow incorporating eToxPred:
Try aspirin (acetylsalicylic acid), CC(=O)Oc1ccccc1C(=O)O, and caffeine, Cn1cnc2c1c(=O)n(c(=O)n2C)C, in one job to compare their predictions. Treat the scores as model outputs for prioritization rather than experimental safety or synthesis guarantees.
Predict absorption, distribution, metabolism, excretion, and toxicity properties for compound libraries.
Profile small-molecule ADMET properties across 22 Chemprop-based prediction models.
Identify toxic and reactive fragments using Brenk structural alert filter.
Screen for lead-like compounds using stricter molecular descriptor criteria for early optimization.
Screen compounds for Pan-Assay Interference patterns using PAINS_A, PAINS_B, and PAINS_C filters.
Evaluate PPI inhibitor potential and drug-likeness from molecular descriptors and SMILES inputs.
Screen compounds for structural toxicity alerts using PAINS filters, BRENK filters, and NIH filters for safer drug discovery.
Evaluate oral bioavailability from rotatable bonds, polar surface area, and molecular descriptors.
Screen compounds for oral bioavailability using molecular weight, LogP, hydrogen bond donor, and acceptor limits.
Predict protein ligand-binding residues from AlphaFold2 pair representations