PROVENANCE · SOFTWARE · DATA

Sources, data, software, and research methods

EvidenceHelix should be able to answer not only “what does the platform claim?” but also “where did the information come from, what code processed it, what version was used, and what assumptions were introduced?” This page is the public provenance shelf.

A citation to software documents the computational method; it does not independently validate a scientific conclusion. Scientific claims still require scientific evidence.

Cite the platform itself

EvidenceHelix now ships a root CITATION.cff file for machine-readable software citation. See the citation and reproducibility guide →

Core software stack

Technology	Role in EvidenceHelix	Official source
Python	Scientific services, evidence processing, mathematics, ingestion, tests, and research tooling.	Python documentation
FastAPI	Python API layer for medical AI, corpus, discovery, intelligence, mathematics, and scientific engines.	FastAPI documentation
Pydantic	Typed request/response models and validation boundaries.	Pydantic documentation
Uvicorn	ASGI application server for Python services.	Uvicorn documentation
React	Interactive scientific workspaces and application UI.	React documentation
TypeScript	Typed frontend application code and interface contracts.	TypeScript documentation
Vite	Frontend development and production build system.	Vite documentation
Three.js	Web 3D rendering for anatomical and organism visualization.	Three.js documentation
Java + Spring	Backend research API and domain services.	Spring Boot documentation
DuckDB	Local analytical queries over research datasets and tabular evidence.	DuckDB documentation
pytest	Regression and scientific-behavior tests for Python services.	pytest documentation
Docker	Containerized, reproducible runtime packaging.	Docker documentation
GitHub Actions	CI verification and automated build/test workflows.	GitHub Actions documentation
Azure Container Apps	Production container hosting for APIs and services.	Microsoft documentation

Major scientific and biomedical source families

Source	What it contributes	Reference
Human Phenotype Ontology (HPO)	Standardized phenotype vocabulary and disease/phenotype relationships.	HPO
Open Targets	Target–disease evidence integrating genetics, genomics, drugs, and literature-derived evidence.	Open Targets Platform
RNAcentral	Integrated non-coding RNA sequence and annotation resources.	RNAcentral
NCBI / NLM	Biomedical literature, genes, variants, clinical terminology, drug and consumer health references.	NCBI · NLM
DailyMed	Structured drug labeling distributed by the U.S. National Library of Medicine.	DailyMed
FDA	Regulatory drug, safety, device, labeling, and approval information.	FDA
NIH NCCIH	Evidence and safety references for complementary health approaches, herbs, and botanicals.	NCCIH
ClinicalTrials.gov	Clinical study registration, status, interventions, outcomes, and results when posted.	ClinicalTrials.gov
Smithsonian Open Access	Open cultural and natural-history collections, including 3D assets and metadata for organism/dissection research where licenses permit.	Smithsonian Open Access
Wikidata / Wikipedia	Fallback identifiers, cross-links, and general reference context; not treated as equivalent to primary biomedical evidence.	Wikidata · Wikipedia

What provenance should preserve

Source identitypublisher, dataset, document, URL/identifier

Release/versionprovider revision, date, software version, model version

Rightslicense, attribution, redistribution limits

Transformationdownload → parse → normalize → derive → index

Integrityhashes, immutable raw copies, reproducible derivatives

Claim linkagewhich exact evidence supports, challenges, or fails to distinguish a claim

Computational citation pattern

A reproducible EvidenceHelix analysis should be able to cite a scientific source and the computational path that transformed it. Example:

claim
 ├─ source: PMID / DOI / dataset release
 ├─ extracted passage / rows
 ├─ transformation: parser + version
 ├─ analysis: algorithm + parameters
 ├─ software: Python / package versions
 ├─ output hash
 └─ interpretation + uncertainty

This makes “we used Python” part of the audit trail without confusing the programming language with scientific evidence. 😄

Research-method references

EvidenceHelix also documents methods such as Bayesian model comparison, causal-invariance screening, change-point detection, multiple-testing control, expected-information-gain experiment planning, contradiction preservation, and replication tracking. Each method should be linked to an implementation version and, where appropriate, a primary methodological reference.

See the Foundational Consciousness & Soul Inquiry for an example of how competing models, anomalies, falsifiers, and experiments are kept separate.