 My plans for today are the following. I could not Protein design is a hot topic and there's a lot of things happening both in research and the world and not to mention big farm industry I Suspect that half of the things I tell you today are going to be completely outdated over there a couple of years But since it's very difficult to find any information in books about this I figured I should share a few slides with you at least and that might take depending on how much questions you have And how interested you are anything between 30 and 45 minutes Then I might not do a break well, it depends a bit on your taste But then I figured then after that I'm going to be around for questions and answers relating to everything in the course repetition Questions about the exam life the universe everything as long as you want But in case you just have one or two questions is stupid for you to go away 15 minutes on a break before we get back And in principle I need to go to the US on Friday morning, but I can be around here until then Hopefully it's not gonna take that long and then let's see We might I might give you a little I don't have prepared any slides about handling task 3 But if your questions about handling task 3 we can certainly repeat that to and go through it a bit So drug design both hand in task 3 and the last lecture I gave you are very much related to drug design, of course And drug design you can approach the top down or bottom up You can think of yourself as Pharma or a doctor being interested in curing a particular disease For some of these diseases we hardly know anything about them on the molecular level while for others such as insulin There's like 50 60 years of research We know pretty much everything in and out is just that we don't have a perfect drug yet Historically as I mentioned last week virtually everything in drug design has been about serendipity You happen to find something that works and then you purify this or somehow improve the doses treatments Whatever and then hopefully you find something that is Well working so that you can administer the patients and actually cure the disease What's happened the last 20 maybe even well roughly 20 years is that there has been a very increased focus on Designing drugs. There are two reasons for that one of them is that we want to be able to Use drugs for diseases that we previously haven't been able to treat There simply might not be anything that bind well enough I mentioned I mentioned that last week that even if you do find something that bind if the binding affinity is not high enough You would need an insanely large dose maybe kilos per day and in practice That's not gonna work you to side effects and those side effects means that all the food and drug Administrations all over the world are gonna be worried and then you might not get the drug approved And even if you did get it approved Most of us would not want to take a lifestyle drug if there are too many side effects Sure, if it's gonna cure fatal cancer disease Then you might do it, but not to say reduce your levels of cholesterol The other challenge there is that all the regulatory Demands have increased we will not accept random deaths and random deaths might sound horrible But in many cases this might correspond to a small genotype variation a deviation in your genome that you just happen to be One in hundred thousand which is too bad if you die from the drug and that's like I realize how horrible that sounds but A lot of health has to be approached in the population level right if there if it is a disease that is otherwise fatal But this drug will cure nine hundred ninety nine nine hundred ninety nine hundred ninety nine people out of the million But one in a million will die. It's probably a pretty wise drug to take because the alternative is dying with 100% certainty What is increasingly happening though is that we're running out of ideas or rather not really running out of ideas But we're not our methods are not good enough We can't design a custom drug to do anything in the body We need some sort of lead. We need something that at least starting to bind to receptor It's very difficult for you to say pick and Forget about buying I talked about binding pockets last week, right? But forget about the binding pocket assuming that there is a new random protein that We know that that protein is relevant to cancer and there is a particular mutation in it But there isn't any binding pockets that might not be involved in binding as part of its normal course of biological interaction and yet I would like to change the way that protein works How do we get something to bind? It doesn't have a binding pocket And that's more difficult right if you don't the second there is a pocket the second there is something to start with The approach I mentioned last week usually works We just need to design something that is better either better at binding than the natural ligand or that somehow blocks the Pockets of the natural ligand can't bind and that's not impossible. Not trivial mind you, but not impossible So increasingly we would like to do things that are completely from scratch or de novo bottom up And what that has increasingly meant is that we no longer limit ourselves to small compounds So that if you talk about modern drug design and there are very few examples of drugs like this on the market Why seems like an obvious right that if there is so many throughout the course you looked at all the diversity of proteas structures What if we could custom design proteins? I'm not sure to compare this to well you can imagine that using small ligands That's kind of like living a shed in the forest Designing proteins is like building you can build houses. We can do absolutely anything only your imagination is the limit So if that's so obvious why haven't we been doing it for 20 years? So was this course trivial? No, and the problem is it's not trivial. It's a pretty darn hard problem Just predicting protein structure is very very very difficult and now you're gonna need to Custom design not just protein structure, right because the structure is not going to get you anywhere You need to custom design a function in a protein And this seems easy just take a protein and if you would like to create a more polar binding site or say more Hydrophobic binding site replace a couple of those polar residues with hydrophobic walls What might happen? Yes, because in general where do the hydrophobic then we talked about that through the entire first half of the course What do the hydrophobic residues want to do? Where do you expect to see hydrophobic residues in a protein in general? So they don't want to be on the outside. So if you it's easy, of course It's completely trivial for me to show a picture like that I imagine if there were some hydrophobic residues here It would be a great binding site the only problem is that nature doesn't agree Nature thinks it's a really stupid idea to put those hydrophobic residues on the surface of a protein So nature is going to fold a different protein You have your sequence that was easy to make the only problem is that you don't get that structure And I'm not sure did the did Lucy or Burke mention that what is the probability of a random sequence folding into a protein? any guess 50 50 1 in 10 1 in 100 1 in a million So what happened remember when I said that we talked about hydrophobic effect And if you take a mixture of hydrophobic and hydrophilic amino acid in a chain What will happen virtually instantly if you throw them in the water in solution? Well, not really full right, but you're gonna get this hydrophobic collapse to what we call multi globular or something But the point is that that's just a mix. That's just like mixing all the ingredients together. It's not just any cake So the question is how many of these will then go over to form some very specific protein structure I was about to say your guess is as good as mine. I don't think it is because I know a little bit more about this But take a wild shot you wish The likelihood is roughly say 1 in 10 to the power of 8 so think a billion rod So if you just randomly start with something up or replace residues by default things, I don't kind of fold There are some exceptions to this because this applies to the entire protein and nature Wants things to be a little bit resistant. So it's normally we don't like proteins where the entire stability is based on one rescue So you usually can't change one or a few residues But in general we can't custom design or randomly to change protein structures. There's no lower-energy proteins That is this is very much related to the priors though that there are some proteins cannot up to states Normally it doesn't happen But say one in ten million or one in a billion or something you might actually have proteins that can occur in two stable states And that's when we tend to get these strange things that have one native state and then one disease state that is also protein But it's very much the exception So the problem here is that we would it seems obvious that we would like to design proteins But it's an exceptionally difficult problem and it's also a problem that we can't just do in the lab in Theory you could of course do random mutations in the lab But again those random mutations means that you're gonna fail Well one time out of hundred million you're gonna be successful and those are not pretty good odds in particular when it's gonna cost you time And effort for every protein There are some solutions to that you can actually do directed evolution in bacteria So we can pretty much try to randomly with radioactivity or something Enforce if you have some sort of biological Process or surrounding that will mean that so that I would like to create proteins that are better at withstanding a high pH If I then create an environment that it's under artificially high pH for this bacteria, and then I randomly force them to do mutations Then I will go through hundreds of millions of changes per day and the bacteria that get good properties They will tend to survive more so we can cheat and again piggyback on evolution But trying to do this with custom design and amino acid by amino acid deciding ourselves in the lab What we didn't do is not gonna work There are some things that work though when I was your age There was nothing of the kind but the last 10 15 years has been a bit of a revolution And I'm gonna go back roughly 10 years to start because this was one of the really cool examples this is partly based on Research and well mostly based on research of membrane proteins and I'm gonna go back to some things one of my students did 10 years ago This is not the design but this is an example of a very simple membrane and Model membrane and a single helix in that and the reason we perform these studies is that at the time We were very interested in understanding membrane protein insertion. Why do membrane proteins insert? Well, we know that we normally don't want hydrophilic helix Sorry, hydrophilic amino acids in a membrane protein, but under some conditions we could have them anyway So we wanted to use she wanted to use simulations to understand the properties of this hydrophilic amino acids and membrane and what happened And I would just saw here initially on this is a helix with a lysine in it And normally a lysine that's even charged We would never expect to see that on the membrane But the reason why we could add that in the membrane is that the lysine will in practice take out of the membrane and form It's not even hide since it's charged. We don't call them hydrogen bonds, but salt bridges with charge residues outside the membrane We can do this in a slightly more advanced fashion if you look if you take two lysis and put them almost in the middle of the helix Again just looking at the sequence I would never say that that could go in the membrane the reason why that will work though Is that you see that they stick out and interact with other parts This was part of a much larger project where we really went after biogenesis designed predictors for what things would insert I can as a small parenthesis you might remember Lucy's lecture about voltage gated iron channels One of the reasons of this at the voltage gated channels They contain a couple of charged residues and we were Everyone in the field at the time were so perplexed how these charged residues could insert and that's where they started Charge residues are exceptionally rare in my brain protein. So but in some conditions Interestingly enough we can find not charged but polar residues serine and three and in and if you simulate if you put that in a computer simulation, it turns out that what they do This OH group would normally hate to interact with the lipid environment But they get by by kind of sharing a bit of a hydrogen bond to the backbone here So they are certainly not happy in the membrane, but it is possible to insert So that was a long parenthesis The interesting thing is that to understand why do these occur in nature sometimes? Why would nature want to insert something slightly polar in the membrane? Because normally we would expect that the more non polar the more hydrophobic the healers are the better They would be a membrane proteins and they are better at the insertion But there are some things that we can do with protein design while having charts was This goes back to another very early discovery by Don Engelman at Yale Don, I think that Don is a candidate for Nobel Prize maybe together with Gunnar von Heine in a few years But what they discovered very early on and worked a long time on is a very simple healers It's a single helix at the time But this helix has a very typical pattern of amino acids They have a glycine and then you have three residues that can be pretty much anything That's why we call them X and then you have to have another glycine and they're also glycine Three X's and then glycine. So you call them sometimes you can call them G X three imparenthesis and G And these healers is are important because they can dimerize Exactly what they do. It's not relevant right now, but it's this is just a toy system But then what they find out is that the importance of this pattern is that you remember what a glycine looks like What is so particular with a glycine in terms of amino acids? What is the most important property of the side chain of the glycine? Yes, it's even so small that it doesn't exist. It's just a hydrogen The glycine doesn't have a side chain This is important actually when it comes to the X and you should I don't expect you to know all these amino acids and draw them by heart You need to understand why glycine is special why proline is special and then a couple of general properties So if you have two healers then we'd like to put them next to each other The other thing that's important might help to have something small, right? Because if you want two things to get close is good not have anything between how many amino acids are there per turn in a helix in an alpha helix? 3.6 That's not an even number, but three is close enough, right? So those two glycines Separated by you have a glycine one two three four. They're gonna be almost on top of each other So what this does is that it defectively creates a depression on the side of the helix. So let's see. I'm lousy at drawing So let's see if I have one helix here You essentially get something like that, right? If these are and here we might have lots of side chains and everything, but there is something here that does not stick out as much Because there are no side chains, right for those two glycines, sorry Imagine now if we take another helix and this is where my drawing is gonna screw thing up that looks roughly like this Then you're gonna have if you if I want my two hands here to cross Right at the crossing point. It's nice not to have extra stuff, right? And if there is no extra stuff at the crossing point, this will enable them to get very close together So in theory this should enable us to create some very stable and nice small interfaces, but there is one problem Why would this helices want to stick together? We want the helices to stick together because that might be very useful for professors in biophysics But if this is in a membrane the helices are the left helix is entirely hydrophobic and the right helix is entirely hydrophobic What is your guess what they would do if you think what about what you've learned in this course entropy in particular? So what will they interact with in a membrane? Yes, and the lipids are Yes, and the lipids are hydro and the lipids are also hydrophobic, right? So lipids are hydrophobic and helices are hydrophobic So if you're a residue, that's hydrophobic. What is it best to interact with hydrophobic or hydrophobic? It's the same So if you now have two helices and there are no significant differences in interaction What is better to stay together or be separate? Yes, you have higher entropy that's ever said putting them together means that we lose entropy. That's bad that will never happen unless there is something else that is advantageous by being together and There goes the plan of the biophysics professor to create a beautiful drug here They're not gonna even if it's theoretically possible for them that it's sterically possible for them to get together There is nothing that prevents them from doing it, but there is no driving force whatsoever. Why should they? Yeah, it's a great idea, but then we have to go back to the traditional type of drug design I can't design these helices to interact but one thing that Don both noticed. They're very rare mutations and Then engineered. What if you have something like this? Is that serine happy there? Why? Well polar, right as your partial charges. It is that free and unhappy there. No same reason, right? It's polar But what if you took that serine helix on the left turned it around 180 degrees? Put it right next to the three in it. So then and if you think about this in terms of free energy, what would happen and why? Or what do we guess might happen? Yeah, so it's a good or bad from an entropy point of view. This is not trivial But that's why I'm asking you the questions And this is why protein design. Okay, we pretty much think about the simplest possible design case We can imagine right and already here we're sampling it, but let's Whenever I ask anybody ask you about something about free energy, what do you do? What is the first thing you do? Stop hand-waving. There is an equation. What is the equation and I would even I Would even strongly advise you don't when we talk about changes. Don't think if talk about Delta F so Delta F Equals Delta E minus T Delta S and that might seem obvious again But in the Delta there means that you have to think about what is the before and what is the after? So the before is when we have to illicit that do not interact the after some way have to illicit that do interact And what I'm essentially asking you what is Delta F? Is it negative? That's good or positive bad And there are only two way there are only two components of that There is a Delta E and Delta S. So let's look at those in this case. What is the Delta E? What could two polar residues do in particular with that little atom involved? What type it could create a hydrogen bond and is that good or bad? Yes, in particular actually, it's that two is not entirely obvious because you could or you have a bit of hydrogen bond here But this is not really a real hydrogen bond. It's sharing it and everything and that having your own hydrogen bond is much better than sharing It's like an apartment. So that the Delta is going to be negative if they interact the Delta S Yeah So that so this week in general we can't say here, right? This will depend a little bit on how much what is the entropy drop for the two helices Versus how much do we gain from the hydrogen bond and in most of these cases the hydrogen bond is more important than the small freedom The helices would have So what happens here is that the reason why we see these serine and three-in-a-s in nature It actually creates a beautiful driving force for nature to form helix interactions So that you can insert the individual helix is not good but we can insert that serine if there's lots of other hydrophobic residues and The second they are in the membrane This is going to create a nice downhill path that you encourage the helices to interact because then they can form hydrogen bonds One of the original reasons we had this in the course is that well This was discovered some 20 years ago people were super excited to be included and I've written my fair share of research grants Where I argued that we should be able to understand membrane protein formation this way But what Lucy might have told you in the membrane protein lecture is that membrane proteins are difficult because they're hydrophobic everywhere It's not easy to predict them So as good hopes we had about this and we hope that this was just the first pattern This is why we are very likely why Lucy didn't bring it up This was just the first pattern and at the time we hope we're gonna find another hundred patterns And we found all those patterns We will be able to use bioinformatics to predict how helices will interact the only problem is that we haven't found any More patterns since it was I wouldn't say an exception We learned lots of things about how helices interact, but it was not the answer to membrane protein structure prediction But the interesting thing here is that just as nature can somehow engineer what things should interact Maybe we too can try to engineer how things will interact So Lucy did talk about the receptor protein tyrosine arty case Tyrosine kinase receptors, right? I'm not gonna go through all their normal function But these are membrane proteins that usually they frequently just have a single helix to the membrane and then they work by under some Conditions you have a ligand that binds and ligand causes this dimers to come together And when they come together good things happen in the cell Well, you usually get signals here that leads to cell growth and signaling on the inside and this Continues to signal on the inside until they eventually spontaneously dissociate again Unfortunately, and this this works to your bodies all the time. Unfortunately, there are some examples where this goes wrong And what happens is you have the normal signaling the ligand binds and You have the functionally active dimer that starts us in signaling and then things should dissociate But unfortunately, there are some examples even with single residue mutants in the silas and then the helices stick together too much And the problem is that if this helices stick together too much It doesn't help if the ligand dissociates and everything if they stick together. They will continue to signal cell growth divide divide divide And you can probably guess what that leads to right very severe tumors And we've even identified we know what this mutation is I don't remember the time either it's an alanine that's an intervaline or vice versa very simple mutation And it's just a single transmembrane helix It doesn't get much more simple than that and it's virtually impossible for us to get anything we can treat it with So the treatment would historically try to be prevent cell growth or something But what we would of course like to do couldn't we have a toolbox that we selectively go in and disrupt this bad interaction between the helices So imagine that we could create some sort of superhero red helix And intercepting peptide so this intercepting peptide I need to design this red peptide So it should have super strong binding affinity to the blue bad helix But it should ideally not bind the green good helices So I should be able to turn off I should be able to bind to the blue helix and block this interaction Because if the red is already bound to the blue this other blue will not be able to bind it right and at least this Is some sort of science fiction manner. I should be able to turn this off So there was a team led by builder garter that actually did this some 12 years ago. I think it was So their idea was literally to use these results of the simple glycophorin dimers and see if we can design Arbitrary peptide can we design a custom create an interaction between peptides? So my peptide would bind say strongly to your peptide, but not your peptide And in principle, it's super easy, right? You just need to find something that to Disable this interaction or something else or create an interaction here so that the two pairs of the native or disease Helices don't interact So in this case, they needed slightly more freedom and they didn't start from this glycophorin helices But what what they started from is that they created a database of They looked at all the helices that are present in the protein data back. There are just hundred thousand proteins Not that many and then they cluster these so that you know, there might be 20 large clusters and Yes, there are probably a lots of small clusters to but in general Let's let's not try to fool nature If it's if there are 10 or 20 very common Conformations of two helices when they stick together. That's likely because that's a favorable good way to pack two helices, right? And if we don't think that we're smarter than nature, let's copy nature so what they did is that they took these 20 largest clusters and Then if we're gonna have two helices interacting these are completely different helices But if I would now like to create a custom interaction say between the red and the green helix here in all likelihood They will have to pack roughly the same way our other helices in nature pack so that they just took the sequences and then they put the sequences in helices and then they tried each of these 20 different confirmations and Then they let computers chew over this for a few weeks or something that can you just try to For each confirmation of two helices that I give you Try to pack the side chains Try to pack the side chains try to pack the side chains pack the side chasing the best possible way you can do and Not only that then we also gonna need to try different side chains right because if I'm gonna create the custom red helix here I can choose the amino acid composition But then I also need to bias myself this has to be something that will go into a membrane So it has to be relatively hydrophobic And then I might be able to insert a few hydrophilic residues like the ones we had in the GX3G and To make a very long story short this worked So they could create helices that they first show that they could pack Well, they form helices which is not obvious a random amino acid sequence will not form a helix But as you showed in the first lab that is something we're fairly good at predicting What they then also showed is that they could create helices that had affinity So that they could show that the helices that we now design they will Well, first you can show that they will bind very well just in a computer. They will have a large buried surface areas They will pack nicely together But they also that showed that if you now put this in a an assay and go and test it with cells it appears to have some sort of effect and That the very final ad they could actually show that if they now introduce this in cells and platelets Actually, they can create very specific effects and block out just the inactive compounds here So this actually worked and I think they they sold the results of pharmaceutical company I'm not sure I haven't followed this the last five ten years to see whether anything has come out of it It might not remember what at the logic most things will fail But I bet that there are a whole lot of other studies trying to go in this direction and Even if it does succeed it might be if if this turn into successful drugs is probably about now that we would see it ten years later or so This was merely the first example It doesn't really get a whole lot simpler than a single helix, but you see the difference with the driving thing We're not here. We're not trying to screen any type of pressure We're not trying to screen a database with a billion compounds We're custom designing one compound to fit a bad pattern that we would like to turn off and it kind of works Here too though, I would expect if this is going to be used as drug you might need to Refine it further and everything. It's also it sounds very easy to just shut off the process The only problem is that if you if you have cells dividing too much I want to reduce it a bit We might not want to completely shut off all cell division in your body because that will likely mean you die So there are there are a billion more things that could go wrong But the point is that there are ways that we can start to custom interact with process in your cells By designing proteins all the way in the computer The cool thing with this is that they're way more potent If you think with a drug We're normally happy if we can get it down to say an on a molar or something That means that the lower the concentration That's the concentration We have 50% of it bound and the lower this concentration is the more potent the more activities Proteins they can easily be a million times more efficient So that extremely small doses They will have super high specificity because again, this is based on protein folding You need a very specific packing of all the sidechains or you're not going to attract at all So this peptide that I showed you it's likely not going to interact with any other protein whatsoever in your body There are no side effects in theory This is starting to be pretty interesting if you're running a pharma company, right? Very potent higher specificity This specificity means you're hardly gonna have any side effects We can custom design it for pretty much any disease. We don't need to rely on anything. That's already out there in nature This is where CEO starts to see lots of dollar signs There are a few problems this requires injection just like incident why There's something before the blood. That's your blood transport proteins. So what's the color of your blood? Red why is it? Yes, it's red because of the hemoglobin binding irons. There are lots of protein in your blood There's something before that it gets into the blood Sorry So what happens when you take a drug? Phil what happens when you take it? Where does it go? Yes? So let's say what happens if you have a if you have a steak for lunch Where does it go in your stomach? And what does your stomach do with the steak? Yes Muscles that's protein right a steak is protein your stomach degrades proteins that if you're now making drugs that are proteins They're gonna be degraded by your stomach. That's what your the enzymes in your stomach are for So we're gonna need to bypass that and the obvious way to bypass it is that you need to inject it directly in the blood and Again, if you're curing a cancer, you're probably happy for that, but you're not you don't want two injections per day for a lifestyle disease They might become very expensive to synthesize because we now need to produce the drug in bacteria on the other hand given the cost of normal drugs That's usually not too much of an issue. The other problem is that they might clear very fast through the body Just the way you need to take normally a pill you need to take one per Well, maybe two or three pills per day right here, too You're gonna need two or three injections per day because you're not just because it's a protein your body doesn't produce the protein It's an external protein that we're injecting. So there are a bunch of problems There is still quite a lot of money to be made though if we're slightly like sexy than those membrane proteins Do you know about the company Novo Nordisk? It's the largest company in Scandinavia What do they do? They specialize in diabetes care It was the largest one of the first companies to start to produce insulin and it has a stock market See the market capitalization is on like 900 billion Swedish kronor far larger than Ericsson already And it's probably a quarter of Denmark's exports Insulin is a complicated drug because in theory it's easy to make but it's also It's a drug that if you're depending on your blood sugar level It takes a long to can take even take hours or something for normal insulin to start acting So you would like drugs that that's very easy to control in particular these modern pumps or something So that if you detect that you would like to adjust your blood sugar level You would like to be able to do it with a course of minutes instead of hours So there are some peptides or something that in principle leads to insulin release that could be used to do this The GLP one is one of them, but it has a half life of roughly 30 minutes in the bloodstream. So it's not gonna work So what nobody's developed a few years ago, they developed a Protein called the Lyra glutide. It's as a market name. Let's see what that is big toads. I haven't heard of the market name Which is a the same protein, but you've used a fatty acid chain to this lysine here and it seems completely arbitrary And suddenly you get a half line of about 10 to 15 hours Which means that you just need to inject it once per day and it will slowly release and the reason for that Is that it will basically binds to fat tissues or something and they will slowly have a little bit of it in the bloodstream Far nicer you get a slow low continuous release and you don't need to take it all the time Magnus used to give this course the reason why I have this slide is actually part of his work So what they've been working a lot on is trying to computationally optimizes first understand why this happens adjust the size of these chains and also understand Why well first why does this happen and can you try to improve these properties? and The reason for the prolonged life to see it binds to albumin in blood and the other part is that these fatty chains They tend to interact so you form these hip tamirs with seven units in them And this is actually good because as long as they were hip tamir They're not gonna interact and then they slowly dissociate and you get the nice low continuous dose where you gradually build up The concentration of insulin in the blood blood rather than having a large instantane. It's so called bolus dose So that in principle regulating your incident level is not that hard But the hard thing is doing it in a way that will make your body survive for 80 years or something so that we don't stress the body And that's usually what leads to the side long-term side effects and diabetes care So here too a whole lot of this is based on computationally trying to optimize the interactions this proteins There was another example a few years ago not by Norbert Nordisk, but Eli Lilly Normal insulin would form hexamers and they wanted to disrupt this to get something that could act slightly faster So they realized that as part of those interactions there was one lysine I'm sorry one proline and one lysine that were interacting and they were if they just swapped the order of those They would kind of destroy the binding interface between the molecules a bit So then they created something called list pro li spro just have a market name for it And that was like 10 times faster release. So again very very small Changes in the proteins that are to tell the truth Probably just based on sitting and looking very carefully at these molecules in a molecular viewer and then coming up with ideas They probably tried 500 other things before they found this one And that's of course not the long-term sustainable view But the point is that we can there are some pretty darn cool things we can do by trying to change protein structure So we can get small changes if we can get some improvements by doing small changes Why don't we do large changes? It seems stupid Why why would you just swap the lysine and proline in the residues couldn't you start to change 20 residues? And then we could get something that maybe has a hundred fold improvement So what did I say when we started about the design the problem is that if you change too much what happens? You no longer have a protein, right? It's not going to fold so that we have to be very very conservative and do small changes This got super hot some eight seven years ago with a number of companies buying other companies doing this type of design I don't know is and you might have heard something in particular when it comes to designing antibodies It's still super hot. Have you heard about the company called Teginero? So this was a Super map they call it a special class of very very very special antibody and that they want to design antibodies to target lymphocytic leukemia And the particular drug actually wasn't drug a drug candidate called TgN 1412 drugs frequent drug candidates frequently just have names in the Companies and this was the CD 28 is the name Relative to the receptors so that the idea here is that we would call this to bind It should activate the CD 28 receptors and T cells and the T cells that then they They kickstart the entire machinery with killer cells and everything and then you would have basically find a way to Jump start the immune defense so that your immune defense would then take care of the cancer cells That sounds like a pretty neat thing. We could accomplish it, right? This one exceptionally well all the preclinical studies were awesome a beautiful binding interfaces you tried it in mice Perfect results it really you really could get the immune system to take care of the cancer cells and Then you went to phase one where it was a complete disaster. So what is phase one a bit of repetition here? What do we do in phase one studies? So when I say disaster, it doesn't sound so bad for the chemistry experiment going wrong Happens all the time in the library Human trials and not just any type of humans Healthy humans you want to make sure that nothing bad happens with you're not trying to treat anything yet You just everything worked really well in the lab So you find some healthy test subjects like students and Give them $50 to try a drug that has it worked really well in the lab There shouldn't be any problem and just to be safe We'll give you a 10 times lower dose per kilo than the mice had it shouldn't be any danger whatsoever They had I think was 10 or 12 students come in and they administer very small amounts within 15 30 minutes They're held hands all the extremities has fallen to twice the size. They were very close to dying I actually think all of them survived, but they had to amputate a whole lot of their extremities and everything You can imagine the company kind of been bankrupt after that they pulled it And we don't know exactly what happened, but there are there is no speculation so One problem. This is what we call a super antagonist. So it's it's not remember that I said we spoke about agonists or as a Super sorry not antagonist super agonist. We spoke about agonists and this is super agonist it's somehow a 10,000 times stronger effect than the normal agonist and The point is that we derived this for humans because obviously that's our long-term goal and And human and mice they share some like 93 percent sequence, which is good But 93 percent is not a hundred percent and in the extracellular part where this is going to bind It's something like two-thirds of the sequence that's shared So the question is how much antibody should I now administer if I administer this in a mouse It's not going to fit perfectly. So I have to I keep increasing the dose a little bit Until I find a level where I activate the mouse the mice is the mice immune system just enough and then it works well Then we take this drug and inject it into you. The only problem is that this drug now fits one hundred point zero percent Your entire immune system is going to be completely berserk So the problem is that we've effectively administered a dose that's ten thousand times too high for you This is never going to release it's it's literally they're made for each other We we designed something that would fit your scd-28 receptors perfectly And what now happens is that when your immune system goes berserk Well, you're activating your immune system to ten thousand percent to that's like giving somebody a Kalashnikov They basically the immune system will try to shoot all over the place So the immune system is not trying to kill everything at campine including your own healthy body Likely we don't know So that would likely happen is that what what we thought went so well Well, what they thought went so well in the test system was likely only a very weak partial activation But when you suddenly got when you administer it in human that it was designed for you got a way too strong activation After those trials this wasn't exactly a matter of say, okay, let's do it again So that I think the company went back up they try they cancelled the entire trials, of course That's actually less than ten years But the point is that this sounds horrible that the approach was not bad. It's super interesting, right? Then it's difficult. It's exceptionally difficult and dangerous But this is opening the door to completely computational design of drugs We have no limitations whatsoever sticking to what we know and I think in particular in cancer This is going to be a revolution within the next decade We can do other things too. I'm gonna spend that I have ten more slides. I will take five minutes more minutes or so Lucy talked about voltage gate that I in channels. I think these are involved in lots of normal behavior But there are also some diseases there. You can have a lot of heart related diseases, but also epilepsy and Some ten years ago people realized that there are certain cases of epilepsy that you can treat with poorly unsaturated fatty acids So it's basically fat with lots of double bonds and if you administer enough of these fatty acids all the symptoms disappear So this is just certain types of epilepsy in particular among children and whether it's based on genetic disorders And if you understand a little bit about these channels fatty acids are interesting in that they have negative charges and as Lucy Hopefully mentioned that a lot of these voltage gated channel have lots of positive charges So maybe just maybe they are interacting with each other So what colleagues of us down linshipping and down for they've worked in these channels in the lab for several years They have this idea that may be a holdout of this disorders They are related to the new tense in these channels and what might actually happen is that you might have channels that they kind of work But they don't open as much as they should and If we now insert lots of fatty acids with a negative charge here that negative charge up here in the head group Is might help pull the protein up a bit and help open it The problem with the fatty acid you need to eat very large amounts of it and it's not very healthy long-term But this is an example where people accept the side effects because the epilepsy is far worse But what if we could design custom drugs that had the same effect that say a thousand times stronger So in principle you just need to start looking at these fatty acids We had a post a few years who started to do this in the simulations. It turns out this one doesn't work at all It doesn't bind while this one is perfect And it's because they have slightly different properties the ones with a double bonds They're slightly more polar double bonds tends to be more polar So these will interact with the charged parts of the protein slightly easier and then of course you have the chart in the head group here While these normal fatty acids they behave pretty much like lipids And I wish that I could give you the final result of this But this is still work going on But what we have done is that they've we've been able to identify the specific residues in this vaulted cell Where they are interacting you can confirm this by now mutating this residues And if you show that if you mutate away those residues the fatty acids will no longer work with it And then Fredrick's team in parallel They've had an organic chemist now try to use docking exactly like you're going to do in handy task 3 So we've been using docking and trying to find interesting compounds and see what are the properties of those compounds that would make The bind here and then they've also been manufacturing these compounds in the lab And this went so well that they have a bunch of them that they have Requested patents for and then published and the idea there now is that we're going to try to turn this long at some point They might very well try to sell it to a company But see if they can actually get something on the market that would selectively treat those types of epilepsy Because it's like 10,000 times better affinity than the normal fatty acids how this will work that remains to be seen But the point is that there is a lot we can do just by looking at things and understanding how they interact 10 years ago This too would have been science fiction and I never thought that this would succeed when we started working on it The final example I'm gonna have that the drawback with all these things. It's still we're basing on simulations We're basing on structures that we know That's partly because we need to walk before we run But the really cool way to do this. What if you could start out with a completely blank sheet? And let's say that I would like to literally design proteins from scratch You have a you're giving me a particular receptor that we would need to combat I'm not going to use anything in the protein data bank and when Lucy spoke about bioinformatics We talked about sequence similarity, right? You talked that we can build models and other models and everything and you can think of some this blue part You can think of this as a sort of arbitrary space of protein sequences or structure and those yellow parts these are native produce produce that do exist and They exist all over this part of space and of course What we did before is a way I try to create a new protein that was just on top of some other protein The stuff that I mentioned about bacteria directed evolution It might be possible if I have a certain function that I would like to achieve I might be able to get lots of sequences to move in that direction And that's what you can call directed evolution that will likely be reasonably close to other proteins But maybe I can move those parts a bit to create some sort of property or a state that is good But the really big pie is there were all the blue parts. What if I could create something there? Not even close to any other proteins completely for us grads not build on a model or any protein But use all the freedom of creating proteins and In particular David Baker's group at University of Washington. They are insanely good at this So they are using computational methods combined bioinformatics and molecular simulation and models You're basically starting from something you're predicting what the entire overall shape of the protein should be you create amino acids That will hopefully create that shape and then you need to start to iteratively or repeatedly optimize your side chains and everything and place the side chains to realize Do we expect it to be a stable fold? Can I change the amino acids to make this fold even more stable? And then hopefully We should be able to find something that so that the lowest RMST to the target should really be the Sequence that has the lowest energy there, too And then we take that sequence on the lab and then we keep our breath and hope that it's going to fall to that sequence People have of course been trying that for 25 years and there's usually a very simple answer to that question And that answer is no it won't But what's happened the last ten years that the answer first started to be one in a hundred Well, maybe and then it was yes once in a hundred and today it's kind of like 50 50 So this is improving every year and this is a remarkably cool example from it from a competition that they have every few years So based on a crystal structure, they predicted. Sorry. There was another group determining a structure This is their prediction whether they did not though the structure. They just knew what the amino acid sequence was after the prediction People told them what the structure was and this is the crystal structure pretty insane. This is not an easy protein We're not talking just packing some helices, right? and There are two papers that I've uploaded to a canvas in particular one of these to look at it I'm not gonna ask you questions specifically about this paper That's why I'm not printing a copy for you, but they go through a whole range of examples What they can do they can create self-assembling materials. So if you create small parts of Beta sheets in particular you can get them to bind in small patterns And then you can get these patterns to assemble into larger patterns and these larger patterns You're essentially treating proteins like there were crystals or Lego building blocks Remember what I did so when we talked about the fibrous proteins your hair and everything, right? So this these aren't fibrous proteins. They're globular proteins But we can use them almost as if they were fibrous proteins to custom design biotechnology building blocks It's not used yet in biotech, but I bet it's only gonna be a matter of time before people use this to say Well create biofuel reactors or something because if we now create make these to turn them into a catalyst too We can get them to catalyze pretty much any reaction we want The other thing that they show in this paper is that they can create hyper stable peptides So these are tiny proteins and then they can engineer in some disulfide bridges So Why would a disulfide bridge make a protein hyper stable? It's a covalent bond, right? It's not just an interaction But you for me those disulfide bridges a bond that will never break But there are the proteins that have that what they also show is that you know if you create a protein Normally you would have the n-terminus there and the c-terminus there But if we design a protein where the n and c-terminus are closed each other under some conditions We can also then create a chemical reaction to fuse that again to close the loop So then you have a little you have a circular protein That's going to be even more stable. So now you said have three links within the protein and I'm not sure if you if you have a piece of yarn in it right on nested, right? If the yarn was circular, it's going to be pretty difficult to unnest it So where we where could you imagine using this? Based on what we talked the entire first part of the lecture today Why would you want a protein to be super stable? Exactly, right? So if you now can custom engineer a protein that's exceptionally stable so stable that would normally not unfold Then we have a protein that would likely be able to survive the digestive tract And then you can do custom-designed proteins that you can administer orally So you get rid of by far the largest problem with protein design There is one more thing that I might have mentioned at the very first lecture or two There's another strategy that people use to try to get through the digestive tract You're going to hear more of David. This might very well be a Nobel Prize for him in the future because there's a amazingly good There could be another way that people use Essentially what you want to create proteins that are bio incompatible Because if you're compatible with the biology in your body your body will digest the proteins So I would like something that is a protein, but I would like it to not be compatible with your proteins You remember something I talked about the very first lecture even or the second. Do you remember these things? So what was that? The handedness of the aminus is right So there's actually you can create that alpha helix that would be a mirror image and this mirror image will mean that it's all the all the Enzymes that would normally say break a peptide bond they recognize normal amino acids But if I have the opposite type of amino acids, it's not going to recognize it But still all the laws of physics apply so in principle they can form stable compounds and everything, but they will actually not be digested But making that type of amino acids is expensive hard and everything this is much more feasible and I said to work with I think that's all I had Both for this lecture and course contents And I don't think there's probably not a whole lot point of recording the other part here But what I figured that we're gonna spend as much time as you want here some questions and answers both for the handling task and the exam