Advanced Data Analytics for Food Industry

Tell me what you eat, and I will tell you who you are. But tell me how much and when you eat it, and I will tell you whom you will become. These are the great puzzles that application researchers in the food industry try to solve every day. Any industry wants to make the most of its raw materials, so knowing which nutrients make a healthy cereal bar, for example, can be as important as knowing when eating that cereal bar will be most restorative when eaten. Otherwise, investing in new nutrients can be ineffective, and the solution to thwart avoidable costs is to decipher the health associations between nutrients and the human organism. These health associations are known as the food-health nexus, one of the most important and exciting challenges for the 21st century, and not just for the food industry.

You are right if the term “food-health nexus” does not sound like a new challenge. It is one of those ever-present goals of humanity that just crossed some game-changing milestones. But before getting into the milestones, let us start with some quick fundamentals. It turns out that, when digesting food, there is one thing more diverse than the nutrients we eat; it is digestion itself. This complexity exists because our metabolism is influenced by anything and everything: age, sex, weight, blood pressure, and even our lifestyle habits, such as exercise, smoking, and drinking. We know this because we can probe our metabolism by looking at the fingerprints it leaves after nutrients get digested. These fingerprints are known as metabolites, and they make up the link (or nexus) between food and health.

So, what is the food industry’s bottleneck? Application researchers do not look for “cereal bar” in our system; they search for the countless metabolites resulting from whole-grain, fruit, and sugar digestion and then study a select few for their effects on human health. However, this shows the classical trade-off between specificity and power versus robustness and generalized applicability. But now, the field of nutritional metabolomics suffered a major analytical leap forward, gaining on previous technological advancements; advanced analytics is narrowing the gap between both sides of the industry’s trade-off. All-in-one analytical toolkits that perform time-series analysis, clustering, and database benchmarking allow application researchers to test which nutrients have the strongest health outcomes (positive or negative) under which circumstances while comparing and reporting results in international curated databases.

Having arrived at this stage, at Wiimer, we developed a conceptual framework, rooted on artificial intelligence, to guide the next steps in advanced analytics for the food industry. First and foremost, we need to profit from such unprecedented amounts of information. On the one hand, we are in a privileged position to train and test machine learning methods on known metabolite-health associations. The expected outcome is a robust toolkit for predicting new associations while keeping up with the discovery of new metabolites. On the other hand, measuring nutrients but mostly health changes will get a significant do-it-yourself element. If smartwatches can read blood pressure and electrocardiogram signals, we should soon have affordable metabolites sampling without a single drop of blood needed.

According to The Economist, the Apple Heart Study and the Fitbit Heart Study each processed bio-data from more than 400,000 people. About 0.5%-1% of participants in each study got an alert about irregular heartbeat and, in both data sets, a third of people suffered atrial fibrillation. Both Fitbit and Apple provide robust predictive models, leading to high true positive results (98% and 84%, correspondingly) and just 0.7% of cases got a false alert.


Fig. 1: The scope of wearable health devices (The Economist, 2022)


A “drop” in costs happened with genetic testing, and in the foreseeable future, physicians could adjust diets that were ineffective due to latent inherited predispositions. Not only that, but they could also design custom diets before genetic predispositions manifest, all with the proper analytical tools at their disposal. Finally, finding the optimal food-health nexus leads to literal profit. Predicting the appropriate amounts of those nutrients with the best health benefit can reduce unnecessary waste in large-scale food production. Moreover, a food-health-genetic nexus can assist other industries, including health insurers and providers. Knowing and predicting the needs of a hospital’s community should lead to fewer drug and supply depletions, and infrastructure overload.


Legend: 50-year-old person, who is a frequent runner and smoker. Lactose intolerance, latent hypertension.


Fig. 2: Changes in metabolite and blood pressure levels after the start of a diet developed from informed advanced analytics