An Intelligent Learning Approach for Identifying Genetic Disorder by Detecting Gene Expression Malfunctions

Misregulation or gene mutations cause the gene to act differently from other patterns of genes, causing a person to inherit a particular disorder from their parents or through environmental changes. The early detection of disorders is made possible by gene expression analysis, which also provides a pathway for the patient’s diagnosis. This proposed work combines both filter and wrapper methodologies to analyze the differential expression of genes and yield the best classification for genetic diseases. Hybrid Ant Colony Optimization is applied for the optimization process that results in the best features. Feature selection starts with building similarity graphs, which aids in the subset gene selection process, Fisher Score is determined using mean, standard deviation, and class information. Next, the Pearson Correlation Coefficient is generated using the correlation of the gene information, which is estimated. The adjacency matrix is created using the correlation vector value as a base. The value of for r0 the threshold correlation vector is 0.32. Finally, the gene subset was extracted based on the highest significance. The proposed Forest Feedforward Neural Network and Forest Backpropagation Deep Neuro-Fuzzy Network algorithms are used for classification, and the outcomes of the methods are estimated based on their performance comparison, where the accuracy estimated as (91.09%,92.78%,93.06%). The performance of the proposed algorithm is more appropriate than the conventional algorithm for solving complex applications like gene classification because they need access to a massive amount of gene data. The Deep Neural Network algorithm's motivation is that, without being expressly stated, it finds features that correlate and then combines them to enhance faster learning.