Scientific specification

All formulas, indices and models

Full scientific reference for everything Nutramilo uses — formulas, units, reference ranges and peer-reviewed citations (PMID/DOI). For patients, clinicians and journalists asking: 'Is this credible?'.

Nutrition — Insulin response

DIL (Dietary Insulin Load) — Инсулинов товар

Insulin load per food — measures total insulinogenic burden of the meal.

DIL_per_100g = 0.69 · Protein_g + max(0, Carbs_g − Fiber_g)

NIS v1.1.9 (Nutramilo Insulin Score) — Инсулинов рейтинг

Normalised 0–100% insulin rating of the meal. Tiers: Low/Moderate/Medium/High/Very high. Frozen v1.1.9 calibration.

raw_NIS = (1.61·C + 0.66·P + 1.20·F − 1.14·Fi) / Energy_kJ × 1000
NIS = clamp(0..100, 0.7275 · raw_NIS + 8.111)
Energy_kJ = (P·4 + C·4 + F·9) · 4.184

Glycemic Load (GL)

Glycemic Load = GI × available carbs / 100.

GL = (GI_food × (Carbs_g − Fiber_g)) / 100

Personalised Daily DIL Budget — Дневен инсулинов бюджет

Daily Insulin Budget — your personal cap on daily insulin load, adjusted for weight, HOMA-IR, activity.

base_budget = 90 (M) / 75 (F)
HOMA-IR ≥ 2.5 → ×0.8 ; ≥ 5.0 → ×0.65
activity ≥ 8000 steps/day → ×1.15

Nutrition — Quality & density

Calorie Density (Volumetrics)

kcal/g. Target ≤ 1.5 kcal/g for volume without overeating.

calorie_density = kcal / weight_g

Nutrient Density (NRF9.3 / NRF11.3)

Nutrient-Rich Foods Index — beneficial/limit nutrients per 100 kcal.

NRF = Σ (positive_nutrients / DRV × 100) − Σ (limit_nutrients / LIM × 100)

Nutri-Score (A–E)

Official EU FOP label — FR, BE, DE, ES, NL, CH.

Score = Σ negative_points (energy + sat fat + sugars + sodium) − Σ positive_points (fiber + protein + fruit/veg/nuts)

Satiety Index

Satiety 0–100 from protein, fiber, energy density and palatability.

satiety_score = w1·protein_density + w2·fiber_density − w3·energy_density + w4·water%

PRAL (Potential Renal Acid Load)

Alkaline/acid load. Relevant for kidneys, stones, osteoporosis.

PRAL = 0.49·P + 0.037·Ph + 0.027·SO4 − 0.021·K − 0.026·Mg − 0.013·Ca

NOVA classification

1 (unprocessed) → 4 (ultra-processed). WHO/PAHO endorsed.

NOVA ∈ {1, 2, 3, 4} — heuristic on ingredient list

FODMAP level

Fermentable carbohydrate level — low/moderate/high. Key for IBS/SIBO.

lookup → Monash University FODMAP DB

Omega-6 : Omega-3 ratio

Anti-inflammatory balance. Target ≤ 4:1, Western diet 15–20:1.

ω6/ω3 = (LA + AA) / (ALA + EPA + DHA)

Inflammation & oxidative stress

DII (Dietary Inflammatory Index)

Composite index across 45 dietary parameters.

DII = Σ (z_i − global_mean_i) · inflammatory_weight_i

EDIP (Empirical Dietary Inflammatory Pattern)

Empirical inflammatory pattern — correlated to CRP, IL-6, TNF-α.

EDIP = Σ food_group_intake · CRP_correlation_weight

NEAP (Net Endogenous Acid Production)

Estimate of daily net acid production.

NEAP = 54.5 · (Protein / Potassium) − 10.2

UIS / Inflammation Index (Nutramilo)

Composite EDIP + DII + NEAP — pro vs anti-inflammatory signals.

UIS = w1·EDIP + w2·DII + w3·NEAP (normalised 0–100, lower better)

Body & metabolism

BMI (Body Mass Index)

Body mass index. Normal 18.5–24.9 kg/m².

BMI = weight_kg / (height_m)²

WHR (Waist-to-Hip Ratio)

Waist/hip. High risk: M > 0.90, F > 0.85.

WHR = waist_cm / hip_cm

WHtR (Waist-to-Height Ratio)

Waist/height. Cut-off: > 0.50 — elevated metabolic risk.

WHtR = waist_cm / height_cm

HOMA-IR (Insulin Resistance)

Insulin resistance marker. Normal < 2.5.

HOMA-IR = (Fasting_glucose_mmol × Fasting_insulin_µU) / 22.5

BMR (Basal Metabolic Rate)

Basal metabolic rate — Mifflin-St Jeor (validated vs indirect calorimetry).

M: BMR = 10·kg + 6.25·cm − 5·age + 5
F: BMR = 10·kg + 6.25·cm − 5·age − 161

TDEE (Total Daily Energy Expenditure)

Total daily energy expenditure = BMR × activity factor.

TDEE = BMR × {sedentary 1.2 | light 1.375 | moderate 1.55 | active 1.725 | very active 1.9}

VAIS (Visceral Adiposity Index Score)

Visceral adiposity estimate without MRI.

M: VAI = (WC / (39.68 + 1.88·BMI)) · (TG/1.03) · (1.31/HDL)
F: VAI = (WC / (36.58 + 1.89·BMI)) · (TG/0.81) · (1.52/HDL)

eGFR (CKD-EPI 2021)

Renal filtration. CKD-EPI 2021 race-free. Normal ≥ 90.

eGFR = 142 · min(Scr/κ,1)^α · max(Scr/κ,1)^−1.200 · 0.9938^age · (1.012 if F)

Ageing & epigenetics

PhenoAge (Levine 2018)

Biological age from 9 biomarkers. Validated vs all-cause mortality.

xb = Σ (β_i · marker_i) − 19.907 (9 markers: albumin, creat, glucose, CRP, lymph%, MCV, RDW, ALP, WBC, age)
M = 1 − exp(−0.0076927·exp(xb))
PhenoAge = 141.50 + ln(−0.00553·ln(1−M)) / 0.090165

DunedinPACE

Pace of biological ageing. Lower = slower ageing.

DunedinPACE = function of 173 CpG methylation sites (epigenetic clock)

Nutramilo Epigenetic Map (модул)

Interactive Sankey diagram of epigenetic pathways across 340+ diseases (incl. rare syndromes). Disease Discovery mode with live cross-pathway search. Links disease → epigenetic pathways → regulator genes → dietary and lifestyle modulators. Based on KEGG, Open Targets and ClinVar.

disease → {epigenetic_pathways[]} → {regulator_genes[]} → {food_modulators[], lifestyle_modulators[]}
Sankey flow weights derived from gene–disease association scores (Open Targets) + KEGG pathway membership.

Gynaecology & hormones

2-OHE1 / 16α-OHE1 Ratio (estrogen metabolism)

Ratio of protective vs proliferative estrogen metabolite. Normal ≥ 0.7.

ratio_2_16 = 2-OHE1 / 16α-OHE1 (от 24h урина или урина с креатинин)

Methylation Activity (estrogen)

Methylation activity — 2-MeOE1 / 2-OHE1. Normal ≥ 0.15.

methylation_activity = 2-MeOE1 / 2-OHE1

4-OHE1 burden

Genotoxic 4-hydroxy-estrone load. Normal < 4.75 µg/g creatinine.

4OH_burden = 4-OHE1 / creatinine_g (urine)

(Breast Cancer Risk, IBIS)

10-year and lifetime breast cancer risk. Standard for high-risk (≥ 20%).

P(BC | age, family, BMI, menarche, parity, BRCA, biopsy, density)

Breast Density (BI-RADS A–D)

Breast density — imaging diagnostic software (mammography).

BI-RADS classification on mammographic image (A almost-fatty → D extremely dense)

COL3 (Collagen Synthesis Score)

Our own score for type-III collagen synthesis support (Vit C, copper, amino acids).

col3 = w1·vit_C + w2·copper + w3·glycine + w4·proline + w5·lysine (normalised 0–100)

Cardio-metabolic risks

Framingham Risk Score

10-year CVD risk. Primary-prevention standard.

logit(P) = β1·age + β2·SBP + β3·smoking + β4·diabetes + β5·TC + β6·HDL

FINDRISC (Diabetes 10y risk)

Finnish T2D risk calculator. Normal < 7.

score = age_pts + BMI_pts + WC_pts + activity_pts + diet_pts + meds_pts + glucose_pts + family_pts

Dentistry

CAMBRA (Caries Management By Risk Assessment)

Caries Management By Risk Assessment. ADA standard.

risk = Σ disease_indicators − Σ protective_factors + Σ risk_factors

EFP 2018 (Periodontitis classification)

Periodontitis staging (I–IV) and grading (A–C).

stage = max(CAL, RBL, tooth_loss); grade = bone_loss% / age

Dietary patterns & pathways

Mediterranean Diet (MEDI-LITE)

Mediterranean diet. PREDIMED: −30% CVD risk in diabetics.

MEDI-LITE = Σ food_group_serving_score (0–18 scale)

DASH score

DASH. Lowers SBP by 8–14 mmHg.

score = Σ quintile_pts (high: F&V, dairy, whole grains; low: sodium, red meat)

MIND score

MIND diet. −53% Alzheimer's risk.

score = Σ MIND_food_pts (15 components: berries, leafy greens, nuts, fish, olive oil, …)

hPDI (Healthful Plant-based Diet Index)

Healthful plant-based diet. −20% T2D risk.

score = Σ quintile (healthy plant foods) − Σ quintile (unhealthy plant + animal foods)

KEGG Metabolic Pathways

500+ pathways from KEGG — mapping condition → biological pathway → food modulator.

lookup: condition_id → pathway_id (KEGG) → food_modulator (peer-reviewed)

AI Agents (RAG + Multi-Bot)

EndoBot

Endocrine interpretation — insulin, cortisol, thyroid, sex hormones. Vector store: 100+ endocrine handbooks.

OpenAI Responses API + file_search (vector_store_id) → Bulgarian-grounded answer with PMID citations

NutriBot

Nutrigenomics, food interactions, micronutrient analysis.

OpenAI Responses API + file_search (vs_69d6c8f5506c81918637cdfe7c9cf4c0) — Bulgarian, no transliteration

NephroBot

Renal load, electrolytes, CKD safety.

OpenAI Responses API + file_search (nephro vector store)

HemoBot

Anaemias, iron, B12, folate, methylation.

OpenAI Responses API + file_search (haematology vector store)

LactoBot

Breastfeeding, lactation, post-partum nutrition.

OpenAI Responses API + file_search (lactation vector store)

GPT-5.2 (Menu generation)

AI agent for 7-day menu of cooked dishes, built only from user-picked foods.

OpenAI gpt-5.2 via emergentintegrations.LlmChat — async job pattern (start + poll/status)

Imaging (AI screenings)

Mammography / Ultrasound / CT / MRI / X-ray

Imaging modalities (ultrasound, mammography, CT, MRI, X-ray). Nutramilo provides AI-assisted visual reading with mandatory clinician validation.

OpenAI Vision (gpt-4o-mini) → structured findings → clinician review

Skin Screening (Dermatology)

Skin malignancy screening — dermatology AI module. Not a diagnosis.

ABCDE + asymmetry/border/colour/diameter/evolution heuristic on AI vision output

Food Match Score (FMS)

Food Match Score (FMS)

Our own composite 0–100 score that ranks each food vs the selected conditions, phenotypes, diets, indices and nutrients. NOT a patent — our own (peer-review pending) algorithm.

FMS = 100 · (w1·disease_compat + w2·phenotype_compat + w3·diet_fit + w4·indices_match + w5·nutrient_dense)
weights: 0.35 / 0.20 / 0.15 / 0.20 / 0.10  (sum = 1.0)
+ cooking_bonus (±8 for cooked vs raw)
Disclaimer: Nutramilo is not a medical device under MDR 2017/745. The formulas and indices are educational tools and do NOT replace medical consultation. Food Match Score (FMS) and Endo Score are our own (peer-review pending) composite scores — NOT patented.
Nutramilo is a wellness and educational platform — NOT a medical deviceLimitationsDental RiskBreast HealthPoliciesComplaint → CPDPEU↔US transfer

Nutramilo — AI платформа за хранене и метаболизъм