Skip to main content

Structurally different, not cosmetically novel

Each of these is a property of how the platform is built. None could be bolted onto a set of disconnected tools after the fact. We describe them as structurally different rather than first, only, or world-leading.

One evidential chain, end to end

Lumen feeds Praxis, which feeds Veraxa, Nexora, Novora, and Oris in sequence, with Labtrack closing the loop back to Praxis. Records persist between every step, so a finished report traces back through execution, planning, validation, and design to the original literature consensus.

Why it’s structural: the persistence and ordering are the architecture itself, decided at the start. That's what makes the chain traceable end to end.

Methodology consensus extraction

Lumen parses full-text Methods sections from the published literature to aggregate concrete parameter values, sequencing depth, fold-change thresholds, correction methods, sample sizes, each with a citation count and strength indicator. Methods used in under 10% of the corpus are flagged as outliers.

Why it’s structural: retracted papers are flagged, not silently counted, and zero-coverage fields are reported honestly rather than filled with a fabricated value.

Full-text Methods section · PubMed Central

Libraries were sequenced to a depth of approximately 30 million reads per sample. Differential expression was assessed with DESeq2; genes with absolute log2 fold-change ≥ 1.5 and Benjamini–Hochberg adjusted p < 0.05 were considered significant. Each condition comprised six biological replicates.

Extracted consensus
Methodology parameters aggregated across the corpus with citation counts and evidence-strength indicators.
ParameterConsensus valueEvidence
Sequencing depth30M reads / sample28 citations· Strong
Fold-change threshold|log2FC| ≥ 1.519 citations· Moderate
Multiple-testing correctionBenjamini–Hochberg FDR 0.0534 citations· Strong
Sample sizen = 6 per group12 citations· Moderate
Batch-effect modelNo literature evidence availableNo literature evidence

The boundary is built in, not bolted on

The decision-support boundary is built into the system itself. Every output is marked research-only, and the platform offers no auto-execute action for any clinical, policy, or institutional decision.

Why it’s structural: the boundary is part of how the platform is built, not a dialog laid on top. The platform suggests; the human decides.

Grounded in live literature

Every output is grounded in live-retrieved literature, not the model's own memory. Each claim traces back to a real retrieved document, nothing invented.

Why it’s structural: when there is no evidence, the platform says so. It does not invent a citation or a value to fill the gap.

Reproducibility

Oris, Novora, Omnix, and Sentinel produce reproducibility certificates that record the methodology, parameters, software versions, and seeds behind every result. Exportable as PDF, DOCX, or JSON.

Why it’s structural: Reproducibility is a user-facing deliverable with a persistent identifier, not an internal log. The certificate is something a reviewer can independently act on.

Wet lab and dry lab, equal

Wet-lab outcomes are first-class evidence. Labtrack captures structured results and feeds them to Praxis, which weighs them alongside the published consensus when proposing the next design.

Why it’s structural: Lab-derived signals are attributed distinctly from literature, so the researcher always knows whether a recommendation rests on their bench or on the published record.

Design-Build-Test-Learn loop. Praxis produces a design; Labtrack builds the protocol, records the experiment, and captures the outcome, yield, purity, and success or failure with reason, which feeds back into the next Praxis design.DesignPraxisBuildLabtrack protocolTestExperimental recordLearnOutcome capturedoutcome signal, yield, purity, success / failure with reasonWet lab and dry lab, equal class

Surveillance data, unified

Prognos ingests surveillance data that today sits in incompatible formats, WHO GLASS, ECDC EARS-Net, AWaRe, BPPL, and national CDCs and harmonises it into one analytical surface, with source attribution and data-freshness preserved.

Why it’s structural: the harmonisation is transparent and annotated, so a public-health analyst can see exactly how two sources were reconciled rather than trusting an opaque merge.

Fragmented sources

  • WHO GLASS90+ countries
  • ECDC EARS-Net30 EU/EEA
  • US CDCnational
  • UKHSAnational
  • Pakistan NIHnational
  • ICMR Indianational

Harmonisation layer

Transparent, annotated reconciliation, not an opaque merge.

  • Antibiotic-name normalisation
  • Specimen-category mapping
  • Breakpoint-standard tracking
  • WHO AWaRe classification

Unified surface

One analytical surface

Source attribution and a data-freshness indicator preserved on every record.

resistance %specimenAWaRebreakpointlast ingested
Surveillance data liberation: WHO GLASS, ECDC EARS-Net, US CDC, UKHSA, Pakistan NIH, and ICMR India flow into a harmonisation layer that normalises antibiotic names, maps specimen categories, tracks breakpoint standards, and applies the WHO AWaRe classification, producing a single analytical surface with source attribution and data-freshness indicators.

Universally accessible

Transparent pricing, published quotas, and no opaque per-use charges, usage dashboards show where you stand in real time. Regional pricing reaches Tier 2 markets as a matter of access, not charity.

Why it’s structural: there are no credits, tokens, or consumed units anywhere in the billing model, and no region is privileged over another.

See the innovations at work in the platform.

How it worksThe ten capabilities