 So what do docking results look like? Well, that depends a lot on what you're starting from. In some cases it can look beautiful. The general challenge here is really that if you provide crap in, you're going to get crap out. And that in particular means that if you have a very high quality x-ray structure, say two angstrom or something, then things might work great. But caveat emptor here, two angstrom doesn't necessarily tell the whole story. Two angstrom, that's the resolution of your entire structure, right? What we really care about here, what is the local resolution right around the binding site? The loops, I could live without them. So you need to look into the specific part that is your binding site and check what the resolution is there. This is even more important if you don't have a real structure but had to build a homology model. Now, if you had a template that was 90% identical to your sequence and that template was very high resolution, you're likely going to have a beautiful binding site in here. But if that is not the case, well, then all bets might be off. Maybe at least you have really good quality of the sequence alignment right next to the binding site or at least 80% of the residues in the binding site are identical. Then I would also trust it. Or maybe you have a structure that's undergoing conformational change or something, then things start to bear a little bit more iffy. On the other hand, this was a COVID target, right? So that's the early part of the plan. During 2020, we didn't really have any choice. We would dock anything we had to it. But you should always have this nagging feeling that it's quite okay to use an homology model, but at the end of the day, you need to be able to argue why your homology model was good enough to do docking. Don't just put it into a program and assume that it's a model so you had a structure because you're going to live to regret that. No model, no matter how good it is, will be able to beat the docking results you get if you actually have a real X-ray structure. And that has increasingly happened now, in particular with COVID. Just the last few weeks Bill Jorgensen and others have provided a whole range of tests. I think in this case they tested 17 predicted docked compounds and 40 of them had activity. That leads me to the second recommendation here. In particular, if somebody else has done these studies, it's not just the docking software or the RMST that matters, but the person involved. I would trust Bill Jorgensen literally with my life. I would take those drugs if you said that they're going to cure COVID and I had it. Random publication you found on the internet? Well, maybe, maybe not. I'm not saying those authors are bad. It's just that I don't know them. But Bill, I know that he will have begun through a ton of details, checked, double checked, triple checked. And if he says it's true, I would likely trust him. So check where your results are coming from. The devil is in the detail and you need to spend more time than you think you're just looking at a molecular viewer.