Development and application of a Phytochemical Food Database (PhytoFooD) to assess the intake of dietary plant bioactives

This paper introduces PhytoFooD, a comprehensive database encompassing 1,067 bioactive compounds across 1,410 plant-based foods, to address gaps in existing literature and enable more accurate assessment of dietary plant bioactive intake and their health impacts.

The Gist

Imagine your diet isn't just a list of calories, protein, and fat. Imagine it's actually a massive, bustling city where every meal is a complex neighborhood filled with thousands of tiny, invisible characters. Some are famous celebrities (like Vitamin C), but most are obscure locals you've never heard of (like specific plant chemicals called "phytochemicals") that might hold the keys to your long-term health.

For a long time, nutritionists have been trying to map this city, but their maps were incomplete. They knew where the big landmarks were, but they missed the alleyways, the hidden gardens, and the secret recipes.

Here is the story of PhytoFooD, a new project that is finally drawing a complete map of this chemical city.

1. The Problem: The "Missing Pages" in the Recipe Book

Think of the food databases we use today (like the ones on nutrition labels or government websites) as old, tattered recipe books. They tell you how much sugar or fat is in an apple, but they often leave out the "secret ingredients" that make plants special.

• The Gap: We know about big players like polyphenols (found in berries) and carotenoids (found in carrots). But what about the lesser-known characters like glucosinolates (in broccoli) or monoterpenoids (in citrus)? They were often missing from the books because the information was scattered, messy, or hard to find.
• The Result: It's like trying to understand a symphony by only listening to the drums. You get the rhythm, but you miss the melody. Without knowing the full chemical composition of our food (what the authors call the "foodome"), we can't truly understand how our diet affects our health.

2. The Solution: Building the "Master Library" (PhytoFooD)

The researchers at the University of Parma decided to build a Master Library called PhytoFooD.

• The Collection: They didn't just look at one book. They gathered information from 13 different national food databases and scoured over 330 scientific studies. It was like sending a team of librarians to every corner of the globe to find every single note about plant chemicals.
• The Catalog: They organized 1,067 different bioactive compounds (the "characters") across 1,410 different plant-based foods (the "neighborhoods").
• The Sorting: They didn't just dump the data in a pile. They used strict rules to clean it up. If a study said a coffee had 10mg of caffeine and another said 1,000mg, they used statistical "detective work" to figure out if one was a mistake (an outlier) or if it was just a real difference. They even added a "Reliability Score" (Robustness Index) to every entry, telling you how confident they are in that number.

3. The Application: Why Does This Matter?

To show off their new library, the researchers used it to answer two big questions.

A. The European Diet Snapshot

They looked at what 22 European countries eat.

• The Findings: They found that while everyone eats different things, coffee is the undisputed king of plant chemicals in Europe. It's the main source of (poly)phenols and caffeine.
• The Surprise: In some countries, people are getting huge amounts of "N-containing compounds" (mostly caffeine) from coffee, while in others, fruits and vegetables provide the bulk of the "Terpenoids" (like the smell of oranges and carrots).
• The Analogy: It's like realizing that while everyone in a city eats lunch, some neighborhoods rely on pizza for their energy, while others rely on pasta. The type of fuel changes the chemical landscape of the city.

B. The "Caffeine Rollercoaster" (The Variability Problem)

This is the most exciting part. The researchers realized that not all foods are created equal, even if they have the same name.

• The Metaphor: Imagine you buy a "cup of coffee." You assume it has a standard amount of caffeine, like a standard-sized soda. But in reality, a cup of espresso is like a mystery box.
  • One cup might have a tiny 21mg of caffeine (a gentle nudge).
  • Another cup from the same shop might have a massive 306mg (a jolt).
• The Danger: The old databases usually just gave an "average" number. If you only looked at the average, you might think you're safe. But if you drink the "mystery box" with the high caffeine, you could accidentally exceed the safe daily limit (400mg) set by health experts.
• The Discovery: When the researchers applied their new database to real people, they found that many individuals were unknowingly drinking "high-caffeine mystery boxes" and potentially overdoing it. The old maps said "You're fine," but the new, detailed map said, "Wait, check your specific cup!"

The Takeaway

PhytoFooD is like upgrading from a blurry, black-and-white photo of your diet to a high-definition, 3D video.

• Before: We knew we ate "healthy plants."
• Now: We can see exactly which chemical characters are in those plants, how much of them there is, and how much they vary from one apple to the next.

This tool helps scientists, doctors, and even you understand that variability matters. It's not just about what you eat, but the specific chemical makeup of that specific piece of food you ate today. It paves the way for a future where we can tailor our diets not just to our weight, but to the complex, chemical symphony of our bodies.

Technical Summary

1. Problem Statement

Current dietary intake assessments are limited by the lack of comprehensive food composition databases for plant bioactives. While macronutrients and vitamins are well-documented, the "foodome" (the total chemical complexity of the diet) remains largely unmapped.

• Data Gaps: Existing databases often focus on major classes like (poly)phenols and carotenoids, excluding lesser-known but biologically active compounds such as glucosinolates, monoterpenoids, and alkaloids.
• Fragmentation: Information is scattered across national databases and scientific literature, often lacking standardization.
• Variability Ignored: Most databases report only average concentrations, failing to account for the significant natural variability in phytochemical levels caused by cultivar, seasonality, and processing. This leads to inaccurate intake assessments and potential misinterpretation of health risks or benefits.

2. Methodology

The authors developed PhytoFooD, a comprehensive database integrating data from multiple sources with a rigorous harmonization and statistical processing pipeline.

A. Data Collection & Curation

• Sources: Data were aggregated from 13 major national food composition databases (e.g., USDA, Phenol-Explorer, EuroFIR eBASIS, Ciqual) and an extensive literature review of over 330 peer-reviewed articles (searched via Scopus from Nov 2022 to March 2025).
• Classification: Compounds were hierarchically classified into four main families based on the PhytoHub system:
  1. (Poly)phenols
  2. Terpenoids
  3. N-containing compounds
  4. Miscellaneous phytochemicals
• Matching: Food items were matched to the Italian BDA (Food Composition Database for Epidemiological Studies) using name similarity and macronutrient profiles. New items were added if missing.
• Processing: Yield factors were applied to estimate phytochemical content in processed foods.

B. Statistical Analysis & Outlier Detection

To address concentration variability, the authors implemented a robust statistical workflow:

• Metrics: Calculated mean, standard deviation (SD), min, max, median, and coefficient of variation (CV) for each food-compound pair.
• Outlier Removal: For datasets with $n \geq 3$, three tests were applied:
  1. Mean $\pm$ 2 SD.
  2. Grubbs' test.
  3. Tukey's test (1.5 $\times$ IQR).
  • Rule: A value was excluded if at least two of the three tests flagged it as an outlier.
• Robustness Index: A quality score (1–3) was assigned based on sample size ($n$):
  • Index 3: $n \geq 5$ (High robustness).
  • Index 2: $3 \leq n < 5$.
  • Index 1: $n < 3$ (Low robustness, data scarcity).

C. Application Studies

1. European Population: Intake was estimated for 22 European countries using the EFSA Comprehensive Food Consumption Database (20 most consumed plant-based foods).
2. Individual Level (OPCT Cohort): Caffeine intake variability was analyzed in 297 Italian volunteers using the Oral (Poly)phenol Challenge Test (OPCT) study data, applying minimum, average, and maximum caffeine concentrations.

3. Key Contributions

• Scale: The database contains 1,067 bioactive compounds across 1,410 plant-based foods, totaling 41,130 mean concentration values.
• Breadth: Unlike previous databases, PhytoFooD includes under-represented classes such as glucosinolates, monoterpenoids, and various alkaloids.
• Variability Data: It is one of the first databases to systematically provide concentration ranges (min/max) and variability metrics (CV) rather than just averages.
• Quality Control: The introduction of a "Robustness Index" allows researchers to gauge the reliability of specific data points based on sample size.

4. Key Results

• Database Composition:
  • 17% of the data has a Robustness Index of 3 (high reliability, $n \geq 5$).
  • ~68% has an Index of 1 ($n < 3$), highlighting the scarcity of data for many compounds but ensuring high-quality selection criteria were used.
• European Intake Patterns:
  • (Poly)phenols: Highest intakes (~600 mg/day) observed in the Netherlands, Finland, and Germany, primarily driven by coffee consumption.
  • Terpenoids: Highest intakes (~500 mg/day) in Portugal, Spain, and Greece, driven by oranges and carrots.
  • N-containing Compounds: Caffeine is the dominant compound, often exceeding 50% of total intake in countries like Serbia, Bosnia, and Romania (800–1200 mg/day total).
  • Miscellaneous: Phytic acid (from potatoes and wheat) is the primary source, with highest intakes in Romania, Czechia, and Hungary.
• Impact of Variability (Caffeine Case Study):
  • Caffeine content in espresso varies drastically (21–306 mg per 30 mL cup; avg 135 mg).
  • Using average values underestimates risk. When maximum concentrations were applied to individual OPCT data, a significant portion of the cohort exceeded the EFSA safe threshold of 400 mg/day.
  • This demonstrates that relying on mean values can mask potential toxicological risks or overestimate safety margins.

5. Significance and Implications

• Research Tool: PhytoFooD serves as a critical resource for nutritionists and epidemiologists to move beyond macronutrient analysis toward a "chemical diet" approach, enabling more accurate associations between diet and health outcomes (e.g., cancer, cardiovascular disease).
• Safety Assessment: The study highlights that ignoring concentration variability can lead to flawed safety assessments. For compounds with narrow safety margins (like caffeine), using average values is insufficient for risk management.
• Future Directions: The authors acknowledge limitations, specifically the lack of data for certain low-molecular-weight terpenoids and the influence of cultivar/processing. They propose future updates to include metadata on cultivars and processing techniques to further refine intake estimates.
• Global Applicability: While initially applied to European data, the database is designed for global use, supporting the "Periodic Table of Food" initiative and the broader understanding of the "foodome."

In conclusion, PhytoFooD represents a paradigm shift in dietary assessment by providing a granular, variability-aware, and chemically diverse database that bridges the gap between food composition science and nutritional epidemiology.