Skip to content
VAMS ATLAS™ · confidence

Confidence you can read at a glance.

Every ATLAS™ output carries a confidence cue — so you can calibrate interpretation and see where limits apply. Trust, governance, and scientific restraint, built into the product.

See a Sample Report How ATLAS Works
Confidence cue on every outputPlain-language levelsVisible caveatsMethod-aware limitsEducational & non-diagnostic
Why confidence matters

Why every output carries a confidence cue

Users need more than labels — they need context. A confidence cue tells you how much weight to give a reading, so interpretation stays proportionate and traceable rather than absolute.

  • ✓Confidence reflects how firmly a pattern can be read
  • ✓It keeps interpretation proportionate
  • ✓It makes the reading experience more credible
  • ✓Stronger visibility never means greater certainty

Input 1 — Evidence quality

Evidence quality

Draws on published reference context
Framed as educational, non-diagnostic information
Stronger evidence supports more confident reading
Evidence posture is shown, not assumed

Input 2 — Method compatibility

Method compatibility

What a sequencing method can support shapes how confidently an output can be read.

16S rRNA amplicon Community-level composition; limited species/strain resolution.
Shotgun metagenomics Species-level and functional-potential context.
Metatranscriptomics Expression context where RNA quality supports it.

Input 3 — Feature availability

Feature availability

Some signals require sufficient depth or coverage
Availability reflects ecosystem, sequencing tier and report type
Where a feature is not supported, it is not shown
What you see reflects what your selection can support

Input 4 — Cohort relevance

Cohort relevance

Interpretation is contextualised against curated reference cohorts
Reference context supports comparative reading
Cohort framing is described transparently
See the Reference Cohorts page for detail

Input 5 — Calibration readiness

Calibration readiness

Confidence reflects how well-established a reading is
Where calibration is still maturing, confidence is shown lower
No fixed performance figures are stated before they are verified
Quantitative calibration is established during laboratory validation

Confidence states

Three plain-language levels

Confidence is communicated in plain language so you know how strongly to read each insight — without exposing internal weighting.

Higher confidence — Well-supported by data quality, evidence and reference context — read with more weight.
Moderate confidence — Reasonable support with visible caveats — read as indicative, not definitive.
Lower / emerging confidence — Limited support or maturing calibration — read as exploratory context only.

Communication of uncertainty

How uncertainty is shown

A confidence cue sits alongside every output Explicit caveats accompany relative patterns Comparative framing, not deterministic verdicts Stronger visibility never implies greater certainty Visible limits where they apply

Claim-safety gates

The principles that guide public ATLAS language

Scientific governance means building restraint and clarity into the product itself, not just having policies.

Educational, non-diagnostic language only No diagnosis, disease prediction or treatment claims No exposure of internal weighting, thresholds or computation No unverified performance or validation figures Reference context framed as context, not clinical verdict

Limitations

Where confidence framing stops

Confidence is not a clinical performance metric It does not turn a pattern into a diagnosis It depends on sample quality, method and reference context It is a reading aid, not a guarantee of outcome

FAQ

Common questions about confidence

What does a confidence cue mean?

It tells you how much weight to give a reading, reflecting data quality, evidence strength, reference context and the limits of your sequencing method.

Does higher confidence mean it is diagnostic?

No. Confidence never turns a pattern into a diagnosis. All outputs are educational and non-diagnostic; stronger visibility never implies greater certainty.

Why is my confidence lower for some outputs?

Confidence reflects evidence quality, method compatibility, feature availability, cohort relevance and calibration readiness — any of these can lower it.

Do you publish exact confidence percentages?

No. Confidence is shown in plain-language levels. We do not publish invented performance figures; quantitative calibration is established during laboratory validation.

Does the confidence framework expose your scoring method?

No. It explains how to read confidence without exposing internal weighting, thresholds or computation, which remain proprietary.

See confidence cues in a report

The sample report shows how confidence cues sit alongside every output — with visible caveats and non-diagnostic framing.

See a Sample Report How ATLAS Works