| 查看: 1227 | 回复: 7 | |||
| 当前主题已经存档。 | |||
[交流]
【求助】MD关于平衡的问题 已有2人参与
|
|||
| 我在建立模型的时候，用discovery动力学平衡，但是我不知道如何选取合适的dynamics time，一般现在很多人都用1ns，但是如果体系太大，选择这个时间运算会很长，我想缩短一点时间，请问有谁知道怎么从平衡图上面看出体系已经达到了平衡啊? |
回复此楼» 猜你喜欢
化学工程321分求调剂
已经有12人回复
-
211本，11408一志愿中科院277分，曾在中科院自动化所实习
已经有4人回复
-
材料专硕326求调剂
已经有5人回复
-
东南大学364求调剂
已经有5人回复
-
国自科面上基金字体
已经有7人回复
-
药学383 求调剂
已经有4人回复
-
286求调剂
已经有5人回复
-
085601求调剂
已经有3人回复
-
302求调剂
已经有5人回复
-
考研化学学硕调剂，一志愿985
已经有5人回复
2楼2010-04-02 10:02:59
★ ★ ★ ★
jings(金币+5):可以加个qq么？57790934，我还不是太懂，想具体问一下你，谢谢了 2010-04-02 10:17
ghcacj(金币+4):谢谢 2010-04-02 10:26
jings(金币+5):可以加个qq么？57790934，我还不是太懂，想具体问一下你，谢谢了 2010-04-02 10:17
ghcacj(金币+4):谢谢 2010-04-02 10:26
|
当然，这只是一个很粗略的标记，具体可以参考如下 How to check when a system reaches it's equilibrium This is a nontrivial question, and the answer partly depends on what you are trying to do. As David points out, if you want to do protein folding, say, your protein needs to have long enough to forget about its starting state (hence his comment about folding and unfolding). Usually that takes a prohibitively long time, though, for large proteins. So assuming you don't want to solve the protein folding problem you have to be content with something less. Definitely you want to run long enough that the protein RMSD quits increasing, temperature and energy begin fluctuating around some relatively stable value, and so on. You can also look for things like number of protein-water hydrogen bonds if you have explicit solvent; this tends to take a much longer time to equilibrate. Probably a good suggestion is to find out what other people have done who are doing something similar to you, and make sure you equilibrate at least as much as they do. Just a remark about the rmsd with respect to the Xray structure: you might want to plot rmsd vs log(time), that is a better check sometimes. Well, rmsd is not that bad. Just that it is easier to judge a converged rmsd from a non-converged one when you plot it against log(t) instead of t. Another convergence check is the cosine content of the traj projected onto some (slow) PCA modes. 详细的可以参考这里 http://www.mdbbs.org/viewthread. ... ghlight=equilibrium |
3楼2010-04-02 10:05:18
4楼2010-04-02 10:09:25
5楼2010-04-02 10:27:45
6楼2010-04-02 10:29:41
guosy1999
木虫 (正式写手)
中纪委反2协会督导员
- 应助: 1 (幼儿园)
- 金币: 2625.6
- 散金: 793
- 红花: 7
- 帖子: 677
- 在线: 334.1小时
- 虫号: 632814
- 注册: 2008-10-21
- 性别: GG
- 专业: 能源化工
7楼2010-04-15 10:17:49
bay__gulf
金虫 (著名写手)
刘苏州
- 模拟EPI: 8
- 应助: 9 (幼儿园)
- 贵宾: 4.85
- 金币: 2332.8
- 红花: 1
- 帖子: 1344
- 在线: 271小时
- 虫号: 592012
- 注册: 2008-09-03
- 专业: 理论和计算化学
- 管辖: 分子模拟
★ ★ ★ ★ ★
小木虫(金币+0.5):给个红包，谢谢回帖交流
lei0736(金币+4):谢谢 2010-04-15 10:42
小木虫(金币+0.5):给个红包，谢谢回帖交流
lei0736(金币+4):谢谢 2010-04-15 10:42
|
判断体系是否到达平衡没有好的方法, 几乎全是靠经验 理论上, 体系到达平衡之后, 所有的状态参数都不会改变了 但某些不改变不一定意味着就到了平衡, 而对所有参数进行检测有很不现实 这些参数中, 使用了控温算法后温度最不可靠, 能量一般很容易收敛, 能量也不可靠 原子数多了, 体系的势能面极为复杂, 很容易落在一个local minimum 亚稳态中, 就像这里 http://muchong.com/bbs/viewthread.php?tid=1557782&fpage=1 的那个例子一样 看到体系像是平衡, 1ns - 2ns 不变了, 未必3ns 时候不变 一个体系越大, 到达平衡所需的时间越长, 如果考虑到充分的各态遍历取样, 几百个ns 对生物大分子体系都是远远不够的的 你必须非常熟悉你的体系, 通过长期的模拟积累的经验, 和灵敏的化学/结构的直觉来判断 模拟的时间是越长越好, 但在特定的历史时期是不允许超长时间的 太短了没杂志要, 太长了又是一种浪费, 多读文献, 看看现在流行多长就用多长吧 [ Last edited by bay__gulf on 2010-4-15 at 10:41 ] |
8楼2010-04-15 10:34:59
