Deep learning application for genomic data analysis Chang Beom
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Modern genomic sequencing techniques have advancedrapidly, thereby improving data production rates and dimensionality. With this accelerated growth, machine learning,especially deep learning, has been leveraged to analyze complexdata and complement conventional bioinformatics methods. Deep learning approaches have been successfully applied ingenomics, leading to the development of state-of-the-art modelsand significantly improved interpretation of genomic data. Here, we review deep learning models in four genomic domains:variant calling, gene expression regulation, motif finding, and3D chromatin interactions. We summarize the key aspects ofmodel development, such as training and generalization, thatenable the efficient application of deep learning models ingenomic research. Real-world applications have demonstratedthe reliability and efficiency of these models for predicting genomicprofiles. Finally, we highlight the future directions of deeplearning approaches in genomics by discussing the challengesrelated to genome tokenization and multi-omics data integration.
Abstract
Modern genomic sequencing techniques have advancedrapidly, thereby improving data production rates and dimensionality. With this accelerated growth, machine learning,especially deep learning, has been leveraged to analyze complexdata and complement conventional bioinformatics methods. Deep learning approaches have been successfully applied ingenomics, leading to the development of state-of-the-art modelsand significantly improved interpretation of genomic data. Here, we review deep learning models in four genomic domains:variant calling, gene expression regulation, motif finding, and3D chromatin interactions. We summarize the key aspects ofmodel development, such as training and generalization, thatenable the efficient application of deep learning models ingenomic research. Real-world applications have demonstratedthe reliability and efficiency of these models for predicting genomicprofiles. Finally, we highlight the future directions of deeplearning approaches in genomics by discussing the challengesrelated to genome tokenization and multi-omics data integration.