Protein Chemical Shift Prediction
Protein Chemical Shift Prediction
复制标题
DOI：
--
复制
发表时间：
2014-09
期刊：
影响因子：
--
通讯作者：
A. S. Larsen
中科院分区：
文献类型：
--
作者：
A. S. Larsen
蛋白质化学位移包含了大量关于蛋白质三维结构的信息。一些化学位移预测的基础上解决与X射线晶体学和相应的实验化学位移的结构之间的关系已经开发。这些经验预测是非常准确的X射线结构，但往往是不敏感的小结构变化。为了克服这一限制，有人建议基于量子力学（QM）计算来制作化学位移预测器。在这篇论文中，QM衍生的化学位移预测器Procs 14的发展。Procs 14基于三肽的235万次密度泛函理论（DFT）计算，并包含对氢键，环电流和前一个和后一个残基的影响的校正。Procs 14能够对13 CA、13 CB、13 CO、15 NH、1HN和1HA骨架原子进行预测。为了对Procs 14进行基准测试，对完整的蛋白质结构进行了许多QM NMR计算。在测试的经验和QM导出的预测值中，Procs 14以最高的准确度再现了QM化学位移。与QM导出的预测器CheShift-2在X射线结构和NMR系综上与实验化学位移数据的比较表明，Procs 14预测化学位移的准确性最好。NMR合奏的预测表现出最好的性能。这表明，未来的工作可能会受益于使用集成采样时进行模拟的蛋白质折叠与化学位移。Procs 14在马尔可夫链蒙特卡罗蛋白质折叠框架PHAISTOS中实现。Procs 14的计算效率实现允许快速预测，因此可能用于蛋白质结构的改进和折叠。
The protein chemical shifts holds a large amount of information about the 3-dimensional structure of the protein. A number of chemical shift predictors based on the relationship between structures resolved with X-ray crystallography and the corresponding experimental chemical shifts have been developed. These empirical predictors are very accurate on X-ray structures but tends to be insensitive to small structural changes. To overcome this limitation it has been suggested to make chemical shift predictors based on quantum mechanical(QM) calculations. In this thesis the development of the QM derived chemical shift predictor Procs14 is presented. Procs14 is based on 2.35 million density functional theory(DFT) calculations on tripeptides and contains corrections for hydrogen bonding, ring current and the effect of the previous and following residue. Procs14 is capable at performing predictions for the 13CA, 13CB, 13CO, 15NH, 1HN and 1HA backbone atoms. In order to benchmark Procs14, a number of QM NMR calculations are performed on full protein structures. Of the tested empirical and QM derived predictors, Procs14 reproduced the QM chemical shifts with the highest accuracy. A comparison with the QM derived predictor CheShift-2 on X-ray structures and NMR ensembles with experimental chemical shift data, showed that Procs14 predicted the chemical shifts with the best accuracy. The predictions on the NMR ensembles exhibited the best performance. This suggests that future work might benefit from using ensemble sampling when performing simulations of protein folding with chemical shifts. Procs14 is implemented in the markov chain monte carlo protein folding framework PHAISTOS. The computational efficient implementation of Procs14 allows for rapid predictions and therefore potential use in refinement and folding of protein structures.