Related Experiment Video
Updated: Jan 11, 2026
An In Ovo Model for Testing Insulin-mimetic Compounds
Published on: April 23, 2018
Neural network modelling of experimental diabetes to study new antidiabetic drugs
Zhansulu Nurgaliyeva1, Ayazhan Dauletkaliyeva2, Serzhan Dauletkaliyev3
1Department of Pharmacology, 374640 Kazakh-Russian Medical University , Almaty, Republic of Kazakhstan.
Objectives:
Diabetes mellitus is a complex metabolic disease characterised by chronic hyperglycaemia, which triggers a cascade of pathological changes in the body. Contemporary methods for developing antidiabetic drugs are often lengthy, expensive, and not always effective. The use of neural networks for modelling diabetes opens up new possibilities for accelerating research and increasing the accuracy of predicting the effectiveness of novel therapeutic strategies. The aim of this study was to develop and validate a neural network for studying experimental diabetes and assessing the efficacy of new antidiabetic drugs, as well as to explore the potential of combined therapeutic strategies.
Methods:
The empirical investigation was conducted using laboratory models of diabetes induced in rats with streptozotocin. Three groups were formed: a control group, a diabetic group without treatment, and a diabetic group treated with experimental drugs.
Results:
A neural network, based on multilayer perceptrons and recurrent architectures, was trained to predict changes in glucose levels, oxidative stress markers, and the condition of pancreatic tissues. The developed model demonstrated high predictive accuracy of metabolic changes, with an average accuracy of 92.3 %. As a result of treatment with experimental drugs, blood glucose levels in rats decreased by 25-30 % over 28 days, accompanied by a 26 % reduction in oxidative stress markers and partial restoration of pancreatic β-cell function in 30 % of cases. Histological analysis confirmed reduced fibrosis and improved tissue condition in the treatment group. The model also identified that combined therapeutic strategies - for example, the combination of antioxidants with gluconeogenesis inhibitors - had a synergistic effect, lowering glucose levels by up to 40 %.
Conclusions:
The study confirmed the effectiveness of the developed neural network for analysing therapeutic strategies and predicting metabolic changes in diabetes models. The proposed neural network represents a promising tool for investigating new antidiabetic drugs, including efficacy assessment within personalised medicine. Its application may accelerate preclinical research, optimise therapeutic approaches, and contribute to reducing drug development costs.
Insights
A new neural network accurately predicts diabetes progression and drug effectiveness, accelerating research. This AI tool aids in developing novel antidiabetic therapies and reducing drug development costs.
Area of Science:
- Biomedical research
- Computational biology
- Pharmacology
Background:
- Diabetes mellitus is a chronic metabolic disorder with significant global health implications.
- Current antidiabetic drug development is often lengthy, costly, and lacks consistent efficacy.
- Artificial intelligence, specifically neural networks, offers potential to enhance diabetes research and drug discovery.
Purpose of the Study:
- To develop and validate a neural network model for studying experimental diabetes.
- To assess the efficacy of novel antidiabetic drugs using the developed model.
- To explore the potential of combined therapeutic strategies for diabetes treatment.
Main Methods:
- Utilized streptozotocin-induced diabetes in rat models.
- Employed multilayer perceptron and recurrent neural network architectures.
- Trained the model to predict glucose levels, oxidative stress, and pancreatic tissue status.
Main Results:
- The neural network achieved 92.3% accuracy in predicting metabolic changes.
- Experimental drugs reduced blood glucose by 25-30% and oxidative stress by 26%.
- Combined therapies, like antioxidants and gluconeogenesis inhibitors, showed synergistic effects, lowering glucose by up to 40%.
Conclusions:
- The developed neural network effectively analyzes therapeutic strategies and predicts metabolic changes in diabetes.
- This AI tool shows promise for accelerating preclinical research and optimizing antidiabetic drug development.
- The model supports personalized medicine approaches and can reduce overall drug development costs.
