Two people eat the same meal. One person's blood glucose spikes sharply and takes two hours to come back down. The other person's barely moves. Same food, same quantities, same time of day. Two completely different metabolic responses.
This is not a marginal difference. In a study published in Cell in 2015, Segal and Elinav at the Weizmann Institute tracked 800 people over a week, measuring continuous glucose responses to identical meals. The variation between individuals was enormous — and it was predictable. Not from the food alone, but from the combination of the food and the person's gut microbiome.
That study cracked something open. It became one of the first pieces of serious evidence that the universal dietary guidelines — eat less fat, eat more fibre, keep carbohydrates moderate — may be less universal than we assumed.
What personalised nutrition actually means
The term gets used loosely. When a supplement brand says "personalised nutrition," they usually mean a questionnaire that places you into one of four categories and then sells you a product. That is not personalised nutrition. That is segmented marketing.
Real personalised nutrition, as it is developing in preventive genomics and nutritional science, is something more specific. It involves:
- Your genetic variants — SNPs (single nucleotide polymorphisms) that affect how you process certain nutrients
- Your gut microbiome composition — which organisms are present and in what ratio
- Your metabolic markers — fasting glucose, insulin sensitivity, lipid panels
- Your lifestyle inputs — sleep, activity, stress, meal timing
The ambition is to combine these into something that actually predicts, for you specifically, how different foods will affect your body. Not average outcomes across a population. Your outcomes.
MTHFR variants — affect folate metabolism and methylation. Relevant for B vitamin intake, neural tube risk, and potentially mood regulation.
LCT gene — determines lactase persistence. Explains why some adults digest milk without issue and others do not. Well-established and clinically useful.
APOE variants — associated with different responses to dietary fat and cardiovascular risk. APOE4 carriers may benefit more from reduced saturated fat intake.
FTO gene — associated with obesity risk and appetite signalling. Not deterministic, but informs context.
The honest limits
Here is what the science does not yet support, despite what some companies will tell you: it cannot currently tell you with clinical certainty what specific diet you should follow based on your genome alone.
The genome is one input. Most nutrition-related traits are polygenic — they are influenced by dozens or hundreds of genetic variants, each contributing a small effect. Our ability to translate polygenic risk scores into reliable individual dietary recommendations is still early. The research is promising. It is not yet a prescription.
The gut microbiome is perhaps more immediately actionable than the genome for dietary response, but it is also more variable — it changes with what you eat, how you sleep, whether you take antibiotics, your environment. A snapshot tells you about today. It does not tell you about next month.
Where it is actually useful today
There are areas where nutritional genomics is genuinely useful right now, not in the future-tense way of much biotech writing:
Nutrient deficiency risk
Variants in genes like MTHFR, VDR (vitamin D receptor), and BCO1 (beta-carotene conversion) give meaningful information about increased risk of specific deficiency states. If you carry certain MTHFR variants, you may need methylated folate rather than standard folic acid. This is not speculative — it is clinically applied in some contexts already.
Lactose and gluten
Lactase persistence is one of the most well-understood gene-diet interactions in human genetics. Testing for it tells you something real. Similarly, HLA-DQ2 and HLA-DQ8 testing has genuine predictive value for coeliac disease risk — not certainty, but meaningful risk stratification.
Caffeine metabolism
The CYP1A2 gene affects how quickly you metabolise caffeine. Slow metabolisers have meaningfully higher cardiovascular risk from high caffeine intake compared to fast metabolisers. This is one of the cleaner gene-diet interactions in the literature.
Omega-3 conversion
The FADS1/FADS2 gene cluster affects your ability to convert short-chain omega-3s (from plant sources) into the long-chain EPA and DHA that the body actually uses. If you carry variants that reduce this conversion, dietary ALA from flaxseed may matter less than you think, and direct EPA/DHA sources become more important.
The Indian context
Most of the large genomic studies underpinning this field have been conducted on predominantly European populations. The polygenic risk scores derived from these studies transfer poorly to other ancestries. This is a significant limitation, and it matters specifically here.
South Asian populations have distinct metabolic risk profiles — higher rates of insulin resistance at lower BMI, different lipid patterns, different gut microbiome compositions shaped by dietary traditions, different disease burden. The recommendations that emerge from European-population genomics studies may not transfer cleanly.
Research consortia like the GenomeAsia 100K project are beginning to build the population-specific reference data needed to make genomic nutrition relevant for South Asian individuals. But the gap between current knowledge and clinically actionable recommendations is larger here than it is for European-ancestry populations.
What this means practically
If you are considering a direct-to-consumer genomic test for dietary guidance, the most reliable value you will get is in the well-established single-gene interactions: lactase persistence, APOE variant (dietary fat and cardiovascular risk context), MTHFR variants (methylated B vitamins), and caffeine metabolism.
For everything else — macronutrient ratios, specific food recommendations, weight management — treat the output as directional context, not a prescription. Combine it with bloodwork, a food diary, and if you are serious about it, a clinician who specialises in preventive medicine rather than a chatbot that interprets your PDF.
Where it is going
The trajectory is clear even if the timeline is not. Continuous glucose monitors are already consumer-grade. Gut microbiome testing is improving in resolution and becoming cheaper. Polygenic score research for South Asian populations is growing. Machine learning applied to multi-omic data — combining genome, microbiome, metabolome, and wearable data — is producing models that predict individual food responses better than population-level guidelines.
Five years from now, the honest version of personalised nutrition will likely be more useful than the oversold version available today. The question is whether we can communicate clearly about what it can and cannot do in the meantime, so people do not spend money on answers that are not yet ready to be given.
The science is genuinely interesting. The premature certainty around it is not.