Melatonin rises at night; cortisol, tryptophan, fat, iron, and other components show daily patterns.
AN UPDATED HUMAN-MILK AUDIT
Mother's
Best.
Study milk as a living signal—not one average recipe—then learn whether formula can respond to the hour, the illness, the heat, and the stage of infancy.
A PUBLIC RESEARCH INITIATIVE · NOT A PRODUCT · NOT A FEEDING INSTRUCTION
A baby should not lose the biology because a mother cannot supply the milk
Human milk is the reference food for infancy. Yet lactation can be limited by delivery complications, anatomy, endocrine conditions, medication, illness, separation, ineffective milk removal, or no wish to breastfeed. Breast size is not a measure of devotion or a dependable measure of supply. A mother's health can influence some milk components, but low supply is not a moral verdict on her body.
Modern formula is not empty powder. In the United States, the FDA requires minimum amounts of 30 nutrients, maximums for ten, manufacturing controls, nutrient testing, and evidence that a new formula supports normal physical growth. That regulatory architecture—not merely fear of lawsuits—is why standard formula is generally nutritionally adequate for healthy term infants when prepared exactly as labeled.
But “adequate average nutrition” and “the full behavior of human milk” are different achievements. U.S. rules specify folate, for example, but do not make L-5-methyltetrahydrofolate the universal required source. DHA and ARA research has produced mixed long-term cognitive results, and the U.S. nutrient table does not settle every bioactive molecule, ratio, molecular form, or time signal. Formula solves the floor. A modern milk atlas can map—and responsibly test—the moving ceiling.
Milk moves on more than one clock
A single pooled sample erases biologically relevant variation.
HUMAN
MILK
MILK
Volume and composition vary within and between feeds; collection protocol can change the result.
Colostrum, transitional milk, and mature milk are different biological states, not one recipe aging in place.
Genetics, secretor status, diet, geography, health, gestational age, and individual biology contribute variation.
A systematic review of 83 reports found circadian variation in melatonin, cortisol and cortisone, tryptophan, fats, cholesterol, and iron, although methods and study quality varied. A 1993 study found night milk melatonin in all sampled mothers and none detectable by day. Modern multi-omics work finds that variation between mothers can exceed variation within one mother, especially for peptides. These are reasons to build a proper atlas—not proof that three finished profiles already exist.
Italianer et al. · circadian systematic review · Illnerová et al. · milk melatonin rhythm · Zhang et al. · proteome, peptidome & metabolome variation
Three principled profiles, used by the same infant
These are working research names. They describe compositional hypotheses to test—not products to buy, diagnoses to make, or permission to switch a baby's feeding without clinical evidence.
Sick Baby
Map milk produced when the infant or mother has a confirmed infection, then test whether reproducible immune-active patterns are beneficial for defined groups: acutely ill, preterm, immunologically vulnerable, or recovering infants. Candidate measurements include leukocytes, secretory IgA, lactoferrin, lysozyme, cytokines, HMOs, antimicrobial peptides, extracellular vesicles, and pathogen-specific antibodies.
The evidence is suggestive, not a recipe. In a cohort of 21 dyads, milk leukocytes rose during maternal or infant infection and returned toward baseline after recovery; a later case-control study found higher milk T-cell proportions when infants had respiratory infections. Neither study establishes which factor to add, its dose, whether it survives processing and digestion, or whether adding it improves outcomes.
Hassiotou et al. · infection and milk leukocytes · Baban et al. · infant respiratory infection
PowerBaby
The standard profile should be built from the most consistently favorable patterns of early rearing while respecting how fast those patterns change from colostrum to mature milk. It should audit molecular forms and ratios—not only totals—including folates, DHA and ARA, amino acids, minerals, choline, HMOs, milk-fat-globule-membrane components, peptides, and cofactors. Morning milk belongs in the time-series, but “morning” should not be declared optimal until infant outcomes show it.
Nighty-Night
Reconstruct the range and relationships found in milk expressed at night: melatonin, tryptophan availability, cortisol, nucleotides, fat, and other signals. The objective is circadian information—not sedation. A survey associated mistimed expressed milk with later sleep onset, and small nutrition trials suggest tryptophan-linked composition can affect sleep measures, but neither supplies a safe melatonin dose for healthy infants.
Any study must separate better circadian entrainment from longer sleep, because longer is not automatically safer or developmentally better. It must measure arousal, feeding adequacy, apnea, sleep architecture, next-day behavior, growth, and adverse events. Endogenous night-milk concentration ranges are a starting observation, not a dosing instruction.
Hahn-Holbrook et al. · mistimed expressed milk · Cubero et al. · chrononutrition trial
morning / ordinary growth→one infant, one longitudinal record→night / circadian timeillness profile only under a validated trigger and protocol
Mothers noticed the output changing before the mechanism was settled
Parents repeatedly describe pumped milk turning yellower, bluer, thicker, or otherwise visibly different around an infant's cold, recovery, or weaning. One mother photographed a “sick baby” sample beside her usual milk; another described a freezer gradient across lactation. Those reports are observations worth capturing prospectively—not clinical tests. Color can also change with fat content, diet, blood, medication, storage, and stage of lactation.
The strongest current evidence is narrower: milk immune-cell composition can shift during infection. How the mammary gland receives the infant-specific signal remains unresolved. Shared maternal exposure and the mother's own immune response are obvious routes. Retrograde movement during nursing and infant oral microbes entering ducts are biologically plausible routes. “Baby saliva tells the breast exactly what antibody to make” is a compelling story, but the complete causal chain has not been demonstrated.
A prospective registry can preserve these observations without treating them as conclusions: record the story, timestamp the sample, confirm the diagnosis, assay the milk, follow the infant, and let thousands of paired observations reveal which anecdotes were early signals.
Color, cream layer, viscosity, infant behavior, feeding pattern, illness onset, recovery.
Leukocyte shifts, immune proteins, circadian hormones, longitudinal proteome and peptidome changes.
Signal route, causal ingredient, useful dose, processing survival, clinical benefit, who should receive it.
A mother's paired illness/recovery observation · A freezer gradient across lactation · Review of milk-microbiota source hypotheses
The atlas must be larger than the average
Recruit across ancestries, geographies, diets, income levels, ages, delivery modes, gestational ages, health states, and lactation stages. “Different gene pools” becomes a study that is genetically and geographically broad without treating ancestry as destiny.
Dense sampling
Collect standardized samples across a full 24 hours, within feeds, weekly through early lactation, monthly afterward, and before, during, and after documented illness. Record pumping, freezing, medication, diet, sleep, environment, and infant intake.
Every molecular scale
Measure macronutrients, vitamins and minerals, lipidome, metabolome, proteome, peptidome, glycome and HMOs, hormones, cytokines, immune cells, microbiome, RNAs, extracellular vesicles, contaminants, and how processing changes bioactivity.
Paired outcomes
Link composition to growth, hydration, renal and metabolic markers, infection severity and duration, vaccine response, allergy, stool and microbiome, feeding tolerance, sleep architecture, vision, and neurodevelopment—not just laboratory resemblance.
Constants before profiles
Find what remains within a narrow band, what follows clocks, what tracks lactation, what varies by mother, and what predicts a better infant outcome. A profile earns its name only after it reproduces a beneficial system, not a photogenic milk sample.
Why new technology changes the audit:
A two-donor longitudinal study quantified more than 1,300 proteins and 2,000 endogenous peptides. A 29-mother study found median between-mother variation of 42.8% for proteins and 85.2% for peptides. A 300-dyad network study shows that cohort-scale protein and peptide systems are now tractable. The tools can see a biological vocabulary that old nutrient tables could not.
Personalized longitudinal profiling · Multi-omics variation · 300-dyad network analysis
The last comprehensive U.S. nutrient review was published in 1998
In 1996, FDA commissioned the Life Sciences Research Office to review infant nutritional needs, invited scientific submissions, and held an open meeting. The resulting 1998 report, Assessment of Nutrient Requirements for Infant Formulas, synthesized the available evidence into recommended nutrient levels for formula. In 2025, FDA identified it as the agency's “last comprehensive review” and requested new evidence on whether the requirements in 21 CFR 107.100 should change.
The regulatory foundation also includes the Infant Formula Act of 1980, strengthened controls enacted in 1986, later manufacturing rules, a 2004 National Academies framework for new ingredients, and the 2007 Codex revision. A new audit can extend that foundation into biology that static nutrient totals do not capture: circadian patterns, molecular forms, peptides, multi-omics, dense longitudinal sampling, and outcomes linked to timing and health state.
Chloride-deficient soy formula harms infants; Congress passes the Infant Formula Act.
Congress strengthens quality control, testing, recordkeeping, and recall provisions.
A ten-mother study demonstrates a pronounced day–night melatonin rhythm in human milk.
FDA solicits public scientific input; LSRO publishes the comprehensive nutrient-requirements review.
The National Academies evaluates new ingredients; Codex substantially revises its 1981 standard.
FDA says the last comprehensive review was in 1998 and requests information for a new assessment.
Mass spectrometry and multi-omics make a time-resolved public atlas possible; a recent review says U.S. and Canadian milk-composition data remain critically lacking.
FDA · 1996 public review notice · LSRO · 1998 comprehensive review · FDA · 2025 request for information · National Academies · 2004 report · FAO/WHO · Codex revision history · North American data gap
Fund the questions in the right order
No infant should become the test bench for an attractive theory. The initiative advances only when the preceding evidence gate closes.
- 0
Listen and register
Build a maternal observation registry with sample timestamps, clinical confirmation, photos, storage history, and consent. Turn anecdotes into testable paired data.
- 1
Map
Create the international, longitudinal milk atlas and publish de-identified methods, reference ranges, and negative findings.
- 2
Find causal candidates
Use cell, organoid, digestion, and animal models where appropriate to test stability, absorption, interactions, and plausible mechanisms.
- 3
Test one variable at a time
Conduct regulator-reviewed dose-ranging and safety studies before combining candidates. Independent monitoring and preregistered stopping rules are mandatory.
- 4
Compare profiles
Only then test time-, illness-, or climate-matched profiles against a nutritionally complete control using clinically meaningful infant outcomes and long follow-up.
No withholding human milk. Support breastfeeding and donor milk wherever available; study participation never requires replacing them.
No profile by appearance. Milk color, fussiness, heat, or a parent's hunch cannot activate an immune or night formula.
No ingredient because it sounds natural. Molecular identity, dose, matrix, timing, processing, and interactions all matter.
No commercial shortcut. Public protocols, independent replication, regulatory review, and conflict-of-interest disclosure come before branding.
Research shelf
What supports the proposal, what complicates it, and what remains unproven.
- Circadian variation · systematic review83 reports and 71 components; real rhythms, high methodological heterogeneity.
- Infection · milk leukocytesImmune-cell proportions rose during maternal or infant infection in a small longitudinal cohort.
- Multi-omics · natural variationProteins, peptides, and metabolites vary on different scales across mothers.
- Maternal diet · updated reviewMilk fatty acids track diet more clearly than many other components; methods and results remain diverse.
- Formula additions · cognition reviewDHA/ARA and other additions show mixed or short-lived cognitive findings, underscoring outcome design.
- Formula · regulatory floorCurrent U.S. requirements for nutrients, quality factors, manufacturing, testing, records, and recalls.
THE INITIATIVE
Measure what a mother changes before deciding what a formula should change
Mother's Best is a grant agenda for nutrition scientists, lactation researchers, pediatricians, analytical chemists, chronobiologists, immunologists, statisticians, ethicists, regulators, and mothers. Its public output is a modern milk atlas, validated reference profiles, and a list of ideas that failed safely in the laboratory. Better formula may follow. The evidence comes first.