Gromacs-plumed元动力学

标准的meta动力学模拟需要设定如下标准参数

metad: ARG=__FILL__ PACE=__FILL__ HEIGHT=__FILL__ BIASFACTOR=__FILL__ SIGMA=__FILL__ FILE=__FILL__ GRID_MIN=__FILL__ GRID_MAX=__FILL__ GRID_BIN=__FILL__ TEMP=__FILL__

ARG: 需要添加高斯偏置势的集合变量CV (e.g. distance, angle, dihedral, etc);

PACE: 每隔多少步注入高斯偏置函数, 设置的步数应足够大以留给非偏置CV的舒展 (常选择100-500 步);

HEIGHT: 高斯偏差函数的高度 (rule of the thump: ~0.1-1.0*kBT KJ/mol, too low->stay trapped for long/too high->lose local minima, 0.02-1.20 for distance and 1.0-2.0 for angles are commonly used values)

BIASFACTOR: 高斯偏差函数的高度增长的速度 (set close to the barriers you want to cross, barriers 20-30 KJ/mol need a bias factor of ~10-15, 10 is reasonable choice for biological systems)

SIGMA: 高斯偏差函数的宽度 (0.5*unbiased range of variable, nm for distance & rad for angle)

FILE: the output file where the Gaussians are deposited along the trajectory

TEMP: the temperature of the system (needed for well-tempered MD to calculate kB)

在长时间模拟中,由于需要在模拟期间存储和处理的高斯值的数量不断增加,元动力学的性能显著下降。如果偏置势能值可以存储在grid中这个问题就可以轻松处理。The following arguments determine the boundaries and bin size of the grid(单位为:埃):

GRID_MIN: the minimum value your CV can adopt

GRID_MAX: the maximum value your CV can adopt

GRID_BIN: the bin size

These arguments should be enough to run a fast and efficient Metadynamics simulation.


plumed.dat文件示例:

# treat each molecule as whole

WHOLEMOLECULES ENTITY0=1-5, ENTITY1=6-10 # 考虑周期性边界对分子完整性的影响,将各组分各自看成一个完整分子

# 声明CV变量
dist: DISTANCE ATOMS=5,10 
dist: DISTANCES GROUPA=ATOMS1-3 GROUPB=ATOMS4-6 LABEL=distAB

# metadynamics参数 ==>将数值输出至HILLS文件中
metad: METAD ARG=dist PACE=500 HEIGHT=0.2 SIGMA=0.05 FILE=HILLS BIASFACTOR=10 GRID_MIN=0.1 GRID_MAX=5.0 GRID_BIN=500 TEMP=300 # in HILLS the following variables will be written: time, dist, sigma, height, biasfactor
                          #时间   CV    高斯偏差函数宽度 高度  偏置因子

# 监测==>将数值输出至COLVAR文件中
PRINT STRIDE=500 ARG=dist,metad.bias FILE=COLVAR # print the calculated distance and bias in COLVAR every 500 steps, COLVAR will contain three columns time, dist and bias

联合gromacs实现元动力学采样模拟

gmx mdrun -deffnm production -plumed plumed.dat -v

你可能感兴趣的:(MD相关,plumed,python,人工智能)