PFAS compounds

Are replacement PFAS chemicals actually safer?

One natural-language query, seven federated knowledge graphs, an answer that joins water-system monitoring, adverse outcome pathways, gene-disease associations, and transcriptomic datasets — work that no single database can do.

Key finding

GenX, the chemical that replaced PFOA, is detected at 73% the frequency of legacy PFOS in U.S. water systems and was the most potent PPARα activator of 16 PFAS tested by the EPA — convergent evidence across seven graphs that compound-by-compound substitution is not reducing risk.

Graphs queried

SAWGraph, AOP-Wiki, SPOKE-OKN, NDE, Wikidata, GeneLab, and Ubergraph.

WOBD contribution

Links contamination observations to molecular mechanisms, disease pathways, and experimental datasets.

Audit trail

Each claim is framed as convergent evidence from named source graphs, not as a single-source assertion.

The question

Manufacturers replaced PFOA with GenX (HFPO-DA) and ADONA, marketed as safer alternatives. Are the replacements actually safer? Answering that requires linking water-system contamination data to molecular mechanism, then to gene-disease associations, then to experimental datasets — across four disciplines that traditionally live in different databases.

Without an OKN-backed federation, this kind of question lands on a research team that has to manually search EPA monitoring data, then mechanistic toxicology literature, then transcriptomic repositories like GEO, then disease association databases — weeks to months of cross-disciplinary work, with the synthesis still done by hand. The MCP server returns convergent evidence in minutes.

One query, federated answer

Natural-language query

Are replacement PFAS chemicals (GenX, ADONA) actually safer than the legacy compounds they replaced?

Unified MCP server federates seven KGs

  • sawgraph — water-system contamination monitoring
  • biobricks-aopwiki — adverse outcome pathways
  • spoke-okn — gene-disease associations
  • nde — transcriptomic dataset metadata
  • wikidata, spoke-genelab, ubergraph — functional annotation and ontology

Integrated answer

GenX is already detected at 73% the frequency of legacy PFOS in U.S. water systems and is the most potent PPARα activator of 16 PFAS tested — converging evidence across seven graphs that compound-by-compound regulation is not reducing risk.

7
federated KGs
25
PFAS in water systems
11
GenX transcriptomic datasets
168
enriched pathways

What the chat looks like

Excerpt from a conversation that produced these findings:

Are GenX and ADONA actually safer than the legacy PFAS they replaced? Pull together evidence from across the Proto-OKN graphs.

Querying sawgraph, biobricks-aopwiki, spoke-okn, nde, wikidata…

Convergent evidence from seven graphs plus PubMed: the replacements are not safer.

  • SAWGraph: GenX appears in 16,954 water-system observations — 73% the frequency of legacy PFOS.
  • PubMed (Evans et al., EPA 2022): in a 16-PFAS PPARα binding panel, GenX was the most potent activator.
  • NDE (GSE212294): in PPARα-knockout mice, GenX's hepatic effects disappear entirely — a purer PPARα activator than PFOA.
  • AOP-Wiki (AOP 166): PPARα activation → hepatocellular carcinoma. SPOKE-OKN: PPARA on liver and kidney cancer pathways.

The replacement chemical hits the same molecular target, more strongly, at concentrations already in U.S. drinking water.

Illustrative excerpt; phrasing is paraphrased and lightly edited from the upstream analysis. Live results may vary as graphs are updated.

What this query unlocks

  • Cross-domain joins. The query spans environmental monitoring (SAWGraph), mechanistic toxicology (AOP-Wiki), gene-disease associations (SPOKE-OKN), and experimental datasets (NDE) — four disciplines that traditionally live in separate portals.
  • Convergent evidence, not single-source claims. The conclusion that replacement PFAS hit the same molecular target as legacy compounds rests on independent lines from environmental data, EPA receptor assays, knockout-mouse transcriptomics, and pathway enrichment — each from a different graph.
  • Mechanism grounded in shared identifiers. Joins on gene symbols, Ensembl IDs, and chemical compound IDs let the assistant move from a chemical in water (SAWGraph) to a target receptor (PubMed + NDE) to a disease outcome (SPOKE-OKN, AOP-Wiki) without losing provenance.
  • Policy-relevant synthesis. The same federated answer that identifies a hazard also surfaces the candidate biomarkers (PPARA, ACOX1, FABP1, SREBF1) needed to monitor it — closing the loop from question to actionable recommendation.

Why this matters — public-health policy

Class-based chemical regulation has been a long-standing aim of public-health agencies, but the evidence required — environmental fate, toxicity mechanism, and disease association, all linked to specific compounds — has historically lived in separately funded, separately curated databases. Assembling it has meant standing up a multi-disciplinary working group for months. The federated MCP server makes a question of that shape a chat session, with provenance preserved back to each source graph.

The implication for emerging-contaminant policy is direct: as the next generation of PFAS replacements (and bisphenols, microplastics, and other chemical classes) enters the regulatory pipeline, the substrate to evaluate them is already in place. The federation also turns single-event policy questions into recurring queries — a regulator can rerun the analysis as new datasets land in the knowledge graphs, without rebuilding the workflow.

Read the full analysis

Specific datasets, mechanistic findings, disease pathways, and the convergence argument in detail.