 So what does this mean if we look at folds. Well the first thing is that folds if I pick a fold with a particular feature say that is very rigid. If the fold is rigid and I'm going to need to fold my sequence into that fold that means that there are lots of constraints on it. So a fold that is rigid and does not have a whole other flexibility will only be able to accommodate a few sequences. But a fold like this one you can probably see the loops here and imagine that this is somewhat flexible. If it has some sort of conformational freedom here it will be able to accommodate more sequences and then we're going to see it more frequently. And this does not sound so bad but imagine that you had a defect that was in the ballpark of 10 k cal. Is that good or bad? Well we know how to compare that. That would mean the relative probability p1 divided by p2 that should be e raised to minus delta g divided by kt right? So if I have my bad fold that is say 10 k cal worse. If delta g is 10 k cal kt was what in k cal? Sadly I can't cross and examine you but it's 0.6. So this would be e raised to roughly minus 16 or which is roughly equal to 1 million or 1 in a million. So suddenly if I have 1 million fold sorry 1 million fold where it was somehow be happier or so just based on this reasoning if I make the energy 10 k cal worse I will only be able to save 1 in a million of those. This looks beautiful and you understand all this because it's the Boltzmann distribution right? We're so close and yet it's completely wrong. You can't use the Boltzmann distribution here. Sorry for fooling you. Why can't you use the Boltzmann distribution? Turn off the camera or stop the video one minute and think about that. Are you back? The problem is that the Boltzmann distribution relied on exchange of states. We have some sort of distribution right? You've seen many of these and there has to be an exchange going in both directions. I need to actually sample things right? You might even remember the so-called detailed balance that at the equilibrium the flux in both directions over the peak have to be the same. What is the flux of one sequence to another sequence? There is no such flux. You don't get alanine spontaneously being changed into phenylalanines. So we do not when we don't have sampling when we don't sample different states we can't use the Boltzmann distribution. The Boltzmann distribution was entirely about sampling. Forget about it. This has to be something different.