Research prototype for semantic toxicology evidence access

ToxCastLite

Assay-scoped SQLite evidence stores, RDF/SPARQL semantic discovery, ToxRefDB in vivo evidence, CPDat product-use context, and drill-down to concentration-response evidence.

Scroll ↓

What is ToxCastLite?

ToxCastLite is a research prototype and methodological use case for portable semantic access to public toxicology evidence.[10] It turns the large ToxCast database (invitrodb) resource into assay-scoped SQLite evidence stores and a compact RDF semantic layer. ToxCast provides high-throughput screening bioactivity data and concentration-response model outputs for toxicological prioritization and interpretation.[1][2] ToxCastLite links these data to ToxRefDB in vivo evidence and CPDat product-use context through DSSTox identifiers. Dense numerical evidence remains in SQLite, while a standards-compliant RDF/SPARQL layer exposes chemicals, assays, endpoints, model results, quality flags, in vivo evidence, and use-context records for semantic querying.

The corresponding preprint is available on bioRxiv with DOI 10.64898/2026.05.16.724895 .

Why it matters

Portable evidence stores

Build smaller SQLite databases from the public invitrodb release for focused assay domains such as DNT-IVB, Developmental Toxicity, Thyroid Bioactivity, or other curated ToxCast assay-list scopes.[3]

Semantic access

Query chemicals, endpoints, model results, flags, and assay scopes through an RDF/SPARQL layer instead of reconstructing complex relational joins by hand. GraphDB was used as the reference backend in the research prototype, but the semantic layer is designed around standards-compliant RDF and SPARQL. [8]

ToxRefDB in vivo evidence

Link in vitro bioactivity to processed animal-study evidence, point-of-departure records, effect summaries, and observation summaries from ToxRefDB v3.0 for transparent evidence-gap and concordance analysis.[4][5]

CPDat context

Link bioactivity evidence to functional-use, list-presence, and product-use context from CPDat v4.0 through DSSTox identifiers, while preserving the distinction between bioactivity, use-context evidence, and measured exposure.[6]

Evidence drill-down

Move from graph-level results back to SQLite concentration-response records for expert review, plotting, and curve-level inspection. Optional R-based workflows can support concentration-response modeling and visualization.[7]

Replaceable RDF backend

GraphDB is treated as a reference RDF/SPARQL backend for the prototype, not as a bundled component of ToxCastLite. The RDF projection, SPARQL examples, and SQLite evidence stores are separate project artifacts and can be adapted to other standards-compliant triple stores.

Designed for local, auditable workflows

ToxCastLite is intended for regulatory and scientific users who need transparent, reproducible, local access to toxicological evidence. The system supports direct SPARQL, schema-grounded query assistance, and traceable drill-down to numerical records. Rather than replacing expert judgement, it acts as a local evidence browser: the RDF backend helps identify relevant chemicals, endpoints, model results, flags, ToxRefDB evidence, and CPDat context, while SQLite preserves the detailed numerical and source-level records needed for review.

Read the bioRxiv preprint Project contact Licensing note References

Core querying and drill-down are designed for local execution without external API dependencies. This statement refers to the prototype architecture and does not grant, transfer, or imply rights to redistribute third-party software.

Research prototype and GraphDB licensing note

The ToxCastLite paper and this project page describe a research prototype and scientific use-case demonstration for semantic toxicology evidence integration. They do not distribute, bundle, sublicense, resell, or provide GraphDB as part of a commercial product, installer, Docker image, or hosted service.

Research prototype No GraphDB redistribution No sublicense grant Replaceable RDF/SPARQL backend

Reference implementation

GraphDB was used as the RDF/SPARQL backend for the reported prototype and case studies. In the architecture, it functions as a semantic query layer over project-generated RDF and local SQLite evidence stores.

Third-party license

GraphDB is third-party software by Ontotext / Graphwise and remains subject to the applicable GraphDB license terms, including the terms for GraphDB Free and GraphDB Enterprise.[9] Any installation, production use, commercial use, or redistribution of GraphDB must be licensed separately by the deploying user or institution.

Project artifacts

ToxCastLite-generated RDF, SPARQL examples, SQLite builders, derived evidence stores, and schema documentation are conceptually separate from the GraphDB software. They are intended to remain usable with standards-compliant RDF triple stores and SPARQL endpoints.

This project page is an implementation and research note. It is not a GraphDB license grant and does not modify Ontotext or Graphwise licensing terms.

References

  1. Dix, D. J., Houck, K. A., Martin, M. T., Richard, A. M., Setzer, R. W., & Kavlock, R. J. (2007). The ToxCast program for prioritizing toxicity testing of environmental chemicals. Toxicological Sciences, 95(1), 5-12.
  2. Richard, A. M., Judson, R. S., Houck, K. A., Grulke, C. M., Volarath, P., Thillainadarajah, I., Yang, C., Rathman, J., Martin, M. T., Wambaugh, J. F., et al. (2016). ToxCast chemical landscape: paving the road to 21st century toxicology. Chemical Research in Toxicology, 29(8), 1225-1251.
  3. EPA's Center for Computational Toxicology and Exposure. (2018). ToxCast Database: invitrodb version 4.3. doi: 10.23645/epacomptox.6062623.v14.
  4. EPA's Center for Computational Toxicology and Exposure. (2018). ToxRefDB v3.0.
  5. Feshuk, M., Kolaczkowski, L., Watford, S., & Friedman, K. P. (2023). ToxRefDB v2.1: update to curated in vivo study data in the Toxicity Reference Database. Frontiers in Toxicology, 5, 1260305.
  6. Handa, S., Isaacs, K. K., Wall, J. T., Larger, A., Burns, S., Koval, L. E., Baron-Furuyama, K., Elonen, C. M., Lyons, D., Dionisio, K. L., et al. (2025). The Chemical and Products Database v4.0, an updated resource supporting chemical exposure evaluations. Scientific Data, 12(1), 950.
  7. Sheffield, T., Brown, J., Davidson, S., Friedman, K. P., & Judson, R. (2022). tcplfit2: an R-language general purpose concentration-response modeling package. Bioinformatics, 38(4), 1157-1158.
  8. Angles, R., & Gutierrez, C. (2008). Survey of graph database models. ACM Computing Surveys, 40(1), 1-39.
  9. Ontotext / Graphwise. GraphDB documentation: licensing information for GraphDB Free and GraphDB Enterprise.
  10. Dönmez, A., Nosov, O., Heck, K., Mosig, A., Fritsche, E., & Koch, K. (2026). ToxCastLite: a portable semantic evidence graph linking in vitro bioactivity, in vivo toxicity, and exposure-use context. bioRxiv. doi: 10.64898/2026.05.16.724895 .