Deep Learning-Based Detection of Disease-Associated DNA motifs in Whole Exome Sequencing Data for Inherited Genetic Disorders

Abstract

This study presents a deep learning framework for detecting disease-associated DNA sequence patterns in Whole Exome Sequencing (WES) data related to inherited genetic disorders. The proposed workflow combines sequence extraction, one-hot encoding, class balancing using the Synthetic Minority Over-sampling Technique (SMOTE), and a one-dimensional convolutional neural network (1D-CNN) classifier. Variant data were obtained from ClinVar and filtered for three disease-associated genes representing diabetes, myopia, and asthma, together with benign variants. Experimental results demonstrate that the proposed framework can effectively classify disease-associated sequence patterns within the available dataset, achieving an overall accuracy of 97%. The findings suggest that deep learning-based sequence analysis may provide a useful computational approach for variant prioritization in genomic research. However, further validation using larger and independent genomic datasets is required before clinical applicability can be established.