Design And Development Of Cloud Platform Of Association Analysis For Complex Traits

Many economically important traits in plants and animals,as well as human complex diseases,are complex.Association analysis method based on mixed linear model can accurately predict the genetic architecture of complex traits and effectively uncovers their genetic mechanisms.Our research group has proposed a mixed linear model approach for genome-wide association analysis and developed a GPU-based software,named QTXNetwork.It consists of three functional modules: QTL analysis(QTLNetwork),genome-wide association analysis(QTS)and multi-omics association analysis(QTT).However,the development of biology software lags far behind the development of algorithms.Existing biocomputing software generally adopts a traditional architecture model,which is not scalable and highly available.And it can only be used by a single user in a single thread,with complex configuration and difficult user operation,which affects the promotion and application of the method.Meanwhile,with the development of biotechnology,the increasing of biological data in exponention has led to more computational complexity of software.Individual server is unable to scale computing power quickly and copying data offline,as a result,the software will become more complex to use.Presently,computation has become a major bottleneck in biology research.To address this problem,we design a solution to turn the software into containers and finally compute on our private cloud platform.Users can remotely use the elastic and scalable computing power of the cloud platform,in addition,the threshold of using GPU,etc.,parallel computing resources will be reduced,while,the efficiency of computation and analysis will be increased.Our cloud platform contains Iaa S,Paa S and Saa S.The Iaa S is mainly implemented through Starling X,the Paa S is built by Kubernetes,and the Saa S includes technologies such as microservice and container.The five analysis tools QTLNetwork,QTS＿Raw,GMDR,QTS＿GPU,and QTT＿GPU are structured into a cloud-native architecture to build QTXCloud,a Saa S-based web services application(http://ibi.zju.edu.cn/QTXCloud/).Software testing has shown that the container-based cloud platform for complex trait association analysis is highly available.It greatly reduces the difficulty of use and improves the utilization of physical resources.On the other hand,the cloud platform has high scalability and elasticity in providing computing resources.While using microservices,platform functions can be added or removed in the form of plugins.