Cluster a protein-ligand system with RMSD using separate sets of atoms for the alignment and distance computation. Both the "alignment" and "distance" atoms for each frame will be loaded, and in each frame, the cartesian center of the alignment atoms is computed and subtracted out of the coordinates of both the alignment and distance atoms.
Distances are calculated between two frames by computing the rotation matrix which optimally overlays the alignment atoms with one and other, and then applying that rotation matrix to the distance atoms in the first frame and computing their root-mean-squared deviation w.r.t the other frame.
Clustering is performed with the k-centers algorithm. You may specify the termination criterion either by the number of states, or a cutoff distance (which ensures that all data points are within that distance of their assigned cluster center).
This script uses hybrid MPI/OpenMP paralleism in addition to highly optimized
SIMD vectorization within the compute kernels. Using multiple MPI processes
requires running this command using your MPI implementation's process manager,
e.g. mpirun, mpiexec, or aprun. The number of OpenMP threads can be
controled by setting the OMP_NUM_THREADS environment variable. For example,
$ export OMP_NUM_THREADS=4
$ mpirun -np 16 clusterLP-MPI <options>
This script requires python, the latest development version of MDTraj available on github, numpy, and mpi4py.
No installation is required. Just call the script or move it into your PATH
if you like.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent