Tissue-based Alzheimer gene expression markers–comparison of multiple machine learning approaches and investigation of redundancy in small biomarker sets

摘要: Alzheimer’s disease has been known for more than 100 years and the underlying molecular mechanisms are not yet completely understood. The identification of genes involved in processes Alzheimer affected brain is an important step towards such understanding. Genes differentially expressed diseased healthy brains promising candidates. Based on microarray data we identify potential biomarkers as well biomarker combinations using three feature selection methods: information gain, mean decrease accuracy random forest a wrapper genetic algorithm support vector machine (GA/SVM). Information gain two commonly used methods. We compare their output to results obtained from GA/SVM. GA/SVM rarely analysis data, but it able capable classifying tissues into different classes at least reference Compared other methods, advantage finding small, less redundant sets that, combination, show superior classification characteristics. biological significance gene pairs discussed.