Creating a predictive model for type 2 diabetes improvement post bariatric surgery based on urine-based protein biomarkers

Alice Murphy; Graham Ball; Ioannis Kyrou; Jana Vrbikova; Vojtech Hainer; Petra Sramkova; Martin Fried; Gyanendra Tripathi; Philip McTernan

doi:10.1530/endoabs.109.P132

P132

Prev Next

Section Contents Cite

Endocrine Abstracts (2025) 109 P132 | DOI: 10.1530/endoabs.109.P132

SFEBES2025 Poster Presentations Metabolism, Obesity and Diabetes (68 abstracts)

Creating a predictive model for type 2 diabetes improvement post bariatric surgery based on urine-based protein biomarkers

Alice Murphy ¹ , Graham Ball ² , Ioannis Kyrou ³ , Jana Vrbikova ⁴ , Vojtech Hainer ⁵ , Petra Sramkova ⁶ , Martin Fried ⁶ , Gyanendra Tripathi ¹ & Philip McTernan ⁷

Author affiliations

¹Nottingham Trent University, Nottingham, United Kingdom; ²Intellegent OMICS Ltd, Nottingham, United Kingdom; ³Warwickshire Institute for the Study of Diabetes, Endocrinology and Metabolism (WISDEM), University Hospitals Coventry and Warwickshire NHS Trust, Coventry, United Kingdom; ⁴Institute of Endocrinology, Prague, United Kingdom; ⁵Institute of Endocrinology, Prague, Czech Republic; ⁶OB Clinic, Prague, Czech Republic; ⁷De Montfort University, Leicester, United Kingdom

Introduction: Bariatric surgery can result in substantial weight loss and type 2 diabetes (T2DM) remission, yet predicting those that will attain T2DM improvement can be challenging with 40-80% variability in remission. Current methods to predict successful bariatric outcomes are limited. Therefore, we aimed to investigate the use of pre-surgical urine samples as a non-invasive method to detect biomarkers and predict and/or monitor T2DM improvement, allowing stratification of patients to improve bariatric outcomes.

Methods: Urine samples were collected from female Caucasian participants (age=52.7±1.39; body mass index=41.6±1.09; n = 40) undergoing bariatric surgery at baseline (pre-surgery) or 6-months post-surgery. Mass spectrometry (MS) was used to assess protein data, identifying 2557 proteins across all samples, and Amica software was used to determine differences in those who improved their T2DM status based on HbA1c (n = 19), and those that didn’t (n = 15). Machine learning techniques, in the form of a swarm of neural networks, were undertaken to interrogate MS data and identify key proteins.

Results: MS identified 57 differentially expressed proteins (P < 0.05) in pre-surgery urine samples between these groups, with differential pathways including immune response and peptidase activity. In addition, neural network analysis used 20 models to identify key proteins that were stable across multiple models. The most stable protein in these models, indicating it had most impact on T2DM status improvement, was immunoglobulin-like cell surface receptor for CD47, with the chance of this being a false result less than 3.25x10^-38.

Discussion: These analyses indicate that pre-surgery urine samples may exhibit a different protein profile based on post-bariatric T2DM outcomes, including key proteins which could be used as biomarkers to predict T2DM status improvement post-surgery. This highlights the use of urine biomarkers as an additional method to stratify bariatric surgery patients to offer more personalised support and improve success rates.