Alzheimer's diagnosis by an efficient pipelined gene selection model based on statistical and biological data...

Diagnosing Alzheimer's disease based on gene expression data extracted from microarrays is still an open field of research. Due to the availability of whole-genome data through microarrays technology, diagnosis accuracy is expected to be improved. Despite the high potential of the data prepared by the technology, their analysis on different platforms shows that they may differ for different samples concerning biomarker status. This affects the diagnosis accuracy because of the existing bias between two experimental conditions. To address this problem, we propose a pipeline-based approach to diagnose Alzheimer's disease using statistical analysis of biological data combined with artificial intelligence techniques. At first, the B-statistics and a new score based on a gene interaction network are used to evaluate genes. The B-statistics helps us to find differentially expressed genes. The new score, called the evidence score, measures the compliance level of the differentially expressed genes with past biological evidence. Next, we use artificial intelligence methods to find the subset of genes that define high separability between normal and affected samples. To this end, we employed a genetic algorithm to find the optimal subset. The performance of the pipeline was compared with other state-of-the-art methods. The results indicate that the proposed method can obtain fruitful predictive performance for diagnosing Alzheimer's disease. All the codes implemented in this study are available online at https://github.com/HamedKAAC/AD-gene-selection.