Investigating Lifestyle Data, microRNAs, and Immune Signals as Early Biomarkers of Pediatric Type 2 Diabetes
CSEF · 2026 Medicine & Physiology (Junior Division)
Overview
Type 2 Diabetes (T2D) is rapidly increasing among youth (2M), and projected to surge by 700% in 2060. Current medical strategies are reactive, focusing on post diagnosis treatment. Motivated by my personal loss, I sought to develop a tool to identify children at risk before symptoms appear. I hypothesised that at-risk youth exhibit coordinated changes in microRNAs, immune signals and lifestyle factors before glucose abnormalities appear. Using cross-sectional dataset NHANES 2013-2020 (ages 12-19), I developed 19 features including insulin resistance (HOMA-IR), beta-cell function (HOMA-B), with inactivity scores, nutritional stress and immune markers (CRP, NLR) as metabolic proxies. I trained Machine learning models with an AUC/ROC of 0.88 in distinguishing risk stages. The model revealed 30.7% of teenagers were in the reversible pre-compensatory phase characterized by normal glucose levels. Higher physical inactivity was identified as a key risk factor strongly associated with accelerated progression towards this phase. This study demonstrates that computational modeling of immune and lifestyle signals can identify critical and reversible pre-diabetes stages. Identification of 1 in 3 teenagers in the reversible pre-diabetes stage offers an opportunity to intervene and avert a public health crisis. To bridge the gap between data and intervention, I have published the model, accompanied by an UI application (on Github), that provides personalized recommendations and stage-specific interventions several years in advance. By utilizing "poor man's molecular marker" NLR, it's optimized for low cost real world usage. Future work includes validating these signals on comprehensive longitudinal datasets such as the TODAY study.
Competition history
- CSEF 2026
Related projects
CSEF · 2018
Type 2 Diabetes Prediction Using Longitudinal Machine Learning Analyses and Integrative Personal Omics Profiling
ISEF · 2023
BioRx: An Integrative NLP Approach to Early Survival and Recurrence Prediction and Novel Biomarker Discovery in Unstructured Text-Based Clinical Narratives for Diabetes Patients
ISEF · 2025
Epigenetic Analysis for Diabetes Risk and Resiliency
ISEF · 2021
Investigating Differential Gene Regulation in the PBMC of Obese Adolescents
CWSF · 2026
In Silico Diabetes Management and Prediction: A Personalized Hybrid Physics-ML System
CSEF · 2009
Down with Diabetes
CSEF · 2026
Leveraging Machine Learning for Equitable Screening of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD)
CSEF · 2017
Using Machine Learning to Predict Postprandial Blood Glucose in Type 1 Diabetics
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects