Abstract
Recent data indicate that up-to 30-40% of cancers can be prevented by dietary and lifestyle measures alone. Herein, we introduce a unique network-based machine learning platform to identify putative food-based cancer-beating molecules. These have been identified through their molecular biological network commonality with clinically approved anti-cancer therapies. A machine-learning algorithm of random walks on graphs (operating within the supercomputing DreamLab platform) was used to simulate drug actions on human interactome networks to obtain genome-wide activity profiles of 1962 approved drugs (199 of which were classified as "anti-cancer" with their primary indications). A supervised approach was employed to predict cancer-beating molecules using these 'learned' interactome activity profiles. The validated model performance predicted anti-cancer therapeutics with classification accuracy of 84-90%. A comprehensive database of 7962 bioactive molecules within foods was fed into the model, which predicted 110 cancer-beating molecules (defined by anti-cancer drug likeness threshold of >70%) with expected capacity comparable to clinically approved anti-cancer drugs from a variety of chemical classes including flavonoids, terpenoids, and polyphenols. This in turn was used to construct a 'food map' with anti-cancer potential of each ingredient defined by the number of cancer-beating molecules found therein. Our analysis underpins the design of next-generation cancer preventative and therapeutic nutrition strategies.
Introduction
While dietary and lifestyle measures are known to prevent a significant portion of cancers, identifying the specific "active ingredients" in food remains a monumental challenge. Traditional nutritional science often evaluates molecules in isolation, failing to account for the complex, synergistic ways food interacts with human biology. Furthermore, experimental validation of thousands of bioactive compounds would take decades.
This study shifts the focus from a molecule's chemical structure to its biological function—specifically, how it perturbs the "interactome," the massive network of protein-protein interactions within the human body. By identifying molecules that mimic the network effects of proven anti-cancer therapies, researchers can map the "cancer-beating" potential of common foods.
Methods and Data Integration
The researchers developed a pipeline that treats the human body as a graph where nodes represent proteins and edges represent biological interactions.
The methodology involved four primary stages:
- Network Propagation: A "random walk with restarts" algorithm simulated how a molecule’s influence ripples through the interactome, generating genome-wide "activity profiles."
- Supervised Learning: Classifiers (SVM and MMC) were trained on the activity profiles of 1,962 approved drugs from DrugBank and DrugCentral. 199 were labeled as "anti-cancer."
- Crowdsourced Supercomputing: The team utilized DreamLab, a mobile cloud platform harnessing idle smartphone processing power to handle the massive computational load.
- Food Mapping: The model analyzed 7,962 bioactive food molecules from FooDB to predict their "Anti-Cancer Likeness" (ACL).
Key Results and Insights
The model demonstrated high predictive accuracy, identifying anti-cancer therapeutics with 84–90% accuracy. Out of nearly 8,000 food molecules, 110 were identified as "cancer-beating molecules" (CBMs) with an ACL threshold >70%. These include various flavonoids, terpenoids, and polyphenols.
The model's "learned" features aligned with critical cancer pathways, such as p53 signaling, DNA replication, and apoptosis. Key findings include:
- Top Ingredients: High CBM diversity was found in cruciferous vegetables (e.g., cabbage, broccoli), citrus fruits (oranges, lemons), and various herbs and teas (green/black tea, coriander, parsley).
- Network Similarity: Molecules can be chemically distinct yet biologically similar if they target the same neighborhood of the protein network.
- Synergy: Whole foods often provide a "consortium" of molecules that hit multiple oncogenic pathways simultaneously, potentially offering greater efficacy than isolated compounds.
Limitations and Future Directions
Despite the promising results, several challenges remain:
- Bioavailability: The model identifies potential efficacy but does not account for whether dietary intake reaches therapeutic concentrations in the bloodstream.
- Directionality: The current model identifies interactions but does not always distinguish between the inhibition or stimulation of biological processes.
- Drug-Food Interactions: Future work must address how these "hyperfoods" interact with the pharmacokinetics of traditional medications.
The ultimate goal is the development of "personalized food passports" and AI-optimized recipes to maximize the retention and efficacy of these cancer-beating molecules.
Scientific Reports | (2019) 9:9237 | DOI: 10.1038/s41598-019-45349-y
Summary updated