This guide covers everything from your first login to exporting assessment results.

Before you start

You’ll need three things:

A ToxIndex account (reach out to sales@toxindex.com for access)
Chemical identifiers for the substances you want to assess: CAS numbers, SMILES strings, InChIKeys, or IUPAC names work
Your regulatory context (REACH, TSCA, cosmetics regulation) so ToxIndex can tailor the assessment

Your first query

The quickest way to get started is the conversational interface. Log in, open Sage, and ask a question about a specific chemical.

Try something like: “What hazard data exists for CAS 80-05-7?” (that’s bisphenol A).

ToxIndex searches the knowledge graph and returns available experimental data, regulatory classifications, and study references for the substance. From there, you can ask follow-ups. For example: “What are the data gaps for a REACH registration at the 100-1000 tonne band?” The system compares available data against REACH Annex IX/X requirements and shows you what’s missing.

Running a full assessment

For a complete, reproducible assessment, use Flow’s pipeline system.

Specify your regulatory context (REACH tonnage band, TSCA PMN, cosmetics safety report)
Enter your substance identifier
ToxIndex runs the full pipeline: data gathering from the knowledge graph and regulatory document search, gap analysis, computational predictions for missing endpoints, weight-of-evidence evaluation, and report generation
You’ll get a notification when it’s done

The pipeline handles sequencing automatically. You don’t need to decide which tools to run or in what order.

Reading the results

Results are organized by endpoint. For each one, you’ll see:

Experimental data from the knowledge graph with source citations, study quality scores, and key results (NOAEL, LD50, positive/negative outcome)
Computational predictions with confidence levels, applicability domain assessment, and which model(s) were used
Read-across evidence from structural analogs identified by the knowledge graph and QSAR Toolbox, with similarity scores and justification
Weight-of-evidence conclusion combining all the above into a synthesized assessment with an overall confidence rating
Audit trail showing every data query, model run, and decision point that produced the result

Digging deeper

Click any endpoint result to expand the full evidence chain.

Experimental data shows the original study reference, species, route, duration, and result with units
Predictions show the model name, version, training data summary, applicability domain analysis, and uncertainty bounds
Read-across shows analog structures, similarity metrics, available data for each analog, and the rationale

Everything links back to its source in the knowledge graph or regulatory document search. You can always trace a finding to its origin.

Exporting

IUCLID format for direct submission to ECHA (REACH, BPR)
PDF report for internal review or regulatory communication
CSV or JSON for integration with your own systems
Audit trail export for QA review or regulatory inspection

Getting Started with ToxIndex