Why building new proteins from scratch is our new superpower | David Baker

honorable laurates excellencies distinguished colleagues ladies and gentlemen welcome to the Nobel lectures in chemistry 2024 so this year's Nobel Prize in chemistry recognizes groundbreaking achievements in two different areas computational protein design and protein structure prediction so it's all about protein structures as you may know the polypeptide chain with protein folds itself up in space into complex three-dimensional structure now the remarkable thing is that this 3D structure is somehow encoded in the sequence of amino acids that make up the chain this finding that a linear sequence of amino acids dictates the 3D structure dates back to the work of Christian anenson who received the Nobel Prize in chemistry in 1972 so over 50 years ago it meant that it should in principle be possible to predict the 3D structure directly from the sequence likewise it should be possible to predict which sequences would produce a certain 3D structure so it has long been a dream of biochemists to computationally solve these two problems and it is clear that one needs to use computers to do it the latter challenge to predict sequence from structure that's what protein design is all about the idea is to draw a new protein typically one that has never been seen before in nature and somehow compute a sequence that will give this structure the sequence could then be made in the lab and completely new proteins with desired functions could thus be created so this challenge was solved by David Baker and co-workers in 2003 they designed an entirely new protein with 93 amino acids and computationally predicted a sequence that would give this structure they then made the protein and verified experimentally that the prediction was correct the reverse problem to predict 3D structure from sequence is an astronomically huge search problem and this was pointed out by Cyrus levental already in the in the late 1960s so progress in this field was slow indeed for many years but Demis hassabis and John jumper and the co-workers they sold it in year 2020 by ingenious neural network engineering which resulted in the now famous program Alpha fold 2 and it was published in 2021 this was undoubtedly the real breakthrough in protein structure prediction and Alpha 2 gave unprecedented accuracy so this network is trained using the data stored in sequence and structured databases to discover correlations and patterns among amino acid sequences and this allows it to produce accurate distance maps that can then be converted into 3D structures we'll probably hear more about it from the laurates so the work of this year's laurates has had an enormous impact David Baker has opened up completely new possibilities to design proteins that have never been seen before with many exciting applications in Pharmacy medicine non science Etc and likewise the work of Demis hassabis and John jumper has caused a revolution in structural biochemistry and we now have access to predicted structures of all 200 million known proteins and this is a phenomenal resource for biochemical and biological research also with many new applications so our first Speaker David Baker he was born 1962 in Seattle Washington United States he got his PhD in 1989 from University of California at Berkeley in California and he is presently professor at University of Washington Seattle and an investigator of Howard use Howard use medical institutes so please join me now in welcoming Professor David Baker for his Nobel lecture all right well well first of all thanks everyone for being here it's really wonderful it's a great honor and it's wonderful to see so many friends and and uh um family and colleagues so today I'm going to tell you about denovo protein design um this is an artist depiction of the interior of um uh cell in our bodies and each of these colorful blobs is a um is a protein and each protein is carrying out um a a a unique job or function and indeed in in in all living things proteins are carrying out they're basically the miniature machines that are carrying out all the important um important functions and they um they evolved over hundreds of millions of years or billions of years to be really perfect at solving the problems that arose during biological evolution um but today we Face new problems there um you know we live longer so disease neurod degenerative disease and other problems now are Rising we heating up and polluting the planet uh and so if we had um so if there was selective pressure over a really long time new proteins might evolve to solve these problems but it would take it would take a very very long time and and not be very pleasant um so uh the the question I'm going to the goal of denovo protein design is to make new proteins that can solve modern-day problems as well as the proteins um that came through Evolution solve the problem um that were relevant during Evolution so first I need to remind you of a little bit of biology and you you already heard this B I'll just give you the basics again the genes in our genomes oh each gene encodes one protein uh it does so by specifying the amino acid sequence of the protein which then folds up to uh the protein with a um which has a particular structure and biological function so in denovo protein design the flow of information is just the opposite we start with a desired function that that doesn't exist we build a protein which um uh we predict will have this function and uh then um uh uh uh compute the amino acid sequence this protein should have and then this is a brand new protein so there's no Gene in nature which already encodes it instead we make a synthetic Gene a synthetic piece of DNA and uh we that encodes this protein we put it into bacteria and the bacteria basically become factories for making the protein we can get the protein out and then test to see whether it actually does what we designed it to do and so today I'm going to be giving you a number of different examples of designed proteins that we've designed to have different functions and in every case we've gone through this computation and then made a synthetic Gene and then made the protein and I'll be showing you what those proteins do um let's see uh so there is a huge space of possible proteins and so there there are 20 different amino acids and a typical protein um is um a sequence of about 100 or more amino acids and so that means the number of possibilities is 20 * 20 * 20 100 times or 300 times depending on how long the protein is and this is an astronomical number and it's kind of depicted by this gray uh gray space here now what's been sampled during evolution is a tiny tiny fraction of the possibilities and um the proteins in nature are kind of indicated by these uh red uh red ovals because proteins nature Pro Evolution proceeds incrementally so proteins tend to be very closely related you know the proteins and us are closely related to the proteins and other mammals and so forth so it kind of bunches up a lot so traditionally and this is really up until the Pres present if um if an engineer or a biologist wanted to solve a new problem and they thought a new protein could could work the way it was done or and still being done is to look all through nature to see if a protein can be found that has that activity and that's called bioprospecting the other thing that's often done is to take a naturally curring protein that does something kind of similar to what you want to do and make small changes to it and um that uh and so you get a protein that's very close to a naturally curring protein and you've hopefully modifi it to do a little bit closer to what you want it to do but what I'm going to talk about today is different I'm going to talk about building proteins completely from scratch that carry out new functions and the advantage of this is you can build the protein do exactly what you want to do you don't have to worry about what the the closest that um uh that nature got or try and modify something that's already very complicated uh so I'm going to begin with a very brief history um in um the the first stage in um in denova protein design was understanding how to build new protein structures from from scratch um and this really culminated in uh 2003 with the work of Brian Coleman designing a brand new protein which had a new structure and a new function and since that time uh a growing and and now quite large community of of people um really wonderful scientists um um a number of whom are quite a few of whom are here um have been um uh working to um both design new protein structures but but also to design new protein functions and in particular to design proteins that that solve current day problems um and in in very recent years um uh this this field has sped up with the um with the uh Advent and you'll hear a lot more about this um emesis talk of of deep learning methods so I'm not going to tell you much more history in my talk I'm going to basically sort of fast forward to the present I'm going to tell you about the Deep learning methods that um we have been developing for protein design and then I'll give you examples of of what protein design can now do um so the um uh so we've been developing uh generative methods for creating new proteins and U these methods or the one I'm going to tell you about called RF diffusion is really inspired by um methods for generating images and these methods um are um uh work by taking a large set of data for example images off the internet um adding different amounts of noise to the images and then training a network to remove the remove the noise and once once this has been done if it's been done well one can start with completely random uh pixelated sets of pixels successively denoise and generate an image that looks like it's image off the internet but in factly in fact it's a completely new um image um and in the same way we took all the structures in the protein structure data bank and I should I have need to highlight that everything I'm telling you about from now on absolutely relied on the work of the tens of thousands of re Searchers who solved the the protein structures in the pdb because that's the data that we use to train and um the also the um the people organized the protein Data Bank it was really an absolutely invaluable resource so we took all the structures in the database we added um increasing amounts of noise and then we trained a network to REM to predict what the uh D noise structure at each stage was so at the end um just as in the case of images we could start with a random uh completely random image and generate something that looked like a very plausible image we could start with completely random uh configurations of amino acids progressively remove the noise and generate a new protein structure um and we found that these proteins folded uh up pretty well we could make them very well we could make them in the lab um but they didn't do anything now in the case of images as you all know you can uh if you if you you can guide or you can condition the uh generation process with for example a text prompt so you can say generate an image of a cat sitting on the table and that's what you get and we can do exactly the same thing here because it's not very interesting just to make new structures what we want to do is make new functions so we can condition the this this protein generation process on the function that we want to make and one way to do that is shown here this is the insulin receptor we can carry out this uh generative diffusion process in the presence of a Target like the insin receptor and uh the during training uh the the generative method has learned that um uh that that proteins bind to other proteins through very shape compliment interfaces and you can see that as this protein gets build up how the shape match of this new protein matches um the Target and we've now used this to uh design proteins to bind to well over 200 different protein targets it's a pretty much a solved problem now I'm not going to tell you about all to 100 of them but I'm going to give you some examples of how one can use these binding proteins uh to solve um problems in a wide range of different areas and the I'm going to tell you about three types of applications today in medicine in technology and in sustainability and I'm going to begin by um uh by talking about applications in medicine so um uh snake poisoning is still a major medical problem particularly in developing countries and snake toxins uh interfere with really fundamental biological processes um now to uh an antidote for snake venom needs to be very it needs to be a very stable protein or set of proteins it has to be cheap to manufacture because where when where they're needed is in countries where there isn't necessarily a cold chain you need them out in the field you need to be administer them almost immediately and we thought this would be a good match for protein design because the proteins that um uh that we design almost always are are it's are very very stable proteins so you can produce them in large amounts quite cheaply um and so here is a designed protein which um is binding to the toxin which is here and um um all this work is highly collaborative um and so our collaborators in in this case um tested to see whether this designed protein could prot protect animals against the snake venom and so in this experiment here the animals who got the um the snake venom alone uh died almost immediately whereas ones which got the designed protein uh were completely protected and so we're very excited about this as a as a general Way Forward for um generating antivenoms that can be more more um more effective and much cheaper than in current approaches now a major uh problem in current medicine is uh is inflammation which is um associated with many autoimmune and other disorders and uh Central to um uh Auto to um inflammation is a protein called the tnf receptor and um it's actually the target of of many um of many drugs currently on the market and uh so here you see um a diffusion trajectory um uh conditioned on the structure of this target you can see how it's building a protein which is really exquisitely uh shape complimentary to the um to the Target and um this protein is uh very effective at blocking inflammation and animals so these are animals that have been that have been exposed to U an inflammatory um treatment with or inflammatory signal from bacteria um they get there's a there's a very large amount of inflammation current drugs like enil that are used to treat inflammation have some effect um this designed protein because it binds so tightly uh is even more effective and so we're very very uh excited now about making drugs to treat a wide range of autoimmune diseases and inflammatory diseases um now cancer of course remains a a a pressing problem and we've been here we've been designing proteins to activate the immune system to um uh to treat cancers and so this protein here this designed protein is bringing together two immune receptors and this results in um a very strong um um activation of the immune system and again uh with collaboration collaborators um Mike and Stephanie Dugan um uh they have been looking at pancreatic cancer which which is one of the most difficult uh forms of cancer to treat currently and uh what they find is that uh this design protein is quite effective more effective than um than other known treatments at in reducing the size of tumors in this model and again we're very excited now about getting this protein out in the world for um you know into clinical development um so uh another important application of design proteins is to make antivirals for pandemic preparedness so this is the uh influenza virus surface protein and here you see the diffusion process generating a binder here but here we've added an additional conditioning um Criterion which is we want this protein to be an antibody which is a particular type of of of protein fold and we wanted to do that because antibodies the pharmaceutical industry knows how to develop antibodies and so they like to work with them and so here you see um an antibod like structure being generated that binds to the um the the the Target and here I'm showing you an experimentally determined structure of this design protein B bound to the um uh to the flu virus protein and you can see that um in one of these Colors Let's see the the the the experimentally determined structure is in purple the um the design structure is in Gray and you can see they're almost identical and that um those of you know about antibodies know that they recognize their targets through um CDR loops and these these Loops which are being built completely from scratch in addition to the docking mode are coming about almost exactly in the designed model so I I think that you know antibodies are currently discovered using um by immunizing animals or pulling them out of people I think we're now seeing a transition over to using uh these kind of design methods to generate antibodies so another sort of um increasingly important uh medical problem are neurodegenerative diseases like Alzheimer's disease which are associated with the formation of long amalo fibral and there are several proteins um among them uh amalo beta or a beta and to which form these fibral that are observed in um the brains of of people with neurod degenerative disease um so here what we've been doing doing is designing proteins that will bind to the disordered portions of these proteins that would normally interact with themselves to form Amid and these trap these these regions so they can't form this extended ameloid and this is this experiment again with uh collaborators is looking at the um if you simply take these proteins put them into solution they aggregate as the amalo is formed and with these um trap proteins uh this um uh this amalo formation is completely blocked so this is measuring the extent of of of amalo or aggregation that's going on and so we're very excited about these as a possible way forward with neur degenerative disease so there are many many other applications we're excited about now now in medicine but I'm going to switch over and now talk about technology so um one of the um one General problem in technology and is in sensing how do we sense um arbitrary molecules and how do we determine sequences of things like DNA and proteins and in nature there are are channel-like proteins or Poe forming proteins that look something like this and people have adapted those for problems like sensing and sequencing but they aren't really optimal for that because they didn't evolve for it and so uh with anastagia vb's group we've been designing developing methods for Designing new pores completely from scratch and this has the advantage that we can completely customize all properties of them of the pore and so forth to be optimal for um uh for the task we want to carry out and this just shows you some examples here this these are three designs with increasing Poe sizes and the larger the poor the more ions that can flow through it and so the larger the current that is measured uh going through these Now to turn these into sensors and we're very interested in constructing molecular noses what we do is to design proteins that bind small molecules that uh we want to um detect and we can put these proteins on top of the pore like shown here and then uh when we put this combined protein into um into a membrane and we measure the current across it we see a sizable current when the small molecule is added basically it closes up the channel and that's shown here um and then the current completely shuts off and this can be measured because one can measure electrical signals very very sensitively so we're excited now about being able to make very general molecular or electronic noses that can tell you what's in a solution um by um by integrating the uh information from many different pores like this um now what you'd really like to do for um for is is to couple this kind of sensing scheme with electron directly what I showed you before was um those were inserting into a hydrophobic lipid uh layer so we've been working on methods for inserting these sorts of pores into um uh into silicon nitride chips so they could ultimately be in for example in in your cell phone to tell you what's around in the environment um and so this requires a totally new class of um of pore which is um uh an example which shows here we want things there of course in proteins never evolve to inter inter interact with silic and nitrate or to insert into it so we have to design new types of protein structures and here are some uh examples of em structures of of sort of this gallery of of pores that we've made to trying to to approach this problem and again with uh uh collaborators um uh we have um this is from Yen goodlock Lab at the W so here we have a pore in a silicon nitride chip there's a lot of current going through um the designed protein pore can insert in and we get a very stable current and now we can start modifying these pores as I showed you with the in a moment ago uh by incorporating binding domains and other sorts of uh domains on the interior to be able to read out sequences and other properties so another General problem in technology is targeting very specific sequences of DNA so you know our our genomes have lots and lots of DNA in it and for there many applications for example fixing genetic uh diseases where you want to Target just a very very specific region of the DNA and so to to do this we've been designing proteins which bind to DNA and have very high sequence specificity so the protein shown here makes very specific interactions with the DNA bases that make it very sequence dependent and this is now some experimental data showing that if how if you take this protein and measure it binding to DNA um it's very sensitive to any changes of the the DNA sequence so when the um when the sequence is CH if the the red color means that the binding is weaker so it can this these sorts of designs can read out DNA sequence very um precisely and we're excited about using these for example to create new ways of smaller more compact ways of fixing uh genetic lesions and other um activities so now we have a very exciting uh Neil King had this very exciting result um last year with the first um approved red uh uh denovo design medicine and this is sort of an interesting case where um Neil started off making these um uh these self assembling nanoparticles sort of just to see if we could do it and then um uh and then after uh he had disc conceeded in building them uh he realized that he could make vaccines out of them by putting um uh pieces of viral proteins on them so during the pandemic he put um uh some of a part of the uh the receptor binding domain of the the um covid surface protein and on the on these and found that this solicited really really strong immune responses and this um is now a clinically approved medicine it's in use in um it's been approved in several countries here is the designed protein in um in a vial and we're very excited about you know a year or two from now being able to have many many examples like this um now uh these sorts of um uh uh nanop particles are very useful also for delivery but there's a problem in that you if you want to deliver a lot of material like say a drug to a specific site in um the body the size of the container matters and a bigger with a bigger container you can make more so um if you uh this is an interesting geometry problem the icosahedron or dedin is the largest completely regular polyhedrin so if you want to you can make it bigger um so this is U 12 pentagons coming together you can make it bigger by inserting hexagons between the PEX pentagons as shown here this is just as in a soccer ball but here you have a problem because the um the uh you can see the angle um at the inside the pentagons is different than the angle inside the hexagons so we've been trying to figure out how to solve this problem and they're basically two solutions I'm going to show you here these are both electron microscopy structures of designs in this case um we've solved the problem by um using just one type of protein that can adopt different uh confirmations in the different uh sites in this um uh in this nanoparticle allowing the interspersal pentagons and hexagons as you can see here in this case uh we've used um different protein chains that sit at the different uh non geometrically equivalent positions and the advantage of this is that both of these procedure methods um and again this is experimental data crym structures um give um much larger particles than um um one gets this is um would get with pure symmetry uh this for example is uh one of the most widely used current delivery vehicles based on a naturally currying virus aav now everything I've told you about so far is kind of static um but in technology we'd like to be able to turn things on and turn things off and um so uh to do that um uh We've designed proteins that um undergo confirmational changes and have two different states so in without um on their own they fold up to one state but then when an affector is added they convert into another state as shown here and if we build now larger assemblies out of these like this triangular one here um in the presence of the affector these will now Bend and so we'll get um and now um we can effectively convert you can see when this becomes a right angle uh we effectively convert a triangle into a square and uh this works really nicely we chose to go for these simple geometric transitions because it's very easy to follow by mic electron microscopy so here's a design just like this one that was designed to be a triangle then the effector added and you see you get a square um and this is electron microscopy data you can clearly see that transition and here's another case where you have a square that's going to a pentagon when the affector is added and for the biochemist in the audience this is clearly allosteric control because this affector is not acting at the interface between these chains it's acting quite far away and working through by changing the confirmation um well that wasn't very particularly useful it was sort of like we that showed that we could build switches but where would you where might you want actually a switch and coming back to immunotherapy we might want to activate the immune system to treat cancer but then we might want to turn it off because overactivated immune system can be a problem so here I'm going to show you um an off switch here's a design protein that brings together two receptors activ ating the immune system as I showed you earlier but now we've built in um um a binding site for an an effector which turns the whole system off and here's a cartoon of how it works so um uh here's the uh the design bringing the two uh chains together this affector comes in it causes an alisic change which kicks off this other receptor subunit and uh this is again a wonderful collaboration this is actually going on on the surface of a Mamon cell you're seeing these two receptors are labeled with fluorescent dyes and they when we add the the drug initially the two receptors come together and then uh when the effector added they come apart as you saw here and so we can see The receptors uh coming apart when the affector is added and um this has a profound effect on self signaling so when we add the drug the the this protein it brings these two together we get an activation of signaling um and then when the affector is added we can shut it off very very rap rapidly and this is useful not only for shutting um off the drug when you no longer need it but for probing exactly what the effect of having these uh receptors together for different lengths of time is which has been kind of an outstanding question in immunology and the final area I'll tell you about is sustainability and one of the some of the most remarkable proteins in nature are um are enzymes and these carry out chemical reactions and the the way they do this is they have these very large complex structures and if you zoom in they have active sites where there are a few key amino acids that are mediating the chemistry and carrying it out on um a bound small molecule and so we can now design catalysts completely from scratch we do that in kind of the opposite way we start by um specifying the active site geometry what the side chains on where they should be and uh the B and um the the small molecule or transition state we want to act upon and then we use the diffusion process to build up a protein around that active site um and so this is kind of neat because we can explore in in nature and biochemists have been and enzymologists have been doing this for many many years you just have one thing in your enzyme and uh you can sort of tell stories about how it works but with this buildup process we can change the what's what the active site what the side chains are in the active site and what their geometries are and look at the effect on um catalysis and so we build up sites of increasing complexity and um uh and um what we find is as we uh as we adjust the geometry and make the sites more complicated we can get we um we get increased activity with a with a sort sort of half with with a partial site we get enzymes that carry out a single round of the reaction but then stop if we build up full active sites with sort of uh more extensive networks of interactions we get uh quite active enzymes and we can solve the structures of these by X-ray crystallography and then we can see um uh that um in that uh to check to see how accurate our designs are and they can be extraordinarily accurate this is now a superposition of the experimentally determined crystal structure and the um actual design and um uh the activities of these are are um much higher than had been found previously straight out of the computer this is that was for a complex multi-step reaction we can also use metal ions to carry out catalysis um and uh here um the uh uh this is actually starting with a a quantum chemistry calculation of the reaction pathway and we um we identify the transition state on this pathway and then we um we basically build a protein around the the metal and the The Binding site and um uh uh this is sort of a blowup of the site and um as I said in all cases I've shown you we are making synthetic genes encoding the designs putting them into bacteria and testing them here we tested 96 designs and we got a number of active designs the the most active of which was um had very very much higher activity than anything that had been made straight out of the computer before and this is starting to approach the activities of native enzymes for this type of reaction which have evolved for hundreds of millions of years so there are many problems um this is a bond breaking reaction and there are many problems um where this is important including plastic degradation so we're now building sight completely synthetic enzymes um to um to break down plastic and other um polymers um using exactly this type of active site um now one of the other remarkable things that proteins do is to harvest um solar energy and this is done of course with um uh with in the photosynthetic process which is very complex which involves very big proteins um but they have at their core um what's called the reaction Center which is um uh made up of two chlorophyll molecules so we've been again kind of building from the outside from the inside out so here we've designed a small protein that binds two chlorophyll molecules exactly as in the reaction Center and uh now we can start building this into larger assemblies um and so this shows taking that this protein Here and Now designing it into these cubic assemblies each one of these has two chlorophyll molecules and so so we can start building up completely artificial synthetic photosynthetic systems um and we can also now tune the um the absorbance profile to go cover wavelengths which are not well covered by natural photosynthesis and my final example again something that nature never tried to do um is in mineralization so in nature we have things like bone and tooth and shells uh where proteins are directing the mineralization of calcium carbonate calcium phosphate to make these really remarkable hybrid materials so we've been really fascinated by this and so we've been approaching it by designing proteins which have arrays of of of amino acids that we predict would drive the nucleation and the mineralization of inorganic compounds in this example though we chose not to apply this to calcium carbonate or calcium phosphate but to zinc oxide which is a semiconductor because you can imagine if you could template semiconductor growth with proteins you could that would open up a whole range of interesting materials so here we've we've lined this part of this structure with um with this interface and uh um when we incubate this with zinc oxide we get um uh zinc oxide filling this part of the triangle as you can see here uh which is opening up really very exciting new ways of making completely new hybrid materials so I've kind of given you a bit of a whirlwind tour of um some of the things that can be done with modern protein design the applications um there there are really a wide range of applications in medicine I've talked about pandemic preparedness um and cancer immunotherapy nerd degeneration um there's applications and Technologies I've showed you and um many interesting problems as sustainability and my prediction will be that over the next 5 to 10 years we'll see an increasing number of designed proteins out in the world uh addressing these problems um and uh so um finally most important part I want to thank all the amazing people who contributed to this work it's really I mean the the prize really celebrates the work of a really large community of people um currently in my research group and more and and then a very very large community of former students post talks colleagues and friends and it's really been just absolutely wonderful to work with with everyone and this prize is really all all all for for everyone who contributed to this and I'm really happy because I think nearly 200 people have come uh come to Stockholm to help me celebrate this which is really really wonderful and I hope I get some time to actually see them so than for