Using Recurrent Neural Networks for Predicting Type-2 Diabetes from Genomic and Tabular Data.

Area of Science:

• Genomics
• Computational Biology
• Biomedical Informatics

Background:

• Genomic technology and artificial intelligence are transforming computer-aided diagnostics and therapies.
• Genomics enables prediction of future illnesses like cancer, Alzheimer's, and diabetes.
• Machine learning accelerates biomedical research and computational biology.

Purpose of the Study:

• To predict type 2 diabetes using gene sequences from genomic DNA fragments.
• To develop and test automated feature selection and extraction for gene pattern matching.
• To evaluate the performance of Recurrent Neural Network (RNN) models, including Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU).

Main Methods:

• Utilized genomic DNA fragments for gene sequence analysis.
• Employed automated feature selection and extraction techniques.
• Tested Recurrent Neural Network (RNN) models, including LSTM and GRU, on tabular data for type 2 diabetes prediction.
• Assessed model performance using Sensitivity, Specificity, Accuracy, F1-Score, and Mathews Correlation Coefficient (MCC).

Main Results:

• The suggested model demonstrated fair accuracy in predicting future illnesses.
• Recurrent Neural Network components (RNN, LSTM, GRU) were evaluated for their effectiveness in processing genetic data.
• The model showed potential for real-world application in disease prediction.

Conclusions:

• The developed model can predict type 2 diabetes with reasonable accuracy using genomic data.
• The research highlights the utility of advanced machine learning models in genomic data analysis for disease prediction.
• The proposed system is suitable for real-world scenarios, with secure data handling capabilities for risk variable evaluation via an Android application.