Can you explain the scale we use for the heat map… specifically what cutoffs would you use to distinguish between a high-contact residue versus a low-contact residue?
I am also very curious in what cellular conditions (pH, water content, membrane-bound) these dockings happen in, can you discuss this a little given your knowledge from your review of these docking tools? I remember that the PDB files were generated by Phyre2.
The color of heatmap is based on the value of accumulated contact frequency for each residue which is in the output result json file of the python script. For every residues that is not found in this result file, they are not the position docked in all previous docking results. All values of contact frequencies found in the file are normalized and transferring to hex color code. Since the contact frequency is negative, hence, the lowest value representing the highest-contact residue will be the red color, vice versa. There is no actual cutoff to distinguish the contact.
Currently, all proteins and ligands are given from other labs or grabbed from Pubchem and PDB, therefore, therefore, there is no given cellular conditions for them. However, it is definitely that there should not be water molecules for each input molecules. To run the Hex docking program in this project with a proper output, correlation type is set as “Shape only” to do basic shape-based docking. The search around the twist angle, so that the calculation can be arranged to make the intermolecular twist angle search is in the innermost loop of the search, is set to be accelerated using a 3D FFT due to the statement from Hex User Manual (http://hex.loria.fr/manual800/hex_manual.pdf) demonstrates that 3D FFT is faster than 5D and 1D for basic shape-based docking. Docking grid dimension is in 0.6 for saving memory to run the server since the usual range to use a fine grid is 0.25 to 0.5Å. Max docking solutions is 5000 to increase the difference among result energies. Receptor and ligand range angles are set as 180 degrees to have a fair fraction of the total possible rotational increments. Docking receptor and ligand step sizes, alpha (i.e. twist) range angle, and alpha step size are set as 7.5, 360 and 5.5 respectively as default settings. Docking main (i.e. Steric) scan was performed at N=16 followed by main search at N=25 since these settings are found to work well in practically all cases. Finally, 500 possible docking result files were saved.
Hi Yuexin,
At 7:20, what do the colours on the ligand represent? I notice some parts of the molecule are red or orange, is this also showing binding probability? And does it use the same scale as the protein colours?
Hey Yue Xin,
I really enjoyed your presentation, I found it well detailed and explained. I was wondering if there is anyway to add a second ligand to the binding site, as a future prediction?
This was a great explanation of your tool! My question is why was rigid body docking chosen over fluid? Is the assumption that the protein and ligand are rigid generally true?
Nice illutration with graph, flow chart and demo. So how long will it take to predict the interaction?
For each protein-ligand pair, it takes around 25 seconds.
Hi Yue Xin,
Can you explain the scale we use for the heat map… specifically what cutoffs would you use to distinguish between a high-contact residue versus a low-contact residue?
I am also very curious in what cellular conditions (pH, water content, membrane-bound) these dockings happen in, can you discuss this a little given your knowledge from your review of these docking tools? I remember that the PDB files were generated by Phyre2.
The color of heatmap is based on the value of accumulated contact frequency for each residue which is in the output result json file of the python script. For every residues that is not found in this result file, they are not the position docked in all previous docking results. All values of contact frequencies found in the file are normalized and transferring to hex color code. Since the contact frequency is negative, hence, the lowest value representing the highest-contact residue will be the red color, vice versa. There is no actual cutoff to distinguish the contact.
Currently, all proteins and ligands are given from other labs or grabbed from Pubchem and PDB, therefore, therefore, there is no given cellular conditions for them. However, it is definitely that there should not be water molecules for each input molecules. To run the Hex docking program in this project with a proper output, correlation type is set as “Shape only” to do basic shape-based docking. The search around the twist angle, so that the calculation can be arranged to make the intermolecular twist angle search is in the innermost loop of the search, is set to be accelerated using a 3D FFT due to the statement from Hex User Manual (http://hex.loria.fr/manual800/hex_manual.pdf) demonstrates that 3D FFT is faster than 5D and 1D for basic shape-based docking. Docking grid dimension is in 0.6 for saving memory to run the server since the usual range to use a fine grid is 0.25 to 0.5Å. Max docking solutions is 5000 to increase the difference among result energies. Receptor and ligand range angles are set as 180 degrees to have a fair fraction of the total possible rotational increments. Docking receptor and ligand step sizes, alpha (i.e. twist) range angle, and alpha step size are set as 7.5, 360 and 5.5 respectively as default settings. Docking main (i.e. Steric) scan was performed at N=16 followed by main search at N=25 since these settings are found to work well in practically all cases. Finally, 500 possible docking result files were saved.
Very cool, Michelle! (: Nice visualization of the workflow.
Thanks
Hi Yuexin,
At 7:20, what do the colours on the ligand represent? I notice some parts of the molecule are red or orange, is this also showing binding probability? And does it use the same scale as the protein colours?
Hey Yue Xin,
I really enjoyed your presentation, I found it well detailed and explained. I was wondering if there is anyway to add a second ligand to the binding site, as a future prediction?
Hello Yue Xin,
This was a great explanation of your tool! My question is why was rigid body docking chosen over fluid? Is the assumption that the protein and ligand are rigid generally true?