Extracting predictive SNPs in Crohn's disease using a vacillating genetic algorithm and a neural classifier in...

Area of Science:

• Genetics
• Computational Biology
• Medical Informatics

Background:

• Crohn's disease is a heritable inflammatory bowel disease.
• Predicting Crohn's disease using single nucleotide polymorphisms (SNPs) is challenging due to the vast number of possible SNP subsets.
• Existing methods struggle to identify optimal SNP sets for accurate disease prediction.

Purpose of the Study:

• To develop a novel computational framework for predicting Crohn's disease.
• To identify an optimal subset of SNPs for maximizing prediction accuracy.
• To overcome limitations of existing SNP selection algorithms.

Main Methods:

• Proposed a new algorithm utilizing chromosomes of varying lengths with feature selection.
• Incorporated novel cross-over and mutation operations within the algorithm.
• Employed a 5-fold cross-validation approach on real-world Crohn's disease data.

Main Results:

• The proposed SNP prediction framework demonstrated superior performance compared to Optimum Random Forest (ORF), USVM, CGSP, CSP, and DNF.
• Achieved 90.4% accuracy in predicting Crohn's disease.
• Reported 87.5% sensitivity and 92.2% specificity.

Conclusions:

• The developed algorithm effectively identifies informative SNP subsets for accurate Crohn's disease prediction.
• This novel approach offers a significant advancement in computational methods for predicting complex genetic diseases.
• The framework's high accuracy and validated performance show promise for clinical applications.