UK Biobank Proteomics Data for Predicting Kidney Complications in Type 2 Diabetes

UK Biobank data from 50,021 participants without type 2 diabetes, tracked from 2006-2010 with an average follow-up of 12.0 to 12.03 years. The dataset includes over 1,400 proteins and more than 280 metabolites analyzed for complications like kidney, cardiovascular, and neurological issues. Author Ming Hao used LASSO Cox and LightGBM models to identify GDF15 as a predictive biomarker for kidney complications.

Use Cases

• Building predictive models for kidney complications based on plasma protein GDF15 levels.
• Comparing biomarker performance against traditional indicators like blood glucose and HbA1c.
• Training machine learning models for early screening of diabetic complications using proteomic data.
• Analyzing the relationship between omics components and various diabetes-related health outcomes.

Strengths

• Dataset includes over 1,400 proteins and more than 280 metabolites per participant.
• Analysis covers 50,021 participants with a prospective follow-up of approximately 12 years.
• Research employed nine specific machine learning methods to validate findings.
• Results show GDF15 achieved an AUC of 0.94 for predicting 5-year kidney complications.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the UK Biobank source.

Provenance

Source

UK Biobank

Collection Method

Prospective study analyzing proteomic and metabolomic data.

Time Range

Baseline from 2006 to 2010, with follow-up averaging 12.0 to 12.03 years.

Freshness

Last updated 2026-05-11 04:26:03; freshness should be verified.

Geography

United Kingdom