Stochastic Simulation Of Excitation-contraction Coupling

工作概述我们已经开始对所谓的自发钙释放事件给予更多的关注，这些事件在形态上从单个火花到局部小波再到全球钙波。这些事件已被证明在控制起搏细胞的频率以及在病理条件下触发心律失常方面发挥着关键的时钟作用。为了模拟这一过程，我们正在构建一个基于簇的三维随机反应扩散模型，该模型描述了嵌入在有限元网络中的钙释放通道簇的动力学，在这个网络中，钙可以扩散并经历缓冲反应。一个专用的Beowulf集群(48个处理器)的硬件刚刚交付，正在安装。同时，我们已经成功地将我们之前的蒙特卡罗激发-收缩模拟代码移植到NIH Biowulf计算集群，在多达160个处理器上进行并行处理。这使得探索局部控制模型的参数依赖性成为可能，这将导致对增益调节机制的新见解。并行化模型将为包括钙释放的三维传播所需的广泛软件开发奠定基础。即使有了新的硬件，在很短的时间内完成这项计算也需要特殊的编程方法。我们开发了一种专门的方法来积分反应扩散方程，它利用了生物物理方程的结构，运行速度比传统的偏微分方程包快100倍。我们最近发明了一种分层积分方案，在该方案中，微分方程求解器递归地调用自己，以便在更接近钙源的区域使用较小的时间步长，同时在构成胞浆体积的开放空间中采取大步长。这些方法现在正在用Fortran对这两个集群进行编码。同时，我们正在收集大量关于心肌细胞钙释放事件统计特性的实验数据，以便与3D随机模拟模型的输出进行比较，后者准备运行。

SUMMARY OF WORK We have begun devoting more attention to so-called spontaneous calcium release events, which vary morphologically from single sparks to localized wavelets to global calcium waves. These events have been shown to play a critical "clock" role in controlling rate in pacemaker cells, as well as triggering arhythmias under pathological condition. In order to model this process, we are constructing a cluster-based thre dimensional stochastic reaction diffusion model describing the dynamics of calcium release channel clusters embedded in a finite-element network in which calcium can diffuse and undergo buffering reactions. Hardware for a dedicated beyowulf cluster (48 processors) has just been delivered and is being installed. Simultaneously, we have successfully ported our previous Monte Carlo excitation-contraction simulation code to the NIH Biowulf computing cluster, parallelized on up to 160 processors. This makes it possible to explore the parameter dependence of the local control model, which is leading to new insights about the mechanism of gain regulation. The parallelized model will form the basis for the extensive software development required to include the three dimensional propagation of calcium release. Even with the new hardware, doing this computation in tractable time will require special progamming methods. We have developed a specialized method for integrating reacdtion diffusion equations that takes advantage of the structure of the biophysical equations to run 100 times faster than a conventional partial-differential equation package. We have recently invented a hierarchical integration scheme in which the differential equation solver calls itself recursively in order to use smaller time steps in regions closer to the calcium sources while taking large steps in the open spaces that constitute most of the volume of the cytosol. These methods are now being coded in Fortran for the two clusters. At the same time, we are collecting extensive experimental data on the statistical properties of calcium release events in cardiac myocytes, for comparison with the output of the 3D stochastic simulation model when the latter is ready to run.