A comparison of classification methods for predicting Chronic Fatigue Syndrome based on genetic data

Area of Science:

• Genomics
• Computational Biology
• Medical Informatics

Background:

• Selecting significant genes is vital for disease susceptibility studies.
• Genetic factors like single nucleotide polymorphisms (SNPs) are key in understanding complex diseases.

Purpose of the Study:

• To compare computational tools with and without feature selection for predicting chronic fatigue syndrome (CFS).
• To identify influential single nucleotide polymorphisms (SNPs) associated with CFS.

Main Methods:

• Utilized a dataset from the CDC Chronic Fatigue Syndrome Research Group.
• Applied classification algorithms: naive Bayes, support vector machine, and C4.5 decision tree.
• Employed hybrid (chi-squared, information-gain) and wrapper-based feature selection methods to identify influential SNPs.

Main Results:

• The naive Bayes model combined with the wrapper-based feature selection approach demonstrated superior performance.
• This model effectively inferred disease susceptibility by analyzing the complex relationship between CFS and SNPs.

Conclusions:

• The developed approach shows promise for assessing associations between CFS and SNPs.
• Feature selection enhances the predictive power of computational tools in genomic association studies.