Solutions · By job to be done
Characterize a chemical with no history from its structure alone.
A novel substance arrives with no LD50, no NOAEL, no mutagenicity result, and a regulator still wants evidence before you manufacture or import. Whether it is a TSCA PMN, a REACH inquiry, or a K-REACH notification, ToxIndex builds a hazard picture from the structure up.
Updated July 18, 2026
- Structure-in
- SMILES, InChI, or MOL to a hazard picture
- 100+
- tools across the endpoint set
- 79
- structural-alert profilers for grouping
The situation
Starting from a structure
All you have is a SMILES, an InChI, or a MOL file. From there the platform runs predictors across the endpoints that matter, including the open-source ToxTransformer, each prediction carrying an applicability-domain assessment so you know whether the chemical sits inside the model's reliable range.
When predictions alone are not enough, analogs carry the weight. The knowledge graph finds structural neighbors, the QSAR Toolbox adds mechanism-based grouping on top, and read-across is generated with the similarity justification a reviewer will ask to see.
How ToxIndex helps
Building the weight of evidence
Predict across the endpoints
ToxJobs scores the structure against validated models, each with a confidence and an applicability-domain read, so uncertainty stays visible.
Find mechanistically relevant analogs
The QSAR Toolbox groups by metabolic and mechanistic similarity, the justification regulators expect to see.
Pull every analog's data
Kiln returns the experimental record and regulatory classifications for each analog, assembled into a matrix of what is known where.
Show your work
Flow captures every query, model run, and decision, so an analog choice or a model selection stands up when a reviewer questions it.
The tools behind it
What runs underneath
FAQ
Questions teams ask
Frequently asked questions
Can ToxIndex assess a chemical with no test data?
Yes. Starting from a SMILES, InChI, or MOL file, it runs predictors across the relevant endpoints and builds read-across from mechanistically relevant analogs, each carrying an applicability-domain read so uncertainty stays visible.
How does it find analogs?
The knowledge graph finds structural neighbors and the QSAR Toolbox adds metabolic and mechanistic grouping across 79 structural-alert profilers, so an analog choice rests on mechanism, not just fingerprint distance.
Which regulatory programs does this support?
The same structure-up hazard picture supports a TSCA PMN, a REACH inquiry, or a K-REACH notification.
See it on your own chemistry.
Try the free version now, or book a walkthrough of the full platform on your substances.