The premise
For most of history, you learned your risks by watching your relatives age. Now you can read them directly: a saliva kit sequences hundreds of thousands of your genetic markers, a blood draw measures the lipids and sugars that actually injure arteries, and an AI assistant can hold all of it at once and tell you where you stand.
That changes what a diet is for. Everyone benefits from the Earth Plate; the evidence for the pattern is universal. But if your genome and your family history say Alzheimer's, fish and the MIND pattern move to the center of the plan. If they say type 2 diabetes, fiber and the overnight fast do. If they say heart disease at 9p21, the plant-rich pattern that has been shown to blunt that very variant becomes non-negotiable, and a calcium scan at 45 becomes a priority instead of an option. Knowing your body is how a general diet becomes a personal one.
Labs first, because they measure the damage actually happening. Genes second, because they tell you which damage to watch for. AI third, to synthesize both into a plan. A clinician throughout, because none of the first three can examine you.
Step 1: the labs that matter
Genes are probabilities; labs are measurements. Most of the panel below is inexpensive, and several items are missing from a routine physical unless you ask. Targets shown are commonly cited optimal ranges for prevention; your clinician sets yours.
| Test | What it tells you | Commonly cited target | How often |
|---|---|---|---|
| ApoB | The count of atherogenic particles; a better predictor of heart disease than LDL cholesterol | Under 90 mg/dL; under 80 if other risks | Yearly |
| Standard lipid panel | LDL, HDL, triglycerides, total cholesterol | LDL under 100; triglycerides under 150 (under 100 better); HDL over 40 (men) or 50 (women) | Yearly |
| Lipoprotein(a) | A largely inherited particle that raises heart and valve risk independent of diet | Under 125 nmol/L (about 50 mg/dL) | Once in a lifetime |
| HbA1c and fasting glucose | Three-month average and fasting blood sugar | HbA1c under 5.7 percent; glucose under 100 mg/dL | Yearly |
| Fasting insulin | The earliest signal of insulin resistance, years before glucose rises | Low single digits in µIU/mL is reassuring | Yearly if diabetes risk |
| hs-CRP | Low-grade vascular inflammation | Under 1 mg/L low risk | Yearly |
| Blood pressure | The most modifiable cardiovascular risk there is | Under 120/80 at home, averaged over two weeks | Monthly at home |
| Waist-to-height ratio | Visceral fat, which drives metabolic risk more than weight does | Waist less than half your height | Monthly, a tape measure |
| Vitamin D (25-OH) | Bone, muscle, and immune function; low in most indoor adults | 30 to 50 ng/mL | Yearly |
| B12, ferritin, TSH, CBC, CMP | Nerve and blood health, iron stores, thyroid, kidney and liver function | Within range; B12 above the low end if plant-based | Yearly |
| Coronary artery calcium (CAC) scan | A low-dose CT that counts actual plaque; the single most informative heart test for a 40-to-60-year-old with any risk | A score of zero | Once around 40 to 50, repeat per result |
| Colorectal screening | Colonoscopy or stool DNA test | Starting at 45, earlier with family history | Per result |
| Grip strength and VO2 max | Two functional measures that predict mortality better than most blood tests | Rising or holding year over year | Yearly |
Step 2: your raw DNA
If you have taken a 23andMe or AncestryDNA test, you already own a file with roughly 600,000 to 700,000 of your genetic markers. The company's reports show you a sliver of it; the raw file holds the rest.
Download the raw data
Sign in, open Settings, find the 23andMe Data section, and request the raw data download; the file arrives by email as a zipped text file. The health reports (including APOE status, if you opted in) are separate and worth saving as PDFs. Keep your own copy: companies change hands, and your genome should not depend on anyone's servers.
Download the raw data
Sign in, open the DNA settings for your test, and choose the raw DNA download; a confirmation link arrives by email. Ancestry does not report health findings, but the raw file contains many of the same markers, including the APOE variants, which a third-party tool or an AI assistant can read.
What a consumer chip can and cannot see. It genotypes a fixed set of common variants; it does not sequence your genome. It reliably reads well-known single-letter variants (APOE ε4, the 9p21 locus, the lactase-persistence variant, caffeine metabolism, many pharmacogenomic markers) and covers enough of the genome to build approximate polygenic scores for common diseases. It cannot reliably call rare variants such as most BRCA mutations, and a negative result for a rare variant is not reassurance. For that, clinical sequencing ordered through a genetic counselor is the only honest answer.
Step 3: polygenic risk, explained
Most common diseases are not caused by one gene. Heart disease, type 2 diabetes, Alzheimer's, breast and colorectal cancer, and high blood pressure are shaped by thousands of variants, each nudging risk a little. A polygenic risk score (PRS) adds those nudges up and places you on a percentile against a reference population: the 50th percentile is average; the 95th is the top twentieth of risk.
- A PRS is relative, not absolute. Being at the 95th percentile for type 2 diabetes might mean a lifetime risk two to three times the average, not a diagnosis. Ask for the absolute number, and remember that lifestyle moves it.
- Ancestry matters. Most scores were built from people of European ancestry and are less accurate for others. Newer, multi-ancestry scores are better; read which one you are given.
- Consumer scores are approximations. A score built from a few dozen lead variants is a sketch; a clinical-grade score uses millions of variants and validated reference ranges. Several clinical laboratories now offer the latter for heart disease and breast cancer through a clinician.
- Where to look. The PGS Catalog (an open registry run with the European Bioinformatics Institute) lists published scores by disease. Third-party interpretation services (Promethease, Genetic Lifehacks, SelfDecode, and others) will read a raw file for a fee; treat their output as a starting point for questions, not a verdict.
- Family history is the oldest PRS. A parent with early heart disease, a grandparent with dementia, a sibling with diabetes: these carry as much predictive weight as most scores and cost nothing to collect. Write it down for three generations.
Step 4: AI as the synthesizer
This is the part that did not exist five years ago. An AI assistant that can run code will read a raw genotype file, pull the variants that matter, compute approximate scores against published weights, lay them beside your lab history and your family history, and write it up in plain language. What used to take a specialist weeks takes an afternoon.
What to give it
- Your raw DNA file (zipped text), with your name removed from the filename.
- Your lab results across as many years as you have; the trend matters more than any single draw.
- A three-generation family history: who had what, and at what age.
- Your blood pressure log, your weight and waist history, and anything from a wearable (resting heart rate, sleep, VO2 max estimates).
- Your current diet, movement, sleep, and supplement habits, honestly described.
What to ask for
- A genetic baseline sorted into favorable, watch, and informational, with the evidence tier for each item.
- Approximate polygenic risk percentiles for coronary disease, type 2 diabetes, Alzheimer's, atrial fibrillation, and the common cancers, with the caveats stated.
- Pharmacogenomic flags worth telling a prescriber about (how you metabolize caffeine, alcohol, statins, common pain relievers, and antidepressants).
- A reconciliation: where the genome and the labs agree, and where they disagree. A near-average cardiovascular score with a high LDL means the lipids are largely lifestyle-driven, which is good news, because lifestyle is yours to change.
- A prioritized action list: which labs to order next, which screenings are due, and which Earth Plate levers address your specific risks.
- A list of what is not a priority, so a sketchy score does not drive panic.
What it cannot do
It cannot examine you, order a test, or know what it was not told. Approximate scores can be wrong in both directions; one version of a self-assembled profile can overstate a risk that a fuller analysis later corrects. Treat the output as a well-organized briefing to carry into a preventive cardiology or primary care visit, not as a diagnosis. The visit is where the CAC scan, the Lp(a) draw, and the statin conversation actually happen.
Step 5: from risk to plate
This is where the two halves of the site meet. Each row names a risk the labs or the genome can flag, and the Earth Plate levers with the best evidence against it.
| If your risk points to | The Earth Plate levers | Beyond the plate | Evidence |
|---|---|---|---|
| Alzheimer's (APOE ε4, family history) | Fatty fish at least twice a week, leafy greens daily, berries daily, olive oil, the MIND pattern; minimal added sugar and processed meat | Sleep seven to nine hours, aerobic fitness, blood pressure control, hearing protection, social and cognitive engagement | Seafood tied to less Alzheimer's pathology in ε4 carriers; MIND pattern tied to slower decline COHORT |
| Coronary disease (9p21, ApoB, Lp(a), family history) | The PREDIMED plate: olive oil, nuts, legumes, fish, vegetables and fruit; soluble fiber from oats and barley; very low added sugar | CAC scan around 45; ApoB-driven statin or other therapy if indicated; zone 2 and strength; no smoking | A prudent plant-rich diet attenuated the 9p21 effect in INTERHEART; PREDIMED cut events about 30 percent COHORTRCT |
| Type 2 diabetes (high PRS, HbA1c creeping, family history) | Thirty-plus grams of fiber daily, legumes most days, intact grains over flour, the Daily Bowl, no sugar-sweetened drinks, nuts and olive oil as fats | Waist below half of height, strength training twice weekly, a walk after meals, the overnight fast, sleep | Fiber and whole grains lower incident diabetes in cohorts; lifestyle beat metformin in the Diabetes Prevention Program COHORTRCT |
| Colorectal cancer (family history, PRS) | Fiber from whole grains and legumes, crucifers, no processed meat, little red meat, little alcohol, calcium from yogurt and sardines | Screening at 45 or earlier, movement, no smoking | World Cancer Research Fund grades fiber and whole grains protective, processed meat and alcohol causal CONSENSUS |
| Breast cancer (family history, PRS) | Mostly plants, fiber, soy foods (edamame, tempeh) are safe and may be protective, minimal alcohol | Weight management, strength and aerobic exercise, screening per family history | Alcohol is a Group 1 carcinogen; body fat after menopause and inactivity raise risk CONSENSUS |
| Hypertension (family history, salt sensitivity) | The DASH overlap: potassium-rich produce, beans, leafy greens, bananas; beets and arugula for nitrate; garlic; less sodium from packaged food | Home blood pressure cuff, movement, sleep, limited alcohol | DASH lowered blood pressure in trials; dietary nitrate and garlic in smaller trials RCT |
| Macular degeneration (AMD variants) | Lutein and zeaxanthin daily: spinach, kale, eggs, pistachios, orange peppers, corn; fatty fish | Sunglasses, no smoking, regular eye exams | AREDS2 formulation slowed progression; dietary lutein tracks with lower risk RCTCOHORT |
| Osteoporosis (family history, low bone density) | Calcium from yogurt, kefir, sardines, tahini; prunes daily; vitamin D and K from greens; adequate protein | Strength training and impact, vitamin D level checked, no smoking | Prunes supported bone density in trials; resistance training is the strongest bone signal RCT |
| Lactose intolerance (lactase variant) | Kefir and aged cheeses are often tolerated; otherwise cultured plant alternatives, tahini, sardines, and greens for calcium | Lactase persistence is one of the clearest single-gene traits a chip reads CONSENSUS | |
| Fast or slow caffeine metabolism (CYP1A2) | Slow metabolizers: cap coffee earlier and lower; everyone: last cup by early afternoon | Fixed sleep window | CYP1A2 variants modify caffeine's cardiovascular association in cohorts COHORT |
| Efficient alcohol metabolism (no ALDH2 brake) | Treat the absence of a flush as permission to drift, and set a hard cap in advance | None within three hours of bed | No safe level has been shown; the brake's absence removes a warning, not the risk CONSENSUS |
The honest word on genes
Twin and family studies put the heritability of human lifespan at roughly 20 to 30 percent, and more careful analyses that account for people marrying people like themselves suggest it may be lower. COHORT The rest is environment, behavior, and chance. APOE ε4 raises Alzheimer's risk but is not destiny: a large share of carriers never develop dementia, and the lifestyle levers above are the ones that separate them. The 9p21 heart-disease variant is the most replicated in the genome, and a plant-rich diet was shown to blunt it. A 99th-percentile diabetes score describes the hand you were dealt and says nothing about how you play it.
So the Earth Diet reads a genome the way a climber reads a weather forecast: to know where the exposure is, not to predict the outcome. The exposures tell you which chapters of this site to read twice. The plate, the sleep, the movement, and the screenings are the same for everyone; the genome tells you which to treat as sacred.
Privacy and care
- Your genome is also your relatives' genome. Uploading it exposes information about parents, children, and siblings who did not consent. Strip names, do not upload other people's files, and think about who else is inferable from yours.
- Read the data policy of any service, AI included, before uploading: whether it trains on your data, how long it keeps files, and whether you can delete them. Prefer services that let you delete, and delete when you are done.
- Keep an offline copy of your raw data and your reports. Consumer genomics companies change ownership; your file should not.
- Findings can land hard. APOE ε4 status, in particular, is information some people would rather not have. Decide in advance whether you want to see it, and consider a genetic counselor if a result is heavy.
- Nothing here is a diagnosis. Every finding on this page is a reason to have a better conversation with a clinician, who can order the tests that confirm or dismiss it.