You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
A standardized meal can produce a highly repeatable glucose response without telling us much about how the same person responds to other foods.
We found this directly when comparing two open CGM datasets. In one dataset, a single standardized breakfast predicted later responses to other meals quite well. In another, the result depended strongly on the challenge meal.
This project asks what makes a metabolic challenge informative rather than merely repeatable.
I will run one harmonized analysis across the existing datasets and any additional open dataset that meets predefined inclusion criteria. For each challenge meal I will measure:
1. same-meal repeatability
2. cross-meal transportability
3. out-of-sample prediction improvement
4. predictive value beyond basic clinical biomarkers
The aim is to identify which properties of a challenge meal make it useful for predicting future glucose responses.
At the minimum level, I will complete the harmonized analysis, robustness checks and public release of code and results.
Additional funding will support more datasets, more challenge definitions, stronger out-of-sample validation and external methodological review.
Main costs are research time, data engineering, computation and replicatio
I am Ilya Nikitin, founder and director of The Null Institute.
The first analyses are already complete in two independent CGM datasets. They produced a clear contradiction: some standardized meals transport well across meals, while others are repeatable but do not.
https://nullinstitute.org/
The main risk is that the effect is highly dataset-specific.
That would still be a useful result. It would show that a standardized meal cannot be treated as a general personalization test without validating its cross-meal transportability.
No external funding has been received for this project. The work so far has been self-funded.