Raw measurement from a system of record.
"James's CBC from Quest Diagnostics"
Precision
Infrastructure
for Bio-Data.
The evidence substrate for living systems. Every biological fact — lab result, AI prediction, researcher hypothesis — structured, provenance-tagged, and cross-species queryable.
Evidence Layer · James, 58
Source Fact
Derived
Normalized
Model Output
James, 58 — 847 evidence assertions visualized across body systems
The Atomic Unit
Every biological fact is an assertion. Every assertion knows its origin.
Not a row in a table. Not a PDF attachment. An EvidenceAssertion — typed, timestamped, traceable. It knows what kind of claim it is, where it came from, how confident you should be, and what it connects to.
source_factVERIFIED
EGFR_EXON_19_DELETION
James, 58 — Homo sapiens · TAX_9606
- Mutation Type
- Deletion
- Exon
- 19
- VAF
- 47.3%
- Coverage
- 2,847×
- Source
- Illumina NovaSeq 6000
- Provenance
- Foundation Medicine · Liquid Biopsy
- Collected
- 2024-03-14 · Status: Current
- → 4 derived_features
- → 2 model_outputs
- → 1 hypothesis
The Friction
Biological data is stuck in the dark ages of static PDFs and broken pipelines.
“James's oncologist ordered 14 tests across 3 systems over 6 months. By treatment decision time, no one could reconstruct which EGFR result came from which instrument run — or which interpretation was from the AI classifier and which from the pathologist.”
Infrastructure Gap Detected
76% of preclinical studies cannot be reproduced. The reason isn't fraud — it's missing provenance. OpenBio is the evidence substrate that fixes the bridge.
What if every biological fact carried its full story?
source_fact
HBA1_HUMAN
Hemoglobin Card
- Atomic Mass
- 64,458 Da
- Sequence Integrity
- 99.98%
- Verified Citations
- 1,204
- Source
- Swiss-Prot · UniProtKB/Swiss-Prot
- Entry
- P69905 (HBA_HUMAN)
- Valid since
- 1988-07-01 · Last reviewed 2024-01
Lineage source_fact → 3 derived_features → 12 model_outputs
100%
Traceability across every digital asset generated within the OpenBio mesh.
The Trust Architecture.
Epistemic Boundaries
Standardized and mapped to a shared ontology.
"EGFR mapped to HGNC:3236"
Derived from documents or images via parsing.
"Physician note, NLP-extracted"
Computed from one or more other assertions.
"Progression score from 3 lab values"
AI/ML prediction with attached confidence.
"Recurrence risk: 74% (AlphaFold3)"
Proposed interpretation, under active review.
"EGFR-TKI resistance via T790M"
How Data Moves Through the Levels
Ingestion
Raw sequence data cleaning and metadata normalization.
Produces source_fact · normalized_fact
Verification
Cryptographic hashing of biological records.
Validates source_fact integrity
Consensus
Multi-node validation of experimental results.
Elevates to normalized_fact · extracted_annotation
Immutable
Clinical-grade archiving of final biological truths.
Archives all levels with full lineage
evidence_vectors
Same schema.
Every organism.
From bacteria to primates — every subject is encoded into the same evidence dimensions. Different shapes, identical axes.
James
TAX_9606 · human
847 assert.
Luna
TAX_9615 · canine
234 assert.
M-4872
TAX_9544 · macaque
512 assert.
HeLa-S3
CELL_HELA · cell line
1,247 pass.
E. coli K-12
TAX_83333 · microbe
4,891 feat.
labsgenomictemporalimagingdrugsEvidenceAssertion · shared schema
The Evidence Mesh
One substrate. Every organism.
The same EvidenceAssertion model spans the entire tree of life. Find an EGFR variant in James. Trace it to macaque trial data. Cross-reference E. coli gene expression. One query. Five organisms.
Query: EGFR-related assertions across subjects
James
847
Luna
234
M-4872
512
The same EGFR exon 19 deletion in James links to macaque trial results in M-4872 through the evidence mesh.
Built For
Infrastructure serves everyone who touches biology.
Research Scientists
Query provenance-complete evidence across organisms. Know whether you're looking at a raw measurement or an AI inference. Compare findings cross-species. Never lose the chain of custody from instrument to publication.
847
assertions on a single human subject
AI & ML Engineers
Structured, typed biological data for training and inference pipelines. Every example knows its trust level — preventing source facts from being mixed with model outputs. The data substrate your agents deserve.
6
trust levels, queryable by filter
Biotech Builders
An evidence API that speaks FHIR, OMOP, GA4GH — and adds provenance none of them offer. Build on infrastructure instead of data engineering heroics. The layer above your existing systems, not a replacement.
5
organism types, one data model