Human genomics · Evidence infrastructure

From genomic data to defensible evidence.

Caelveris is building a modular analysis and interpretation platform that transforms WGS and WES data into traceable evidence, prioritized candidates and structured outputs designed for expert review.

In development. Intended for professional and research workflows. Caelveris does not independently provide a clinical diagnosis and is not presented as a clinically validated medical device.

Evidence
Engine
Variant evidencePopulation frequencyPhenotype contextGene–disease validity
TraceableSource and version provenance
ModularExecutor-independent workflow design
DeterministicEvidence before narrative
Expert-ledReviewable, not autonomous diagnosis
Future workspace

One platform, role-specific views.

The public website is the first layer. The planned authenticated portal will separate laboratory operations, expert review and administration so each role sees the data and controls needed for its responsibility. These panels are a product-direction preview—not a live clinical system.

Operations workspace

Submit, monitor and verify analysis runs.

Planned capabilities include authorized case intake, controlled upload, run state, QC review, artifact inventory and delivery tracking. Large genomic files will use dedicated object storage and signed transfers rather than the application database.

Case intakeValidated metadata and consent context
Run monitoringNextflow status and QC checkpoints
DeliveryVersioned reports and audit trail
Platform direction

Analysis is only the beginning.

The platform is designed as a reusable analysis core rather than a single FASTQ-to-VCF script. It connects reproducible primary analysis with evidence normalization, phenotype-aware reasoning, candidate prioritization, reporting and auditability. Components remain separable so laboratories can adopt the layer they need and retain control of expert interpretation.

01 / ANALYZE

Reproducible WGS and WES processing

Quality control, GRCh38 alignment, alignment metrics, DeepVariant calling, normalization and VEP annotation organized as modular, testable processes.

02 / CONNECT

Evidence with preserved meaning

Variant, population, gene–disease, phenotype and disease-ontology sources are integrated without collapsing distinct evidence types into misleading binary conclusions.

03 / REVIEW

Structured outputs for specialists

Candidate ranking, inheritance context, evidence summaries and provenance are delivered as reviewable structured data before any narrative layer is generated.

Analysis flow

A transparent path from reads to review.

Each stage is expected to emit explicit artifacts, quality metrics and machine-readable provenance. The long-term objective is to make reruns, comparison, verification and expert sign-off practical across local, cloud and HPC execution environments.

Raw data quality controlFASTQ integrity and sequencing metrics
Alignment and BAM qualityGRCh38 alignment with measurable QC
Variant calling and normalizationDeepVariant output prepared for consistent annotation
Functional and clinical evidenceVEP, ClinVar, gnomAD and curated evidence sources
Phenotype and inheritance contextHPO, disease normalization and family-aware logic
Candidate ranking and reportingStructured evidence for qualified expert assessment
Interpretation principles

Meaning is kept explicit.

Caelveris is designed around the idea that each evidence source answers a different question. Absence from one source is not automatically negative evidence, and gene-level evidence is not treated as proof of variant pathogenicity.

Ranking is not classification

Candidate prioritization and ACMG/AMP clinical classification are treated as separate workflows with separate outputs and responsibilities.

Sources are not interchangeable

PanelApp supports panel relevance; GenCC supports gene–disease validity; HPO provides phenotype context; MONDO supports disease normalization; ClinVar and gnomAD answer different variant-level questions.

NOT_FOUND is not REJECTED

Missing evidence is represented explicitly rather than silently converted into a negative conclusion.

Narrative follows structured evidence

Any future NLP layer will explain deterministic outputs. It will not independently decide which finding is clinically significant.

Trust foundation

Built toward accountable operation.

Clinical and genetic data require more than computational accuracy. The product roadmap includes role-based access control, encryption, audit logging, retention and deletion policies, licensing controls, data-processing agreements and intended-use governance before real patient data is accepted.

PROVENANCE

Evidence lineage

Source name, release, retrieval date, license context and transformation history are intended to accompany evidence records and reports.

VALIDATION

Benchmark discipline

Independent technical benchmarking is planned with GIAB HG002 after the first production VCF is frozen, reducing the risk of tuning development against the truth set.

GOVERNANCE

Professional boundaries

The initial commercial direction is B2B analysis and decision support for authorized organizations—not direct-to-consumer autonomous diagnosis.

Questions

Clear boundaries from the start.

The answers below describe the intended direction of the product while it is under development. Final capabilities, validated use and contractual terms will be defined before commercial deployment.

Does Caelveris provide a clinical diagnosis?

No. The platform is intended to support qualified professionals with structured analysis and evidence. It does not independently diagnose disease or replace clinical responsibility.

Is candidate ranking the same as ACMG/AMP classification?

No. Ranking helps experts navigate potentially relevant findings. Clinical classification is a separate controlled process requiring explicit criteria, evidence evaluation and professional sign-off.

Where will genomic files be stored?

The production design will keep large genomic objects outside the transactional SQL database, using encrypted object storage, controlled transfers, tenant separation, retention policies and audited access.

Can laboratories request a pilot?

Early conversations are welcome. Pilot work will be limited to clearly defined scope, data governance, technical validation and intended-use boundaries.

Early collaboration

Building for laboratories and research teams.

Caelveris is seeking conversations with laboratories, healthcare professionals, bioinformatics teams and research organizations interested in modular genomics analysis, evidence integration or pilot validation.