Rosetta - patrickchirdon/AIdrugdiscovery GitHub Wiki

https://www.rosettacommons.org/ http://rosie.rosettacommons.org/ note- there is a fee for commercial users

http://www.meilerlab.org/index.php/rosetta-tutorials

https://en.wikipedia.org/wiki/Rosetta@home

I provided a protein ligand docking shell script that you can run Rosetta from. Please note that the full 60GB package of Rosetta needs to be downloaded to a cloud of your choosing. I also provided a docking script. The only lines that need to be edited are the lines that contain directories in the .sh script and the line of the dock.xml file which specifies the coordinates of pocket for the protein of interest (search for move from). The coordinates of various binding pockets can be found using servers such as--

http://fpocket.sourceforge.net/ (note you must use pymol to find the x,y,z coordinates once fpocket highlights the pockets)

http://www.scfbio-iitd.res.in/dock/ActiveSite.jsp

http://altair.sci.hokudai.ac.jp/g6/service/pocasa/

For a description of how the algorithm works see-- https://www.ncbi.nlm.nih.gov/pubmed/16972285

Rosetta ligand docking is based on Monte Carlo minimization. Its Monte carlo method has three components--

  1. random starting postion of the ligand in a binding pocket- move 1 angstrom in each direction, random rotations of mean .05 degrees around each axis

  2. side chain conformations are repacked using rotamer trials

3.Rigid body orientation and side chain X angles of the ligand are optimized using Davidson-Fletcher-Powell algorithm

The move is allowed if energy decreases.

Rosetta uses an energy function that accounts for Van der Waals forces, hydrogen bonding potential, and electrostatics. The method is notable for including protein flexibility, which is important for accuracy. Attractive interactions are measured with Lennard Jones potential. There's a repulsive term, solvation term, hydrogen bonding term, and coloumb charge model based on the CHARM27 forcefield. Electrostatics inside the protein are calculated from pair potential from PDB statistics.

**The goal of using protein ligand docking software is to find the lowest free energy pose of the ligand in its binding site. Note- Correlation and prediction quality vary largely depending on the type of protein.

The binding energy correlation between experimental small molecule binding with the virtual binding is R= .63