Nvidia Generative AI for Docking & Virtual Screening

[@Adventure_with_Homen]

Your description mentions that the platform is primarily designed to support tasks related to master’s and PhD theses. How does it ensure high-quality publications, particularly when compared to established tools like AutoDock and Vina?

[@vishwajeetgore3240]

hello pritam ,you have share a much helpful information .But i might have found little confusion .I also have perform mol docking by using autodock vina ,pymol ,openbabel and now i have also perform this ai diff docking .Now the result for this docking of different software are so much different so i just wanna know how much authentic or sucess do this software have and what are the factors are there which can tell the user that his /her docking is 100%accurate ,not 100% but what is authencity .Beacause if we look as a reserch point of view then do these blind docking show correct result .please help me😢

[@BioinfoCopilot]

Thank you 🙏. I would suggest to take a pdb structure which is bound to ligand and the active site is already known. Then use Diffdock to predict the conformation. If it aligns with the conformation predicted vs experimental, then you can consider diffdock as another tool to validate your results. Another approach is to compare all the tools you have been using for mol dock and then take the final conformation for MD simulation. Check after 100 ns whether it has any significant meaning or not. These are static poses generated by these different tools like autodock, diffdock etc. what you can consider is flexible docking! This approach is very useful to determine the exact conformation. Don’t consider scoring values as your way to define or select the binding poses, rather consider the interaction profiles with binding site residues. Sometimes the affinity scores may mislead you in determining the binding poses. So in conclusion: Validate using Flexible docking Do MD simulation Perform MMPBSA and FEP calculations Sometimes experiences matters in these scenarios 🙏

[@penscriber9320]

Its really cool. Thanks

[@BioinfoCopilot]

My pleasure

[@ES-yd1ze]

So diff dock can achieve by Python shell script or by gui of nvidia ? and same result ?

[@BioinfoCopilot]

Both are same!

[@sinhasuman1944]

Is this for just single molecule docking or it can do virtual screening?

[@BioinfoCopilot]

I have shown an example for single docking but you can perform virtual screening as well.

[@KamalSingh-dn7gv]

Hello Pritam, Introducing Nvidia's DiffDock, which uses generative AI for the docking of small molecules is indeed a promising idea. However, there are some inaccuracies in your video that should be addressed. It's important to ensure the information shared is scientifically sound, as misleading content can misinform viewers who may trust your statements. Specifically, there are two points of concern regarding your docking of remdesivir into the structure of SARS-CoV-2 Mpro: (i) Remdesivir is not a substrate of Mpro. (ii) The form of remdesivir you docked is not its active form. It is a prodrug form that requires metabolic conversion to a form that is accepted as a substrate by the target enzyme. I recommend reviewing these aspects and conducting further research to ensure accuracy in future content. It's essential to have a solid understanding of the science behind such AI techniques to maximize the potential of tools like Nvidia DiffDock and ensure their effective use. Otherwise, these AI tools can also be considered useless! Thank you - Kamal

[@BioinfoCopilot]

Hello Kamal, Thank you for your insights. The tutorials I made so far related to docking or dynamics are for demonstration purposes. The tutorials must not be taken into consideration for replicating or to publish etc. I think people who watch my videos know it very well. I don’t have to put a disclaimer or something in the beginning of the videos. Thank you 😊