Multisite λ Dynamics

1. Hardware requirements
   Linux system with NVIDIA GPU
   !! root or superuser privileges are NOT required !!

2. Software requirements
   CHARMM program
   miniconda
   PyMOL
   MMTSB
   LOMAP -- Optional

3. References literature

   • λD original paper
   • MSλD paper
   • Adaptive Landscape Flattening (ALF)
   • Protein Mutations - I
   • Protein Mutations - II
   • Biasing potential replica exchange (BPREX)
   • msld_py_prep and charge corrections
   • MSλD benchmark paper
   • AMBER & OPLS in MSλD
   • FEP analysis guidelines
   • Best practices FEP
   • MBAR
   • charge changing mutations
   • charge changing mutations FEP+

4. Github Resources

   • ALF
   • pyCHARMM workshop
   • MSλD workshop
   • protein ligand benchmark datasets
   • protein tutorial
   • precomputed mutation free energies

Compiling CHARMM

** You can follow the steps from the pyCHARMM workshop as well, below is just a part of the things that are on the pyCHARMM workshop page **

1. Install miniconda and create a new environment
   conda create --name charmm
conda activate charmm

• install mamba within the conda environment to make the next steps fast
  conda install conda-forge::mamba
• Please see more info on pyCHARMM workshop's page regarding cuda and GCC
  • check the version of NVIDIA drivers
    nvidia-smi
  • install cuda version that is compatible with your NVIDIA drivers
    mamba install -y -c "nvidia/label/cuda-x.y.y" cuda
  • use GCC version that is recommended for your cuda version
    mamba install -y -c conda-forge gcc==X.X gxx==X.X gfortran==X.X
• a few more packages through conda
  mamba install -y -c conda-forge make cmake binutils fftw openmpi openmm sysroot_linux-64==2.17 readline==8.2 rdkit openbabel pymol-open-source pymbar

2. Compile CHARMM

• active your conda environment
  conda activate charmm
• untar CHARMM
  tar -xvf charmm.tgz
cd charmm
mkdir build_charmm
cd build_charmm
../configure -u --with-blade --without-mkl
  I'm using 8 cores (see next command) to compile/build charmm, you can use more cores eg. 10 or 20 or less eg 1 or 4.
  make -j 8 all
  DONE