Deep-learning based representation and recognition for genome variants—from SNVs to structural variants  

作　　者：Songbo Wang Kai Ye 

机构地区：[1]School of Automation Science and Engineering,Faculty of Electronic and Information Engineering,Xi’an Jiaotong University,China [2]MOE Key Lab for Intelligent Networks&Networks Security,Faculty of Electronic and Information Engineering,Xi’an Jiaotong University,China [3]School of Life Science and Technology,Xi’an Jiaotong University,China [4]Faculty of Science,Leiden University,The Netherlands [5]Genome Institute,The First Affiliated Hospital of Xi’an Jiaotong University,China

出　　处：《National Science Review》2024年第11期3-6,共4页国家科学评论（英文版）

基　　金：supported by the National Natural Science Foundation of China(323B2015);K.Y.is supported by the National Natural Science Foundation of China(32125009);the National Key R&D Program of China(2022YFC3400300).

摘　　要：The evolution of genome sequencing and artificial intelligence(AI)has ushered in a new era of variant calling.Deep-learning methods have notably advanced the de-tection of both small-scale and large-scale variants,overcoming limitations and un-resolved issues faced by traditional meth-ods built on statistics and modeling.Here,we review eight deep-learning vari-ant callers,deconstructing their compu-tational procedures into two key mod-ules,representation and recognition.The representation module takes genome se-quencing data as input and encodes ge-nomic features with greater depth.The recognition module,built upon deep-learning models,tackles complex variant detection tasks and outputs variant prop-erties with superior accuracy.This per-spective aims to pave the way for leverag-ing AI in future variant research.

关 键 词：REPRESENTATION Deep LEARNING 

分 类 号：Q987[生物学—遗传学]