An Improved Integrated Tempering Sampling Method

With the rapid development of computer technology, computer simulation has be-come an important research tool in the research of biomolecule and polymer. The mostchanlleging problem in molecular simulation of biomolecule is the complexity of thebiomolecule. The complexity of biomolecule partially comes from the complex struc-ture of biomolecule and partially comes from the long time-scale of the bioprocess. Toensure the stability of all-atom molecular simulation, it is necesary to adopt the time-step at fetomsecond timescale. To observe the bioprocess in milisecond timescale byall-atom molecular simulation, the required computational resourses are generally notavailable now. Therefore, a lot of efforts have been devoted in the last decades to studyhow to use molecular simulation method to study the long-timescale events, and var-ious enhanced sampling methods have been proposed and have been demonstrated tobe efficient in the study of biomolecules.In the molecular simulation of polymers, there are currently few enhanced sam-pling methods. Similar to biomolecule, how to use molecule simulation to study themulti-timescale events has been a very challenging problem. So, the development ofenhanced sampling method which is suited to polymer system is vitally significant forsolving a number of key problems related to polymer, for example, the glass transition.Because the biomolecule and the polymer have similar physical nature, the enhancedsampling methods which are mainly used in the research of biomolecule are possibleto be used in the research of polymer.Recently, GPU has experienced a dramatic development, now GPU has been ahighly parallelized, multi-threads, multi-cores processor, with outstanding computa- tional ability and broad memory band-width. CUDA is a program model and softwareenvironment designed for GPU, it is implemented as the extension of C language inGPU. The ocurrence of CUDA has made programmable GPU much easier to use. GPUhasbeenwidelyapliedtovariousfields,theheterogeneouscomputersystemwithGPUshas been established in our country such as Tianhe-1. Because GPU is very suitable forthe massive floating-point calculation with high computaional density such as molec-ular simulation, now how to use the huge computational power of GPU has been a hotissue.Professor Yi Qin Gao in Peking University recently proposed an integrated tem-pering sampling (ITS) method, this method has been demonstrated to be efficient inenhancing the sampling of biomolecule system. However, when this method is appliedto polymer system, we encounter problems such as it is hard to get the weighting fac-tor required to launch an integrated tempering sampling simulation. In this thesis, wepropose an improved integrated tempering sampling method. We propose an easy andquick way to determine the weighting factors. Based on the optimization of the energydistribution generated in the integrated tempering sampling simulation, this methodcoulddeterminetheweightingfactorseasilyandquicklywithoutiteration. Wealsopro-posed an easy-to-use way to determine the temperature distribution used in integratedtempering sampling simulation. By using the intrisic similarity of replica exchangemethod and integrated tempering sampling method, we could easily determine the tem-perature distribution with enough overlap between the adjacent temperatures. We notonly solve the problem and extend its applicability in the original ITS method but alsomakeadeepinsightintotheessenceoftheintegratedtemperingsamplingmethod. Thismethod could be implemented in typical GPU accelerated simulation software packageeasily and thus very suitable to be combined with GPU.Inordertobenchmarkourmethod,weapplythismethodtothestudiesofLennard- Jones fluid, the trans/cis comformational transition of ALA-PRO dipeptide and theCoil-Globule transition of a flexible polymer chain. The results demonstrate that ourmethod is valid, robust, effcient and could be used to study complex polymer systems.Thus, this method is very promising to make contribution to solving some importantissues of polymer, such as glass transition.