DNA sequence classification for diabetes mellitus using NuSVC and XGBoost: A comparative

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Precision Medicine

Background:

• Diabetes Mellitus is a global health issue requiring early risk identification.
• Traditional diagnostics miss genetic predispositions.
• Genomic analysis offers potential for identifying diabetes-related genetic markers.

Purpose of the Study:

• To classify DNA sequences for predicting Diabetes Mellitus susceptibility.
• To compare the performance of NuSVC and XGBoost machine learning models for this task.

Main Methods:

• Utilized NLP techniques to process DNA sequences as text, converting them to numerical features via TF-IDF.
• Applied SMOTE to address class imbalance in the dataset.
• Trained and validated NuSVC and XGBoost models using 10-fold cross-validation.

Main Results:

• XGBoost achieved 98% accuracy, 0.0650 log loss, and 1.00 AUC.
• NuSVC achieved 87% accuracy, 0.2649 log loss, and 0.95 AUC.
• XGBoost demonstrated superior performance across all evaluated metrics.

Conclusions:

• XGBoost is robust for analyzing complex genetic data in Diabetes Mellitus prediction.
• Machine learning, particularly XGBoost, shows significant potential for early diabetes diagnosis.
• Integrating advanced ML models into healthcare can enhance predictive diagnostics for diabetes.