Methodology

Every biomarker Hemix understands lives in a single catalog of 190+ markers. Each one has a stable identifier and one canonical definition — its name, its unit, its category, and its reference range. The catalog is the single source of truth: the app, your reports, and any AI you connect all read the same definitions, so a marker means the same thing everywhere.

Because labs name and format the same marker in many different ways, the catalog also carries a large set of aliases across several languages. That is how a result labelled differently by two labs still lands on the same marker in your record.

You can upload a lab report in almost any format. Before a value is added to your record, it passes a fixed, deterministic set of checks:

• Name matchingThe extracted name is matched to a catalog marker through several tiers, from exact alias to close fuzzy match.
• PlausibilityThe value is checked against the expected range for that marker.
• Unit normalisationThe reported unit is converted to the marker's canonical unit; an unsafe conversion is rejected rather than guessed.
• Reading qualityLow-confidence text extraction is flagged rather than trusted.
• Section contextA value is checked against the part of the report it came from.
• De-duplicationAn exact repeat of a value already in the same test is removed.
• ConfidenceEach value is graded; high-confidence values are added, lower-confidence ones are set aside for review.

Nothing is silently dropped. A value that can't be matched with confidence is held for review, never discarded — so your record stays a complete account of what your reports contained.

Each value is shown next to a general-population reference range. Where the science supports it, that range is sex- and age-aware, so a result is compared against the right population rather than a single generic adult average. These ranges describe the general population — they are not personalised clinical targets.

Every account sees every value and its reference range. Pro accounts additionally see the catalog's narrower optimal ranges — tighter wellness bands that sit inside the reference range. Free or Pro, the underlying value is the same; the difference is only how much context is shown around it.

Related markers are grouped into ten areas — Heart & Vessels, Metabolism & Energy, Liver, Kidney, Thyroid, Hormonal Balance, Blood Health, Immunity & Inflammation, Bone & Muscle, and Nutrients. Each area gets a descriptive 0–100 summary of where it stands based on your most recent results.

A pillar score is a description of the present, not a forecast: it summarises where your current markers sit relative to general-population ranges. The same ranges are used for everyone — any conditions you record inform educational context only, and never change a score.

For Pro accounts, Hemix can show an estimate of biological age, calculated from a standard set of blood markers together with your chronological age. It is presented as an estimate of where a set of markers typically sits, alongside your chronological age for context — not a prediction about the future.

Method · PhenoAge (Levine, 2018)

Your record is more than blood work. Alongside your labs, Hemix holds the supplements you log — name, dose, schedule, and the dates you started or stopped — your journal entries, and, where you connect a device, read-only wearable data such as steps, sleep, and workouts.

These sit together on a shared timeline, so you can see your readings, what you were taking, and what your device recorded side by side. Hemix co-presents them; it does not assert that one caused another. Any connection between them is yours to read — or to ask your own AI to read. Wearable data is shown as your device measured it; Hemix doesn't change or re-score it.

You can read your record in the app, export it, or connect your own AI assistant to it. What is shared in every case is your stored data — values, units, ranges, and dates — your record, not an interpretation of it. See connecting an AI agent.

Any analysis produced from that record — whether Hemix's own AI report or your own assistant reading the data — is observational. It describes what is in your results; it does not diagnose, prescribe, or predict.

Plain-language descriptions of markers and conditions are written for a general audience and draw on patient-education sources such as NHS, Mayo Clinic, Cleveland Clinic, and others. They describe conditions in general terms — they are not guidance about your specific case.