基于数字宏基因组的病毒DNA片段群落检测及致病预警平台

Excavation and identification of new viruses is a key to prevent, diagnose and treat the viral diseases. The traditional diagnosis techniques exist many limitations and shortcomings in research of excavating new viruses, which can be overcomed by metagenomics. As doesn't rely on traditional culture techniques and could fully tap the virus community in specific environment, metegenomics is widely promoted in clinical medicine in recent years. With the development of sequencing technologies, more and more metegenomes are sequenced, thus how to tap richer viral information from these massive virus genes is the urgent problem needed to be solved by us. This project intends to treat metagenomics as the research object, and plans to design a pathogenic warning platform by combining some pattern recognition knowledge from the perspective of bioinformatics, which can be used to identify the virus populations in metagenomics. This platform has three research issues: (1) how to optimize the high-dimensional characteristics extracted from viral DNA fragments? (2) how to design a pattern statistical method utilized to automatically identify biological communities of viral DNA fragments? (3) how to design a reasoning model for the unknown virus? The successful construction of this platform can not only quickly identify known viruses and find unknown viruses, but also visually depict the viral abundance of samples from variety of environments, and real-time monitor changes of several potentially pathogenic viruses. So this project has an important clinical guidance for the prevention and treatment of new infectious diseases.

发掘和鉴定新病毒是预防、诊断和治疗病毒性传染病的首要任务。传统的诊断技术在挖掘新病毒的研究中存在很多不足。不依赖传统培养技术的宏基因组学能克服这些不足，充分挖掘特定环境中的病毒群落，近年来，在临床医学中得到推广。随着测序技术的发展，越来越多疾病的宏基因组被测序，如何从这些海量宏基因中挖掘丰富的病毒信息，是我们迫切需要解决的问题。本项目拟以宏基基因组为研究对象，从生物信息学的角度，结合模式识别知识，搭建基于宏基因组的病毒DNA片段群落检测、预警平台，该平台围绕三方面研究：(1).如何优化病毒DNA片段中提取的高维数字特征？(2).如何设计模式统计方法自动鉴定病毒DNA片段的群落？(3).对未知病毒，如何设计推理模型？该平台的搭建不仅可以迅速鉴定已知病毒，发现未知病毒，并可以直观地描绘出各种疾病样本中的病毒丰度，实时监测一些潜在致病病毒的变化情况。对新发传染病的预防与治疗具有重要的临床指导意义

近年，由于不依赖传统培养技术的宏基因组学能充分挖掘特定环境中的病毒群落，在临床医学中得到推广。然而，如何从这些海量宏基因中挖掘丰富的病毒信息，是我们迫切需要解决的问题。本项目以宏基基因组为研究对象，从生物信息学的角度，结合模式识别知识，搭建了基于宏基因组的病毒DNA片段群落检测、预警平台，研究内容如下：(1).利用粗糙集算法优化了从病毒DNA片段中提取的高维数字特征。(2).基于有监督的SVDD算法和无监督的模糊c-mean算法，自动鉴定已知病毒DNA片段的生物群落。(3).利用深度学习算法，设计了未知病毒推理模型。该项目当前的研究，初步实现了对已知病毒和一些未知病毒的鉴定，可以直观地描绘出各种疾病样本中的病毒丰度。未来的进一步完善，可实现监测一些潜在致病病毒的变化情况，对新发传染病的预防与治疗具有重要的临床指导意义。