Which endpoints can be predicted in silico?

We routinely cover mutagenicity, carcinogenicity, reproductive and developmental toxicity, skin and eye irritation, sensitization, acute and repeated-dose organ toxicity, and ecotoxicological endpoints. Coverage for any specific endpoint is confirmed during scoping, because it depends on the availability of validated models and supporting data.

Are in silico results accepted by regulators?

Yes, when they are generated under the OECD principles for (Q)SAR validation and supported by adequate documentation. Acceptance depends on the regulatory framework and the endpoint in question, so our reports make the model, the applicability domain, and the uncertainty explicit. We recommend confirming expectations with your regulator during scoping.

How do you handle compounds outside a model's applicability domain?

We report an applicability-domain score for every prediction and flag compounds that fall outside it. For those cases we rely on read-across from structurally and mechanistically similar analogues, or recommend targeted experimental testing rather than overstating what the model output supports.

Can you work with confidential structures?

Yes. Structures and study data are handled under confidentiality terms, and we can work with internal identifiers rather than compound names. Please confirm NDA coverage with us before sharing proprietary structures or unpublished study results.

How long does a standard assessment take?

Turnaround is scoped during consultation and depends on the number of compounds, the endpoints requested, and whether a read-across case is required. Please contact our team for a timeline specific to your project.

Services

Online Inquiry

Name *

Email *

Subject *

Phone

Your Message

How Did You Hear About Us

Computational Toxicology Services

In Silico Safety & Toxicology

Predict toxicity, prioritize compounds, and assemble defensible safety arguments from structure and mechanistic data. Our in silico workflows support every stage from early lead triage through regulatory submission.

Structure- and ligand-based prediction across major toxicity endpoints
OECD-aligned (Q)SAR models with documented applicability domains
Read-across and weight-of-evidence dossiers for regulatory submission
Mechanistic interpretation linked to adverse outcome pathways

Request a Safety Assessment

Why Choose Our Computational Toxicology Service?

We combine curated toxicology databases, validated models, and expert review so that every prediction can be traced back to its evidence.

Expert toxicologists review every model output before reporting
Endpoint coverage aligned with ICH, REACH, and OECD guidance
Transparent model documentation with applicability-domain statements
Confidential handling of proprietary structures and study data

Auditable

Regulatory

Reproducible

Our Approach

Evidence-Led In Silico Safety Assessment

A prediction on its own is not a safety argument. We curate the input structures, run complementary models, quantify the uncertainty, and then place the results in mechanistic and regulatory context so that every conclusion can be defended.

Toxicology Prediction Pipeline

A structured workflow that converts structures and existing experimental data into endpoint-specific predictions with quantified confidence.

Structure curation, salt stripping, and tautomer standardization
Physicochemical, ADME, and pharmacokinetic property profiling
Endpoint prediction using structural alerts and (Q)SAR models
Applicability-domain scoring for every predicted value

Read-Across & Weight of Evidence

Where analogues or mechanistic data are stronger than a model, we build a documented read-across case rather than forcing a prediction.

Analogue identification by structural and mechanistic similarity
Data-gap analysis against the relevant regulatory endpoint
Uncertainty assessment with explicit confidence categories
Weight-of-evidence narrative ready for submission

Regulatory Aligned

Deliverables are structured around ICH, REACH, and OECD frameworks so they slot directly into a dossier.

Mechanistically Informed

Predictions are interpreted against adverse outcome pathways rather than reported as isolated numbers.

Transparent

Every model, version, input file, and applicability-domain statement is documented and reproducible.

Deliverables

What You Receive

A complete toxicology data package, documented so that a reviewer can trace every conclusion back to its input and the model that produced it.

Data Deliverables

Curated structure file with a standardization and QC log
Endpoint prediction tables with confidence and applicability-domain scores
Supporting experimental data drawn from public toxicology databases
Machine-readable results (CSV or SDF) for use in your own models

Reporting & Support

Toxicity assessment report with explicit confidence statements
Read-across and weight-of-evidence justification where applicable
Model documentation suitable for regulatory or internal audit
Scientific consultation to interpret results with your team

Service Scope

Customizable Assessment Parameters

Each assessment is scoped individually after a consultation. The table below outlines typical parameters and options.

Parameter	Typical project scope
Input Data	SMILES, SDF, or InChI structures; new compounds, impurities, degradants, or metabolites; single compounds or screening libraries
Compound Classes	Small molecules, impurities and degradants, metabolites, or structurally related analogue series
Endpoint Coverage	Genotoxicity, carcinogenicity, reproductive and developmental toxicity, organ toxicity, and ecotoxicity
Molecular Descriptors	2D and 3D descriptors, structural fingerprints, and mechanistic structural alerts
Model Suite	Statistical (Q)SAR, rule-based expert systems, and machine-learning models, applied in combination
Applicability Domain	Similarity-based assessment with documented thresholds and explicit confidence categories
Read-Across	Analogue selection, data-gap analysis, and weight-of-evidence justification when required
Reporting Standard	ICH M7, REACH Annex XI, OECD principles for (Q)SAR validation, or client-specified internal criteria
Technical Support	Subject to customer service consultation

Workflow

From Structure to Defensible Conclusion

Our five-step process turns structures and experimental data into an assessment a reviewer can follow, endpoint by endpoint and model by model.

Structure Curation & QC

Normalize structures, strip salts, standardize tautomers, and confirm stereochemistry before any prediction is run.

Property & ADME Profiling

Calculate the physicochemical and ADME parameters that govern exposure and inform endpoint-specific predictions.

Endpoint Prediction

Apply complementary (Q)SAR, expert-rule, and machine-learning models to each endpoint and score the applicability domain.

Read-Across & Uncertainty

Where model coverage is limited, build a documented analogue-based argument and quantify the remaining uncertainty.

Weight of Evidence & Reporting

Integrate predictions, experimental data, and mechanistic reasoning into a defensible assessment report.

Applications

Where In Silico Assessment Pays Off

Our computational toxicology workflows are used across discovery, process development, and regulatory submission.

Qualify Impurities Without New Studies

Support ICH M7 assessments by combining structural alerts, (Q)SAR predictions, and analogue data to classify impurities and justify a control strategy.

Bacterial mutagenicity prediction with complementary models
Class assignment and cohort-of-concern analysis
Documentation aligned with ICH M7 submission expectations

Fail Fast on Avoidable Liability

Rank series and deprioritize compounds with unfavourable safety profiles before synthesis and in vivo work consume the budget.

Consensus predictions for a full library in a single batch
Side-by-side comparison of analogues and series
Flags for structural alerts linked to known liabilities

Strengthen Submission Packages

Provide the in silico evidence and justification that regulators expect when animal testing is reduced or waived.

OECD-aligned validation documentation for each model
Applicability-domain and uncertainty statements
Read-across justification with a weight-of-evidence narrative

FAQ

Frequently Asked Questions

Common questions about our computational toxicology service.

We routinely cover mutagenicity, carcinogenicity, reproductive and developmental toxicity, skin and eye irritation, sensitization, acute and repeated-dose organ toxicity, and ecotoxicological endpoints. Coverage for any specific endpoint is confirmed during scoping, because it depends on the availability of validated models and supporting data.
Yes, when they are generated under the OECD principles for (Q)SAR validation and supported by adequate documentation. Acceptance depends on the regulatory framework and the endpoint in question, so our reports make the model, the applicability domain, and the uncertainty explicit. We recommend confirming expectations with your regulator during scoping.
We report an applicability-domain score for every prediction and flag compounds that fall outside it. For those cases we rely on read-across from structurally and mechanistically similar analogues, or recommend targeted experimental testing rather than overstating what the model output supports.
Yes. Structures and study data are handled under confidentiality terms, and we can work with internal identifiers rather than compound names. Please confirm NDA coverage with us before sharing proprietary structures or unpublished study results.
Turnaround is scoped during consultation and depends on the number of compounds, the endpoints requested, and whether a read-across case is required. Please contact our team for a timeline specific to your project.

Scientific Literature

References

Pawar G, Madden JC, Ebbrell D, et al. In Silico Toxicology Data Resources to Support Read-Across and (Q)SAR. Frontiers in pharmacology. 2019;10:561. View on PubMed
Zhang J, Li H, Zhang Y, et al. Computational toxicology in drug discovery: applications of artificial intelligence in ADMET and toxicity prediction. Briefings in bioinformatics. 2025;26(5):bbaf533. View on PubMed
Fischer I, Milton C, Wallace H. Toxicity testing is evolving! Toxicology research. 2020;9(2):67-80. View on PubMed

Please kindly note that our products and services are for research use only.