Hi Dien, great job, it was a pleasure to work with us! Two questions, so you are continuing Richard, Rico, Metyu, Carlo and my own contributions. I think the biggest achievement you have is the OOP refactoring of the script. How easy would it be for you to extend the Docking class and its methods if we were to implement say another docking program (e.g. PLIP/BINANA) in the pipeline post-HEX?
And for your BRI1/BLD proof of concept, are figures a and b the same? The authors are simply moving the 3D structure around and choose to show the ribbons vs surface? Curious how yours would look if you had the surface/mesh view!
Since the structure of the docking module is compartmentalized using OOP refactoring, it will be easy to extend the Docking class to implement other docking programs post-Hex. Since the path to the results folder is stored in every Docking instance, we just have to write another method, for example, plip_docking() in the Docking class to run PLIP.
Yes, figures a and b are the same, they are just moved around and displayed with different representations of the receptor protein. I have tried using the same style (surface) when visualizing it in PyMOL, but the surface of the receptor covers the ligand almost completely, so I decided to switch back to ribbons.
Hi Dien! This sounds like a very great addition to the GAIA website! I see that you have done a comparison to the Hex Docking result to the publication figure. I’m wondering would their be any risk on the Hex Docking not producing the correct result? If so, what will be the approximate percentage of the undesired results?
Hi Emma! That’s a great question. Yes, there is always a chance that Hex doesn’t produce the correct docking conformation. Hex successfully identified good docking
orientations for two of the seven target complexes presented in the blind CAPRI (Critical Assessment of Predicted Interactions) docking experiment, and can place a good solution within the top 20 orientations for four of the seven targets. The specific accuracy percentage varies a lot, depending on the molecules used for docking. Hex is a rigid docking algorithm, which overall is not as accurate as flexible docking algorithms, but compared to other rigid docking algorithms, it ranks second (Pagadala NS et al. 2017) (Wang Z et al., 2016).
Hi Dien, fantastic job on the OOP refactoring and presentation! Your explanation made the docking module’s structure very clear. I’m curious about the average time it takes to complete one docking process. Could you provide an approximate timeframe for this? Also, since I understand this involves docking prediction, how do you assess the likelihood of a successful docking prediction? What metrics or methods do you use to measure this?
I don’t keep track of all the dockings we’ve done so far to generate a specific average, but based on the many times I have used it while testing to dock different molecules, it takes around 30 seconds. Larger receptors will take longer, possibly a minute. Since the docking and post-docking analysis are computationally expensive, users are limited to only 2 dockings per minute. These docking times only apply for new dockings, so if the docking has already been done, the results would be shown almost instantly.
Since this project is more of an extension and continuation of a previous project, assessing the likelihood of a successful docking using Hex was not one of the priorities. It has also been done by other labs in the past. In the CARPI trials, RMSD (Root Mean Square Deviation) was used to measure how accurate the dockings are. These are the results reported by the authors of Hex: “Overall, in Rounds 1-2 Hex did quite well, scoring 2 close hits for two of the seven targets. In Rounds 3-5, Hex scored a 1.8A RMS hit at rank 6 for target 12 (cohesin/dockerin) and got some further low to medium accuracy hits for targets 10, 11, 12 and 13, but missed the rest.” (Ritchie DW, 2002)
Hi Dien, great job, it was a pleasure to work with us! Two questions, so you are continuing Richard, Rico, Metyu, Carlo and my own contributions. I think the biggest achievement you have is the OOP refactoring of the script. How easy would it be for you to extend the Docking class and its methods if we were to implement say another docking program (e.g. PLIP/BINANA) in the pipeline post-HEX?
And for your BRI1/BLD proof of concept, are figures a and b the same? The authors are simply moving the 3D structure around and choose to show the ribbons vs surface? Curious how yours would look if you had the surface/mesh view!
Since the structure of the docking module is compartmentalized using OOP refactoring, it will be easy to extend the Docking class to implement other docking programs post-Hex. Since the path to the results folder is stored in every Docking instance, we just have to write another method, for example, plip_docking() in the Docking class to run PLIP.
Yes, figures a and b are the same, they are just moved around and displayed with different representations of the receptor protein. I have tried using the same style (surface) when visualizing it in PyMOL, but the surface of the receptor covers the ligand almost completely, so I decided to switch back to ribbons.
Hi Dien! This sounds like a very great addition to the GAIA website! I see that you have done a comparison to the Hex Docking result to the publication figure. I’m wondering would their be any risk on the Hex Docking not producing the correct result? If so, what will be the approximate percentage of the undesired results?
Hi Emma! That’s a great question. Yes, there is always a chance that Hex doesn’t produce the correct docking conformation. Hex successfully identified good docking
orientations for two of the seven target complexes presented in the blind CAPRI (Critical Assessment of Predicted Interactions) docking experiment, and can place a good solution within the top 20 orientations for four of the seven targets. The specific accuracy percentage varies a lot, depending on the molecules used for docking. Hex is a rigid docking algorithm, which overall is not as accurate as flexible docking algorithms, but compared to other rigid docking algorithms, it ranks second (Pagadala NS et al. 2017) (Wang Z et al., 2016).
Hi Dien, fantastic job on the OOP refactoring and presentation! Your explanation made the docking module’s structure very clear. I’m curious about the average time it takes to complete one docking process. Could you provide an approximate timeframe for this? Also, since I understand this involves docking prediction, how do you assess the likelihood of a successful docking prediction? What metrics or methods do you use to measure this?
I don’t keep track of all the dockings we’ve done so far to generate a specific average, but based on the many times I have used it while testing to dock different molecules, it takes around 30 seconds. Larger receptors will take longer, possibly a minute. Since the docking and post-docking analysis are computationally expensive, users are limited to only 2 dockings per minute. These docking times only apply for new dockings, so if the docking has already been done, the results would be shown almost instantly.
Since this project is more of an extension and continuation of a previous project, assessing the likelihood of a successful docking using Hex was not one of the priorities. It has also been done by other labs in the past. In the CARPI trials, RMSD (Root Mean Square Deviation) was used to measure how accurate the dockings are. These are the results reported by the authors of Hex: “Overall, in Rounds 1-2 Hex did quite well, scoring 2 close hits for two of the seven targets. In Rounds 3-5, Hex scored a 1.8A RMS hit at rank 6 for target 12 (cohesin/dockerin) and got some further low to medium accuracy hits for targets 10, 11, 12 and 13, but missed the rest.” (Ritchie DW, 2002)