Introduction to the Drug Discovery Process

Part 1: Introduction to Drug Discovery

00:00:05.05 Hi, my name's Jim Wells. I'm with the departments of Pharmaceutical Chemistry and Cellular and Molecular Pharmacology at UCSF.
00:00:13.11 And I'm going to tell you today about the process of drug discovery and development
00:00:17.28 in two parts: part one it will be screening of compounds and in that regard I'll be joined by my colleague
00:00:24.02 Michelle Arkin.
00:00:27.26 Hi, I'm going to talk after Jim gives an early history of drug discovery
00:00:31.12 and talks a little bit about target identification, then I'll talk about the process of screening
00:00:36.15 and hit identification.
00:00:38.15 Great, see you in a bit.
00:00:40.29 This slide shows some of the products of modern drug discoveries,
00:00:44.27 such as Lipitor, which is used for cholesterol lowering,
00:00:48.20 or a more recent drug, Gleevec, which an anti-cancer drug.
00:00:52.19 These compounds were discovered through a very rational, systematic
00:00:57.07 process, involves a lot of exciting scientific discoveries
00:01:01.28 as well as a lot of serendipity, luck and hard work.
00:01:07.00 To understand how we found these compounds, it's useful for us to review how drug discovery came to be,
00:01:14.25 what's the sort of brief history of drug discovery, as I'll show you on the following slide.
00:01:19.18 To understand the modern drug discovery development process today
00:01:23.10 it's useful to review the history, briefly,
00:01:25.13 of drug discovery. Prior to 1900, most drugs, in fact only a few really,
00:01:32.24 were identified through human screening.
00:01:35.07 Natural products, for instance, aspirin, was discovered from tree bark.
00:01:40.22 Quinine was discovered. And even illicit drugs like cocaine were discovered.
00:01:46.22 Long about the turn of the century, 1906,
00:01:49.26 the Food and Drug Administration was established
00:01:52.23 because a number of these kinds of potions or elixirs were found to neither be safe nor efficacious
00:01:59.07 and it was necessary to regulate these in a systematic way.
00:02:03.13 And this led to the development of animal based screening, for example, to discover anesthetics,
00:02:11.12 bacterial screening to identify antibiotics, and the like,
00:02:17.13 tissue screening to identify compounds that could react with neurological receptors,
00:02:23.09 like GPCRs, HTS, high throughput screening,
00:02:28.21 now very common discovery technology, as you'll hear a lot more about in this talk,
00:02:34.12 for discovering target-based compounds.
00:02:40.08 And then, lastly, mechanism-based discovery, which was used for HIV drugs and the like
00:02:46.06 as well as molecular and cellular based screening for kinase inhibitors.
00:02:53.04 And finally, genomics, to actually profile patients to determine who will be affected and who won't be affected.
00:03:00.25 So, in fact, this process, the history of drug discovery,
00:03:04.18 had gone from the human to the molecular target
00:03:09.09 and this now in reverse reflects what we actually do today.
00:03:14.24 Shown on this slide is what, in sort of general terms,
00:03:19.04 the modern drug discovery process.
00:03:21.10 And this process starts off with a disease,
00:03:26.08 and from that disease one tries to, through a lot of biochemical
00:03:33.07 cell-based, genetic and other means identify
00:03:37.10 what is the target or the molecular species in a cell, in an organism,
00:03:43.12 that's causing that disease.
00:03:46.13 One then develops a drug to that target, as you can see here,
00:03:52.12 and then, having identified that target, one then needs to identify
00:03:56.28 a compound that will interact with that target in a phase called lead discovery.
00:04:01.00 This is a chemical process where we identify the first compounds that are actually important in modifying a disease
00:04:10.08 and then once compounds are identified, typically in cells and then in animals,
00:04:15.06 they're prepared for clinical trials in this process called drug development
00:04:20.18 which is, this phase is really about interfacing the compounds that we discovered here,
00:04:28.14 to the human biology that we wish to effect here.
00:04:31.27 And if successful, we'll come out of this process with a drug.
00:04:35.17 Now, this process is a long winded process. It typically takes about now about 10-15 years
00:04:42.26 to discover a drug and it's expensive too.
00:04:46.00 It's about half a billion to a billion dollars to develop a drug.
00:04:50.21 So, when you're thinking about the pills that you take in a bottle,
00:04:55.00 think about a shopping mall, because that's easily the cost that it takes to get to the drugs that we end up using.
00:05:03.01 Ok, I wanted to just review quickly what kind of classes, what kind of molecules constitute drugs.
00:05:11.10 There's actually three basic classes and they include
00:05:15.29 the small molecule, organic compounds,
00:05:19.27 typically, these are compounds whose molecular weight is less than five hundred
00:05:23.25 and they're taken, generally, orally,
00:05:25.19 although they can also be taken as an injectable.
00:05:28.06 And they represent the kind of classic drug that you think of when you go to a pharmacy,
00:05:32.14 that you would buy over the counter, for example.
00:05:35.12 There's another class of very important drugs known as the protein therapeutic drugs.
00:05:39.27 These are typically injectable drugs, molecular weights of over ten thousand,
00:05:47.26 often up to a hundred thousand, or even higher.
00:05:51.03 And they are the important class of biotherapeutics
00:05:55.21 and they represent about thirty percent of drug sales today.
00:06:00.17 The other class of drugs, actually one of the very first to be developed
00:06:03.09 are the vaccines. And these are
00:06:07.03 basically viruses, pieces of viruses, that are used to elicit an immune response to a disease.
00:06:12.28 So these are the basic categories and today I'm going to focus on small molecule drug discovery,
00:06:18.28 leaving these other two categories alone for another talk.
00:06:23.18 So, the process of small molecule discovery is a long and winding road.
00:06:29.29 And it starts off with identifying what is the most critical target that's involved
00:06:36.10 in mediating the disease. So, identifying the disease target.
00:06:41.01 Having identified that target, generally a protein target,
00:06:44.17 one then goes through a process known as hit identification, shown here,
00:06:51.01 and the role of hit identification is to get the first compounds that actually engage the target.
00:06:59.05 Which compounds actually bind to the protein of interest
00:07:03.06 and can begin our drug discovery process.
00:07:07.01 From there, taking that isolated protein in a test tube, we need to show that that compound
00:07:11.20 actually works in a cell.
00:07:14.04 And so, this begins this process called hit to lead
00:07:18.01 which is to generate a compound which has cellular potency.
00:07:21.18 The next stage, sort of drawing from there, to a larger scale,
00:07:26.28 is the lead optimization stage.
00:07:28.23 This is a critical stage in which one actually shows that these compounds
00:07:33.28 that have been generated have animal efficacy and actually work in a pharmacological model
00:07:41.17 for the disease. The next stage after that is the IND enabling stage,
00:07:47.07 this is the stage that is preparing compounds for clinical trials.
00:07:52.05 Primarily, it involves animal tox experiments, in addition, chemical synthesis, scale,
00:07:59.28 and formulations experiments, and at the end of this process, one would hope to have a package
00:08:05.26 that you could convince the food and drug administration that you have a compound
00:08:11.05 that is going to be both safe and efficacious when administered to humans.
00:08:16.26 Then begins the all-important human clinical trials
00:08:21.04 if the FDA agrees with you.
00:08:24.20 In the first trail, is for human safety. This is typically done in a dose escalation,
00:08:29.12 kind of trial, with healthy volunteers, although in certain disease settings, like cancer,
00:08:35.18 you can use people with cancer.
00:08:38.08 And the goal of this is really to find out
00:08:40.23 what is the circulatory lifetime of the drug in humans
00:08:46.12 and how safe is the drug if its dose is increased.
00:08:52.03 The next phase, phase two, is involved in determining the efficacy of the drug
00:08:57.19 in a disease setting. So, this would be taking patients with the disease,
00:09:01.10 treating them with your drug at a level that's below any toxicity that was observed in phase one
00:09:11.05 and in ranging doses to find out what is the efficacy of the drug as a function of its dose
00:09:19.07 and what's the best dose to best effect the disease.
00:09:23.03 So, from these small trials, then, one then moves into a much larger, what's called registration trial,
00:09:29.21 phase three, in which one then fixes the dose, fixes the disease,
00:09:35.22 fixes the formulation, and then treats a large number of cohorts, both with and without the drug
00:09:45.18 to determine how effective the drug is. And at the end of this time, if your drug is safe and efficacious, you'll submit
00:09:53.23 what's known as a new drug application, an NDA,
00:09:58.00 to the FDA. They will either, they will review it and agree with you or not,
00:10:03.05 that you have a drug that's ready to go into humans
00:10:05.29 and at the end of that process, you have this pill down here,
00:10:11.27 which will then be launched with great fanfare, because this process is, as you'll see, a very long and arduous one.
00:10:19.08 Ok so I'm going to start at the beginning here with target ID.
00:10:23.25 What's causing the disease? How is it, what is the actual molecular target that we want to go after?
00:10:30.29 This link to the disease of interest.
00:10:33.08 And this is actually a very, very, can be a very long process
00:10:37.11 to find out what causing the disease.
00:10:39.16 Many diseases we don't have a clue as to what their cause is.
00:10:42.28 And in fact, ironically, even with all the tests that we might do to validate a target,
00:10:49.01 the final validation of a target is not known until you get down here with the pill itself,
00:10:54.09 to see if that is actually effective in a human.
00:10:58.02 Ok, so just briefly, what are the general causes of disease, what are the things that we think about.
00:11:03.13 First thing is, I like to think is it a bug or is it in the body?
00:11:07.21 Is it an infectious agent
00:11:09.27 or is it a host imbalance? So, for instance, if it's an infectious agent,
00:11:15.03 that's causing the disease, generally these days, we have sequences of the pathogenic bacteria.
00:11:20.27 We'll find a target that's not in humans
00:11:24.17 and then we'll take that protein target and go after that
00:11:30.07 in the drug discovery process.
00:11:32.11 If however, it's a disease like host imbalance, maybe it's a metabolic disease or cardiovascular disease or cancer
00:11:40.07 you first have to decide is it due, is the disease caused by an underactive protein,
00:11:47.05 for instance, people with diabetes,
00:11:50.06 they're not as responsive to insulin and so by giving them back insulin
00:11:54.17 you can hope to modify and ameliorate that condition.
00:11:59.02 Other diseases, for instance, here, many cancers are caused by overactive proteins
00:12:06.13 such as kinases, and so there's a lot of interest in discovering drugs
00:12:10.28 that would inhibit specific kinases for cancer.
00:12:15.20 Ok, that is just sort of a very skimmed view of this process, but just to give you a sense.
00:12:22.13 Once you have identified the target, this target process actually can be very complex.
00:12:30.16 So, the human genome is vast, there's some twenty thousand genes
00:12:35.07 that code for proteins. And finding exactly which one is causing the disease
00:12:40.27 can be challenging. So, one you've come up with the protein,
00:12:45.20 and the gene that encodes it for that particular disease,
00:12:49.17 you're ready to go on to another very important consideration.
00:12:52.26 Which is that not only do you need to go after the biology of the target,
00:12:58.22 the target itself, which is causing the disease, but that target is
00:13:05.00 itself has to be amenable to small molecule discovery.
00:13:07.22 And by that I mean it has to be something that we think could bind a small molecule.
00:13:13.10 because that in the end is ultimately what we want to do.
00:13:16.16 What I show here is a recent drug target
00:13:22.22 known as the BCR-Abl protein, shown in this space filling view here,
00:13:29.25 and in it is the small molecule known as Gleevec,
00:13:34.05 which was found to bind to it.
00:13:35.11 And you can see that these are, it was thought that this would be a good drug target because it was known
00:13:40.15 that kinases such as this bind ATP, and ATP targets have pockets in them for which we can find small molecule surrogates.
00:13:49.04 And indeed, they did find them for Gleevec.
00:13:52.18 Another property to look at is does the site have a cavity or a hot-spot,
00:14:00.05 an energetic region in the molecule that can bind a small molecule
00:14:03.28 So these kinds of considerations sort of define the druggability of the target.
00:14:09.10 So, first you have to determine the biology of it,
00:14:11.23 and its link to a disease, and next the druggability of the target.
00:14:16.25 based on whether or not we think we can find compounds that will engage the target.
00:14:21.07 Ok, now certain targets really like to bind small molecules.
00:14:25.23 And in fact, one of the favored disease targets are GPCRs,
00:14:32.07 G-coupled protein receptors. They naturally bind small molecules
00:14:37.17 and they've been traditionally great targets for going after for small molecules.
00:14:41.19 and in fact about forty five percent of the approved drugs are GPCRs.
00:14:48.27 Other targets which are of great interest in terms of the biology are kinases, that I had mentioned
00:14:55.07 proteases and protein-protein targets.
00:14:57.17 These targets can bind small molecules, but in general, they bind them weaker
00:15:03.21 than GPCRs and that probably accounts for why we see so much activity in finding GPCR inhibitors than others.
00:15:13.10 We then want to talk about what is it that we're looking for at the end?
00:15:18.19 What is it we hope this drug will be?
00:15:22.19 Well, first of all, most oral drugs, we'd love them to be just a single, daily dose.
00:15:27.05 That you only have to take it once
00:15:28.26 and typically, that would mean something less than a hundred milligrams of the drug that you would take.
00:15:35.25 Just for example, here's a pill bottle of Ibuprofen,
00:15:41.25 in here are tablets which are about in total two hundred milligrams or so, less than half of that is the drug itself
00:15:48.21 because the rest is the formulation
00:15:51.05 for the drug.
00:15:52.24 Now, in order for that to be the case, in order for us to be
00:15:57.03 at a drug dose of a hundred milligrams per day,
00:15:59.16 there's certain molecular properties that a compound's going to have to show.
00:16:03.00 And one of them is it's going to have to bind to the target with a high affinity
00:16:08.14 and selectivity, so that it binds just one target
00:16:14.03 ideally, so that, and does so with a great deal of potency,
00:16:19.01 so that you don't need much compound to trigger that.
00:16:20.13 We can measure that in this process over here, where we show
00:16:25.08 direct binding of a compound to a protein.
00:16:28.20 So, what we can do is we can titrate the compound in, increasing concentration from left to right
00:16:34.02 of the compound and measure the binding ability
00:16:37.12 of the compound to the target
00:16:39.08 and this case we can measure, then, at what concentration we get fifty percent binding
00:16:43.06 and that's called the Kd
00:16:44.29 and in this case it's ten nanomolar. That's a nice binding compound. We would like our compounds to be
00:16:50.02 that potent or more.
00:16:52.20 In addition, we can measure the potency of the compound in a cell,
00:16:56.09 because that's obviously, going to have to bind to the protein in the cell.
00:17:03.10 And in order to do that, we would like potency of this in a cell based assay
00:17:07.16 to be at least a hundred nanomolar
00:17:10.09 for the midpoint here in binding to the cell.
00:17:13.00 If it meets those criteria, then it has the potency potential
00:17:17.28 to be at a once daily, less than a hundred milligrams dose.
00:17:20.11 But there's some other very important considerations too.
00:17:23.03 And we'll get to that in the lead optimization part of this talk.
00:17:26.27 Which is the half-life of the compound in the body.
00:17:30.24 So, we would like the compound to not be cleared too rapidly,
00:17:34.17 the body wants to clear small molecule compounds, does so through the liver and the kidney,
00:17:41.28 and in other means. And we would like that compound to have a half-life in blood
00:17:49.05 of greater than about three hours
00:17:52.09 in order to have an effectiveness over twenty four hours for the drug.
00:17:57.04 Also, we do want the drug once we take it to be orally active,
00:18:02.02 so that we would want the oral uptake to be at least fifty percent of the drug ingested.
00:18:07.21 to be taken up. Ok, well with these considerations in mind,
00:18:12.21 let me just go also into the chemical considerations of the drug
00:18:17.13 that we want. So those are some of the biological considerations,
00:18:21.03 what are some of the chemical considerations.
00:18:23.26 And here we have a list of four guidelines that were provided by Chris Lipinski
00:18:29.06 and his colleagues at Pfizer that studied
00:18:31.04 a whole variety of orally active drugs
00:18:35.01 and identified several properties that are important for making good, potent
00:18:43.17 and orally active drugs. So, for example, one of the things that most orally active drugs
00:18:49.22 one of their properties, is that they have a molecular weight
00:18:52.10 less than five hundred Daltons.
00:18:56.29 So, we would look to be building compounds that are less than that.
00:18:58.01 They have good solubility, things that are not very soluble don't dissolve so well
00:19:04.28 so they don't go through the gut. And so we'd like to have this parameter called cLog P
00:19:09.12 less than 5.
00:19:12.07 Looking at the structure of the drug too,
00:19:13.24 they found that drugs that had fewer than ten hydrogen bond acceptors
00:19:17.15 such as these moieties here, and here, were better at being orally active drugs
00:19:29.21 as were compounds that had fewer than 5 hydrogen bond donors,
00:19:34.01 such as these species here. And so, these are just general guidelines,
00:19:39.26 they're not, they're many drugs that actually break these rules,
00:19:42.08 and more recently, people are realizing that we can finesse these rules,
00:19:48.00 finesse compounds so that they can actually exceed these guidelines.
00:19:51.16 But they do serve as useful guidelines in thinking about it.
00:19:54.24 Ok, the next consideration that one needs to make is that
00:20:00.25 chemical space, that is we now have to move on to finding compounds that will actually engage our target.
00:20:05.19 The chemome, it turns out, as it were, these are compounds less than five hundred molecular weight
00:20:12.04 is a huge chemical space.
00:20:15.02 If we were to calculate, as has been done, all the compounds that can be made
00:20:21.12 with a molecular weight five hundred from carbon, hydrogen, nitrogen and oxygen,
00:20:28.02 typical constituents in an organic drug,
00:20:32.14 one could build compounds that would be about ten to the sixty in diversity.
00:20:38.06 Now, this number ten to the sixty two is greater than the number of particles in the universe,
00:20:44.04 so we're not going to be able to build all these things.
00:20:46.22 And so what we want is we want methods where we can actually search chemical space
00:20:51.12 effectively to find the area where we want the drug to be.
00:20:58.25 Generally, it turns out that most targets will yield hits
00:21:02.03 to the target when we screen on the range of a thousand or a few thousand compounds
00:21:08.25 we'll find a handful of hits from that process.
00:21:13.03 So the more compounds you screen, the better the likelihood is of getting to hits to your target.
00:21:19.22 And there's no way that we're going to screen
00:21:21.26 all the compounds that are out there, that could be made out there
00:21:26.02 because there's just far too much chemical diversity for us to sample everything.
00:21:32.00 In this regard, bringing a small molecule to a big molecule, involves a lot of considerations
00:21:37.21 that I talked about and also involves a lot of faith,
00:21:40.09 the same of kind of faith that Michelangelo had when drawing this picture.
00:21:43.10 And I'd now like to turn it over to my colleague,
00:21:45.19 Michelle Arkin, who's going to tell you about how we actually make this happen
00:21:49.02 in the laboratory.
00:21:50.22 Alright, so today we're going to talk about hit identification
00:21:54.25 so I say this, getting on the board. So, getting a small molecule
00:21:58.21 that has some of the initial properties
00:21:59.21 that you want in your final drug. And as you see, it's
00:22:01.27 very early in the process. And so having a good starting point here
00:22:06.09 with good molecular properties and good selectivity
00:22:08.20 is really going to help as you winnow down based on other properties
00:22:11.21 that Jim will talk about in the second park of his lecture.
00:22:15.17 So, how are we going to identify that initial chemical starting point.
00:22:19.11 One way is to start with a natural substrate and make it drug-like.
00:22:23.24 If you know the substrate, for example, the ATP-like analogues that Jim gave as an example earlier,
00:22:29.14 you can make these more drug like. That's a sensible place to start.
00:22:32.28 You can start with somebody else's starting point.
00:22:35.14 We call this, jokingly, patent busting, but it's also
00:22:39.03 a very interesting way of getting a late stage compound.
00:22:44.09 If you know the liabilities of your competitors
00:22:46.12 or your own compounds, then you can improve those liabilities
00:22:51.04 and really get a better molecule right off the bat.
00:22:54.00 You can also design a hit from scratch, de novo,
00:22:57.17 using structure based design if you really know a lot about your binding site
00:23:01.16 and really know a lot about that protein's structure and this is difficult
00:23:05.02 but it is an area of a lot of improvement and a lot of research.
00:23:08.28 And finally, probably the most widely used method nowadays is screening.
00:23:14.01 HTS stands for high throughput screening,
00:23:16.12 a more technical, and a little bit new approach is called fragment based screening,
00:23:20.26 but we're going to focus today on screening by high throughput screening.
00:23:24.18 And there are two reasons to do this; one is it's very
00:23:27.09 generally applicable, widely used in the industry,
00:23:30.12 and also more and more people in academics are using this technique
00:23:34.02 as well, to find molecules, not only as starting points for drugs,
00:23:37.25 but also as tools to investigate their biology.
00:23:40.10 So you can think, yourself, your own biology, would it be helped by having a small molecule
00:23:44.09 that specifically interrogates that biology
00:23:46.24 and high throughput screening is one approach for getting those tool compounds.
00:23:49.26 So, especially when you're starting with high throughput screening, you have to screen a lot of molecules
00:23:56.10 to find a drug. So, we're starting here, a high throughput screen can be anywhere from a hundred thousand to a million compounds.
00:24:02.22 Maybe even more than a million compounds.
00:24:05.03 And from that you're going to have some hit rate,
00:24:07.19 so most of those molecules will not be active against your biology.
00:24:10.24 But then a large subset will, say five hundred,
00:24:12.16 to five thousand. Then we have other metrics that we're going to use to winnow those compounds
00:24:17.20 down, based on potency, other in vitro properties, to a hundred to five hundred
00:24:24.12 and you can see that the process keeps going down, here you can see that we're doing toxicology and safety studies,
00:24:28.08 and now through clinical trials, we have attrition all the way until we get to that final molecule.
00:24:33.10 So, we're going to be testing a lot of compounds, we're going to be synthesizing a lot of compounds.
00:24:38.08 So, we need to have robust ways of doing that, precisely and with high throughput.
00:24:43.04 So, what do these molecules look like?
00:24:45.12 We start with libraries of drug-like molecules,
00:24:47.13 and they can come from several places. Three common places that these
00:24:50.24 libraries can be assembled from include natural products,
00:24:54.03 so compounds that come out of extracts or now partially purified compounds from fungi and other species,
00:25:03.21 structure design, so if we know something about the molecular targets that we're looking
00:25:08.00 at, for example, GPCRs are a very common target, as Jim mentioned,
00:25:13.06 and there are a lot libraries that have compounds that look like other GPCR agonists or antagonists.
00:25:17.14 So, you again have a designed library based on particular biological and structural classes.
00:25:23.13 And then finally, there are diversity libraries.
00:25:27.05 These are just selections of molecules selected to be as different from each other as possible
00:25:30.26 still trying to fall within these drug-like properties, like Lipinski's rules,
00:25:35.17 as Jim mentioned, or other drug-like parameters that you might consider.
00:25:40.05 And this is a schematic example of what those drugs would look like.
00:25:43.16 The straight lines are scaffolds, so you'll have different shapes to the molecules,
00:25:47.27 and then the different polygons around those are different functional groups.
00:25:51.15 So, you'll have an array of different shapes, different sizes,
00:25:54.10 and different functionalities to come to about half a million compounds
00:25:58.03 plus or minus. You're then going to screen those compounds
00:26:01.09 in some assay that is amenable to screening half a million compounds,
00:26:06.14 and then at the end, hope that you get some number of molecules
00:26:09.05 that bind to the target in the way that you want.
00:26:11.20 So, how do we handle this many compounds?
00:26:15.02 There are two general ways to think about handling them, one is miniaturization and the other is automation.
00:26:19.29 So, in miniaturization, here we have a standard format plate,
00:26:23.22 and each of these wells is another experiment.
00:26:28.11 Like a test tube or an Eppendorf tube, but they're arrayed here in 96, 96 times 4,
00:26:34.05 384 well format, and even smaller, 4 times smaller than that, fifteen thirty six well format.
00:26:40.09 So, here we can do one thousand five hundred and thirty six experiments
00:26:44.10 on a single plate.
00:26:45.11 And we'll go into the lab in a minute and I'll show you what these plates look like.
00:26:48.07 And then there's automation, so we have automation,
00:26:51.06 that allows us to prosecute many examples of these plates, one right after the other,
00:26:56.13 we have parallel pipetters, so that we can
00:26:58.11 add reagents with high reproducibility to each of these experiments.
00:27:01.28 Robotics to handle those plates and those pipettes,
00:27:05.16 and then plate readers that are able to read all of these experiments in a high-throughput fashion.
00:27:10.23 So, you can hear now and see now that we're in the laboratory,
00:27:15.29 can you hear the hum of activity?
00:27:18.07 We were just talking about miniaturization into small format plates
00:27:23.08 and automation, so I thought we would show you where we keep our libraries
00:27:26.06 and then show you some automation.
00:27:28.26 We have several of these freezers and in each freezer
00:27:31.05 we have five shelves, and in each shelf we have large numbers of racks of compounds.
00:27:35.26 Here you can see they're stored in 96- or 384- well format and they're barcoded
00:27:41.02 so that we can directly go back to our database and find out what's in those wells.
00:27:47.17 So that we can identify the structures of the compounds and we know what to do next.
00:27:51.01 Ok, now let's go look at some automation.
00:27:55.23 Now we're in the automation room, and I wanted to show you
00:27:59.22 what the plates look like.
00:28:02.01 At the small molecule discovery center, we work mostly in 96-well format
00:28:06.12 and 384-well format and different color plates depending on the assay.
00:28:10.04 Here's a 96-well plate used for luminescence assay,
00:28:12.28 you want to have maximum scatter,
00:28:14.25 to get all of your photons from the luminescence.
00:28:17.17 Then we have black plates, here's a 384-well format,
00:28:20.12 the black bottom plate, this is used for fluorescence,
00:28:22.13 which we measure from the top.
00:28:24.20 And then here's a 384-black sided plate with a clear bottom, and this we use for doing high content imaging for microscopy.
00:28:31.00 Here we have the liquid handler, so these liquid handlers are set up to do automatic pipetting,
00:28:40.19 when you have a lot of plates to work with.
00:28:43.11 We have an incubated plate hotel so your cells can stay warm and cozy,
00:28:47.06 in their plates while they're waiting to be processed.
00:28:49.25 And then over in the back, you probably can't see it here, but in the back,
00:28:53.09 we have a large plate hotel that holds tip boxes, so pipette tips,
00:28:57.00 and also plates for processing.
00:28:59.21 So, now we're going to go through one of these routines,
00:29:02.22 so you can see how the pipetter works.
00:29:05.16 Now we're watching an automation run, where we're going to be adding media to a plate.
00:29:11.25 That will eventually contain cells.
00:29:14.02 So you see the 384-well plate is coming out of an incubated plate hotel,
00:29:18.08 this is where the cells can sit while they're being processed,
00:29:21.19 so that they can stay warm and cozy.
00:29:33.05 You'll notice we're going from a 96-well plate into a 384-well plate.
00:29:36.15 So we're using a 96 tip head,
00:29:59.05 And the pipetter is going to pull media from the 96-well plate
00:30:03.14 and pipette it into all four quadrants
00:30:05.26 of the 384-well plate.
00:30:23.19 Now we say thank you to the tips,
00:30:28.11 thank you to the 96-well plate,
00:30:36.09 and we'll send that 384-well plate of cells back into its hotel.
00:30:50.07 And this plate hotel can hold 42 plates at a time.
00:30:54.24 So you can imagine that pipetting 384 experiments by 42 times would be quite laborious without an instrument like this.
00:31:02.01 Fluorescence is one of the most popular assay
00:31:05.24 formats to use both for biochemical and cell based assays.
00:31:08.20 There are a lot great things about fluorescence. It's very sensitive,
00:31:11.28 often in the nanomolar range of sensitivity,
00:31:14.27 it has a very wide dynamic range, so you can see over several orders of magnitude
00:31:19.26 of intensity.
00:31:21.28 And you can also tune the wavelengths so that you can look at multiple probes at one time
00:31:25.12 that fluoresce in different colors or you can look at the interaction between probes.
00:31:37.00 So, naturally, we don't look at fluorescent 384-well plate
00:31:40.25 visually and score it, we use a plate reader.
00:31:43.21 And so here's the very same plate, on that plate reader,
00:31:46.17 collects the data very quickly, and you can see the image coming up on the screen,
00:31:52.02 as we go.
00:31:53.24 The red is going to be the lowest fluorescence
00:31:56.09 and the blue is among the highest fluorescence on the plate.
00:31:59.15 From there, if you're screening a whole bunch of compounds, you can see that you would have a range of colors
00:32:02.28 in between those. And the top and the bottom of this plate show a dose response curve.
00:32:07.02 So that you can see how the fluorescence changes intensity as the concentration drops.
00:32:12.24 So, this allows us to quantify how much fluorescence we have
00:32:16.08 and therefore how much active enzyme or inactive enzyme
00:32:18.22 we have present in the mix.
00:32:19.28 We just looked at some of the automation equipment
00:32:32.03 that we have in the laboratory, showed you the freezers,
00:32:35.10 where the compounds are stored.
00:32:37.03 Now let's talk about what assays you're actually going to put into those little wells.
00:32:40.01 What little miniaturized experiments are we going to do?
00:32:42.05 So that depends very much on your biology and what's known about the biology.
00:32:46.20 Now, if you have a purified protein and you can make an activity that you can visualize,
00:32:51.00 very typically fluorescence or luminescence,
00:32:53.16 you can do a biochemical based assay.
00:32:56.23 So here's a very simple example where we have an enzyme that binds to a substrate
00:33:00.17 and this substrate has this caged fluorophore in green
00:33:03.13 so that it doesn't fluoresce when it's not bound,
00:33:06.23 then when it binds to the enzyme and is activated by the enzyme,
00:33:11.09 is acted on by the enzyme, it releases
00:33:13.16 this fluorophore, which we can now detect as an increase in fluorescence here at 520 nanometers.
00:33:19.08 If we're looking for an activator of this enzyme,
00:33:21.23 we'd be looking for a low signal that becomes a high signal.
00:33:24.14 If we're looking for an inhibitor of the enzyme,
00:33:27.08 we'll be starting with a situation like this, with a high fluorescence signal,
00:33:30.24 and looking for things that inhibit that signal.
00:33:33.01 So there are several instances where a biochemical assay is really an appropriate high-throughput screening assay
00:33:40.00 especially if you know what protein you want to inhibit.
00:33:43.11 Bcr-Abl is a good example, it's the known agent of causing
00:33:48.25 CML. So inhibiting that enzyme is going to inhibit that disease progression.
00:33:52.27 You also need to be able to express and purify that protein in large enough quantities to run many experiments on it
00:34:01.16 and that protein needs to be similar enough to what it is in the cell to make sense.
00:34:08.01 Also, the selectivity among similar proteins is important.
00:34:11.28 So, again, the kinase example. There are hundreds of kinases in the body,
00:34:15.29 some you want to inhibit for your disease, some you don't.
00:34:18.22 They're critically involved in homeostasis.
00:34:20.24 So, if you have a whole panel of those similar proteins,
00:34:24.12 then you can test activity against your protein of interest
00:34:27.10 without having activity against the proteins you're not interested in.
00:34:30.10 Ok, there are other examples. So when you have a biochemical assay,
00:34:36.17 you come out of it knowing the molecular target.
00:34:39.05 You know what protein your compound is binding to.
00:34:41.28 What you don't know coming out of that assay is whether
00:34:44.21 that compound is going to survive in the cell,
00:34:47.10 whether it's still going to bind to your target when it's in the cell,
00:34:50.03 whether it will even be permeable and get into that cell.
00:34:52.28 On the other hand, if you use a cell based assay, then you'll know the opposite.
00:34:57.26 You'll know that the compound is active
00:34:58.29 in the cell, because that's how you identified it,
00:35:01.12 but you won't necessarily know the molecular target,
00:35:03.10 the protein or other biomolecule that that molecule is targeting.
00:35:06.18 So, when would you want to screen in a whole-cell context.
00:35:10.03 There are several cases where you'd want to screen in cells,
00:35:15.12 and this is an area that's really gotten to be much more popular recently,
00:35:19.01 as high throughput cell culture has become available.
00:35:21.25 For example, if you know the molecular target, but you can't isolate from it the cell,
00:35:26.18 GPCRs are a good example, ion channels are an example,
00:35:30.02 and more subtly, some kinases even are regulated
00:35:34.23 by subcellular localization, by time in the cell cycle and if you take that protein
00:35:39.13 out of its context, it might have a very different inhibitory profile
00:35:43.03 then in the cellular context.
00:35:44.23 Also, sometimes your goal is to alter a cellular phenotype.
00:35:49.03 Maybe you're not that interested in the molecular target,
00:35:51.28 but you want to turn on a whole pathway,
00:35:55.08 or you want to kill a bacteria,
00:35:56.06 and you're a little less concerned about how you kill that bacteria
00:35:58.21 or for example, you get your phenotype and you'll use that molecule to understand the biology
00:36:03.21 well enough, use that molecule to tell you what the important targets are
00:36:09.04 to cause that phenotype.
00:36:10.18 For example, death of a bacterium.
00:36:13.03 Also, if we know the pathway that we want to inhibit,
00:36:16.18 but we don't know the best or most druggable target
00:36:18.27 in that pathway, we can allow the molecules to tell us that information
00:36:22.03 by using an assay that measures some parameter at the bottom of that pathway
00:36:26.28 and see which molecules inhibit or activate that function.
00:36:30.25 How are we going to read out these cell based assays?
00:36:37.05 So, remember we have cells cultured in this small, microtiter plates now,
00:36:42.05 and we want to read them out in some way that a plate reader will be able to detect them.
00:36:46.25 There are two very common, basic approaches
00:36:48.26 to looking at cell-based assays.
00:36:51.01 One is to look at a whole well parameter,
00:36:52.25 such as fluorescence or luminescence.
00:36:55.11 So, this would be an example of reporter cell lines,
00:36:58.21 where you have a luciferase or you have a green fluorescent protein
00:37:02.12 that's downstream of a promoter
00:37:04.19 that you want to turn on or off.
00:37:05.09 And if you alter the state of that promoter, then you're going to
00:37:08.19 get more or less of that fluorescent or luminescent protein.
00:37:12.21 Another more recent advance that's getting to be very popular also is to look at those
00:37:19.02 properties cell by cell, and we're going to go through some examples
00:37:22.00 of high-content imaging, where we're looking at an automated microscope
00:37:26.10 so we can look at each cell in that well,
00:37:28.10 and multiple parameters that are affected by the drugs.
00:37:33.22 So this instrument is our high-content screening instrument,
00:37:38.11 it's an In Cell Analyzer, by GE Healthcare.
00:37:40.01 It's basically automated microscopy. It allows to automatically
00:37:44.18 collect data, fluorescence and absorbance, transmitted light data,
00:37:49.09 across several wavelengths, sequentially, in a multi-well plate. Several plates at a time.
00:37:54.13 And then later, we'll analyze that data.
00:37:56.12 So the assay we're going to look at today is a pretty simple, straight-forward assay.
00:38:00.00 It just has one stain, we can do up to six stains.
00:38:03.05 This is just a nuclear stain, which we'll see in blue,
00:38:06.04 and we're looking for trypanosomes. Trypanosomes are parasites,
00:38:10.02 they're the causative agent of Chagas disease, which is a terrible disease
00:38:13.03 that infects people in South America.
00:38:15.07 And these trypanosomes, these little parasites, live inside
00:38:19.03 host cells. So, this assay is going to have a host cell that's infected with parasites
00:38:23.21 and we'll be able to detect and quantify the number of parasites per host cell
00:38:28.08 based on the nucleic acid staining.
00:38:30.20 So, let's get started.
00:38:31.21 So, now Kenny, our high-content screening expert,
00:38:34.23 starting the instrumentation and it's loading the plate
00:38:39.00 in the plate reader, it's setting the objectives,
00:38:41.03 the lenses to 10x objective, the correct filters to detect
00:38:47.07 fluorescence in the blue channel. And now you can see
00:38:50.01 the data are coming up.
00:38:53.04 We're scanning twenty fields per well
00:38:55.11 see here, and as the data are collected, they're coming up on the screen.
00:39:00.06 And they're also being saved into the computer.
00:39:04.08 They'll go to our database for analysis later.
00:39:06.27 You can see the large spots are host nuclei and the smaller spots
00:39:12.07 in some of them are trypanosome kinetoplastid DNA.
00:39:15.08 This is very dense DNA that is next to the nucleus.
00:39:20.03 We just looked at this experiment on the high content imager
00:39:22.08 and now we're going to look at how we analyze that data.
00:39:24.27 So, you saw that we can automatically collect these experiments
00:39:27.15 and in the blue channel, here, we're looking at nucleic acid.
00:39:31.24 We see the large circles here, are the host nuclei and the small dots are kinetoplastid DNA
00:39:40.07 from the parasite. So it's very straightforward for the instrument,
00:39:45.23 for us to teach the instrument which of these blobs belong to host nuclei and
00:39:50.05 which of these blobs belong to the parasite.
00:39:52.12 And here you see circled in green what the instrument has identified as host nucleus
00:39:57.23 and in yellow, circled, what the instrument has identified as kinetoplastid, or parasite DNA.
00:40:03.06 We can then very simply count the number of cells that have parasites in them
00:40:08.05 and look at a reduction in the number of infected cells
00:40:10.26 with drug treatment.
00:40:12.08 Or look at the ratio of parasites to host cells.
00:40:15.26 One really nice thing about an assay like this is instead of a whole well assay, where we just see the properties overall,
00:40:21.25 here we can look as well at the health of the cells.
00:40:25.09 So, are we killing the host nuclei? Are we causing them to be apoptotic?
00:40:29.05 Or to otherwise change their morphology?
00:40:31.09 We don't want to alter that cell, we only want to remove the number of parasites
00:40:35.29 that are infecting those cells.
00:40:37.08 So, we can get a lot of selectivity and some early toxicity information from an experiment like this.
00:40:42.06 So, now we have run our screen and this is an example of a ninety thousand
00:40:48.02 compound screen. So, a sort of moderate size high-throughput screen.
00:40:51.27 And you can see the three types of experiments that you get out of this.
00:40:56.08 Here you have on the y-axis, we have a percent inhibition.
00:41:00.13 So, we're looking at an enzyme that's active at the bottom and
00:41:05.07 fully inhibited, or inactive, at the top.
00:41:07.25 In green, we see what the average fluorescence is,
00:41:12.07 these are maximum signal controls on each plate
00:41:16.01 that show you the variance in the activity of the enzyme across that plate.
00:41:24.03 And then a fully inhibited enzyme, here these are minimum signal controls
00:41:27.24 going across the plate and you see some variation,
00:41:30.23 but overall, this is a minimum signal, a highly inhibited enzyme signal.
00:41:36.02 And it's normalized to the average here, being 100 percent inhibited
00:41:40.21 and the average of the green being 100 percent active.
00:41:43.26 Now, in blue, are the compounds that we've actually screened.
00:41:48.07 So, how do we identify which of these compounds are inhibitors,
00:41:52.02 or partial inhibitors that we want to follow up in further assays.
00:41:54.26 We have an assay parameter that's commonly used that's called a Z-prime value
00:41:58.25 and the Z-prime value measures the width of the minimum signal,
00:42:03.14 the width of the maximum signal,
00:42:05.19 and the distance between them to come up with an assay parameter
00:42:08.07 that varies between negative one and one,
00:42:11.09 one being a perfect assay, point five being adequate for high-throughput screening.
00:42:15.12 To give you an example, this has a Z-prime factor across the entire assay of about point eight.
00:42:22.08 So this is a very high quality assay.
00:42:24.01 Now we go to find hits in that assay, here the grey line is that average
00:42:29.18 minimum signal, it's the average fully inhibited enzyme.
00:42:32.24 And here, this black line, shows you three standard deviations above the noise.
00:42:37.18 So, most of the compounds are totally inactive in this assay,
00:42:40.25 the enzyme was fully active. And then three standard deviations above that
00:42:45.00 these would be considered statistically relevant changes
00:42:48.06 from active or partially inhibited.
00:42:52.15 We would go in and select several of these compounds and test them some more.
00:42:56.04 So the next steps are to re-test those compounds to see if they
00:43:00.06 recapitulate, run does responses of those compounds to see how potent they are,
00:43:04.27 determine the mechanism of inhibition.
00:43:06.23 This is really important aspect of high throughput screening
00:43:09.28 because a lot of compounds will inhibit a cellular function or protein
00:43:15.04 for reasons that are not drug-like. So we need to tease
00:43:16.22 those apart before we put a lot of effort into the compounds.
00:43:19.26 Then once we're confident that the compounds are truly active,
00:43:23.04 and acting in the way we want them to be acting,
00:43:25.09 then we go into optimize.
00:43:26.21 And this is what Jim will be talking about in the next segment of the seminar.
00:43:31.05 So, just to go back to where we started,
00:43:33.15 hit identification takes some time, and there are a lot of experiments that go there.
00:43:38.25 But it's just the first step to discovering a drug.
00:43:42.02 So, if we start here with a hundred thousand to a million compounds,
00:43:45.22 we really want to make sure that what's going through this funnel
00:43:49.15 has a chance of being successful at the end.

Part 2: From "Hit" to Pill

00:00:04.10 Hi, this is Jim Wells again. I'm going to tell you about part two: from hit to pill.
00:00:10.10 Now, last time we left off, we had taken you up through the stages of these nine stages.
00:00:17.26 We'd taken you up from the target identification stage, to hit ID, and now we're at the hit to lead stage.
00:00:24.23 And the key thing about this stage
00:00:26.26 is that we want to identify compounds that don't just bind to the protein,
00:00:30.17 they actually work inside a cell. And they actually show selectivity in a cell.
00:00:35.25 And this is yet another elevated level
00:00:40.10 and gets us closer to what we want, which is a drug that's safe and efficacious in people.
00:00:44.26 The key aspect of hit to lead stage is an iterative process in which we
00:00:50.27 not only show that the compound works in a biochemical assay,
00:00:56.14 but we also demonstrate that it works effectively and selectively in a cell-based assay.
00:01:01.11 So, it can actually go through the cell membrane, reach the target inside the cell,
00:01:06.03 if it is an intracellular target,
00:01:08.13 and engage that protein in a cell-based assay.
00:01:14.07 So, in starting this process, the compounds start off with potencies
00:01:20.22 that are weaker than we would like.
00:01:23.09 As shown in this biochemical binding assay
00:01:25.16 what we're looking for is compounds that will make the medicinal chemistry
00:01:29.19 that will improve the potency of the hit compound at least a factor of ten,
00:01:35.23 ideally a factor of twenty, in the biochemical assay.
00:01:39.15 We'll also be looking for things that start off with, from the hit stage,
00:01:44.00 that have weak cellular potency, shown here,
00:01:46.26 but with medicinal chemistry that correlates with the biochemical potency above,
00:01:51.02 drives the cellular potency to be more potent in the cell.
00:01:56.01 And this is all toward, the goal here is to get potent compounds
00:02:00.04 that are cell active.
00:02:02.11 Now, we also look in this stage for several other important properties.
00:02:06.13 For instance, we don't want it to bind to other off-targets
00:02:11.26 that are related to it that may cause toxicity.
00:02:14.18 So, what we'd prefer is that the compounds have potency that are at least ten-fold weaker
00:02:20.11 to the closest related target.
00:02:22.08 We'll also be looking to see that there's chemical evidence that we can advance these compounds.
00:02:27.00 When we look at compound's structures,
00:02:29.19 you should see that there are compounds which both
00:02:33.24 affect the target and compounds which don't affect the target
00:02:36.27 that are fairly closely related
00:02:39.03 suggesting that they're binding to a single site.
00:02:40.22 Because one of the properties of a drug is that it bind to a single site
00:02:45.03 on the target protein.
00:02:46.16 We'd like to know the molecular target; Michelle mentioned
00:02:50.17 that you can start off these hit-finding expeditions
00:02:54.10 just looking for cellular activity.
00:02:56.08 In that case, we don't really know the target that's involved.
00:03:00.02 At this stage, we really would like to have a molecular understanding of what that
00:03:04.04 small molecule is engaging.
00:03:08.06 Lastly, and very importantly,
00:03:10.08 in this especially for formulations reasons,
00:03:12.11 we look to have compounds which have solubilities above a hundred micromolar.
00:03:17.08 And that's because we wish the compounds to be
00:03:20.28 soluble and dissolve well once they're administered to an animal.
00:03:27.00 Ok, having successfully passed this stage,
00:03:31.00 of a hit to lead stage,
00:03:32.21 we're now into this very important and pivotal stage,
00:03:36.07 called lead optimization.
00:03:38.06 This is where we're looking to see that compounds that we've created
00:03:41.15 can actually work in a whole animal, can reach the
00:03:45.08 protein target, through a cell, and through the circulatory system of the animal.
00:03:51.25 Ok, so I think at this stage it would be good to understand
00:03:56.04 what happens when you swallow a pill.
00:03:58.18 Here's our patient right here, going to take a pill,
00:04:03.21 there's the pills, they go down through the saliva
00:04:08.15 into the stomach, where they're
00:04:11.13 subjected to pH one, so they have to survive that.
00:04:15.10 They come out of the stomach and they go straight into the duodenum,
00:04:19.15 which is this arrow here, which you see goes straight into the intestines.
00:04:23.18 Now, the intestinal, the intestines have about
00:04:27.25 two hundred square meters of surface area to absorb.
00:04:30.10 That's about the size of a tennis court.
00:04:31.10 There's a lot of opportunity to absorb a drug,
00:04:35.19 but it has to have the right physical properties to get across that intestinal barrier
00:04:40.27 so it actually enters into the blood stream.
00:04:43.04 Once it does get across that, into the intestine,
00:04:47.08 and into the bloodstream, the first thing that is does is
00:04:49.29 it goes straight into the liver through the portal vein, here,
00:04:53.01 and then meets a series of very important obstacles.
00:04:58.03 They include a variety of enzymes that are in the liver that are meant to detoxify,
00:05:02.25 get rid of, these foreign small molecules.
00:05:06.08 So, they would be things like p450s, which oxidize the compound,
00:05:09.26 or hydrolases, protease, lipases, esterases,
00:05:14.14 that would be hydrolyzing compounds.
00:05:18.12 There would be glucaronidation enzymes and the like,
00:05:19.21 which tag them so that they can be rapidly excreted.
00:05:23.14 Of course, all of these things can reduce the potency and the availability of the drug
00:05:28.29 to have its effect in the peripheral tissues where itâ€™s probably got to act.
00:05:33.16 So, if really does, the drug has to get through this very important gauntlet here,
00:05:37.29 of the liver in order to go on from there, throughout the circulatory system
00:05:48.03 first into the right atrium, then into the lungs,
00:05:53.14 and then back through the lungs, into the left atrium,
00:05:56.08 and then through the arterial system all the way throughout the body.
00:05:59.19 So, understanding the pathway that a pill has to take
00:06:05.08 is actually very important in this drug discovery process
00:06:07.27 because even though a compound could be great in a biochemical assay or a cell based assay
00:06:12.27 if it doesn't work in this system over here
00:06:15.14 then it's not going to be a drug.
00:06:19.24 Ok, once, now understanding that
00:06:25.14 and having successfully passed the hit-to-lead stage we're now ready to do lead optimization.
00:06:31.03 And that again is an iterative process of chemistry and biology
00:06:35.11 paired together. But this time, at the animal level.
00:06:39.02 So, the very first thing that happens when you have compounds from the hit-to-lead stage
00:06:43.09 is that you'd want to determine how well do they survive in the body.
00:06:47.16 How long do they live in the body?
00:06:49.21 And to do this, you typically give either an injection of the compound
00:06:54.06 or feed an animal this compound
00:06:56.26 and then determine the PK of the compound
00:07:00.01 in other words, how well does that compound,
00:07:03.12 this is time along the x-axis here and amount of compound
00:07:07.19 that's in an animal along the y-axis
00:07:10.05 and as the compound is injected, you see that there's a rise
00:07:15.24 in compound levels as it reaches, as it goes into the circulatory system
00:07:20.11 and then it decays.
00:07:22.09 And this half-life and other parameters
00:07:24.25 are used to judge how well will that compound work in an animal
00:07:29.08 or how long it will last in an animal.
00:07:31.12 The next stage is called the pharmacodynamics part of this
00:07:37.20 and there we're actually looking to see how active is this compound
00:07:41.00 in an animal model.
00:07:42.23 This, I show here, is an example of a xenograph model for cancer.
00:07:46.05 This, for instance, is time along the x-axis and along the y-axis is tumor volume.
00:07:52.16 You can see if the animal is untreated with your drug,
00:07:56.13 your drug candidate, the tumor grows rapidly,
00:08:00.04 that's this blue line here,
00:08:01.29 and then, in the case of a compound that shows some efficacy,
00:08:06.17 if treated with the animal, treating the animal with that, you can see you can suppress the growth of the tumor.
00:08:12.02 That would be a good result.
00:08:13.15 Now, one would continue along this process
00:08:16.25 of lead optimization until one has gotten compounds
00:08:20.24 that show good PK and good animal efficacy
00:08:24.11 according to those guidelines that we discussed at the beginning of this lecture.
00:08:28.11 The notion of lead optimization is really one of
00:08:35.18 trying to push compounds from the vast amount, trying to identify compounds
00:08:40.20 from the vast chemical space that's out there.
00:08:44.22 We, though the hit-to-lead process, we've identified potent compounds,
00:08:48.10 both biochemically and in cell assay,
00:08:52.02 we use pharmacokinetics to identify those compounds that have both potency and good pharmacokinetics
00:08:59.10 so we take compounds within this area of the Venn
00:09:02.17 and then using our animal efficacy experiments, our PD or pharmacodynamic models,
00:09:08.17 we're really interested the compounds that are sitting right in this narrow area here.
00:09:13.13 So, it's this iterative process of testing these compounds
00:09:17.00 in pharmacokinetics, pharmacodynamics and cell and biochemical assays
00:09:22.18 that ultimately then identifies a compound that can be deemed a clinical candidate.
00:09:27.25 This would be a compound that would be ready for IND enabling studies
00:09:32.16 and hopefully on the pathway to being a drug.
00:09:36.10 So, the goals for an oral drug, then, after this process
00:09:42.29 again are that we're looking for compounds that can be dosed
00:09:46.15 once a day, at hopefully less than a hundred milligrams per day,
00:09:51.24 we want reasonable protein binding potency
00:09:57.28 and reasonable cell activity.
00:10:00.29 We want a decent half-life, so that it only has to be taken once a day
00:10:04.10 and we want good oral uptake
00:10:07.26 so that it can be taken orally.
00:10:10.05 Ok, all of this work I told you about,
00:10:13.09 this reminds me of this Greek myth of this character down here,
00:10:18.15 who you can see is exhausted
00:10:20.13 and that's how you feel once you get to the clinical candidacy stage.
00:10:23.29 You've just spent a lot of time, a lot of effort,
00:10:28.08 from target ID, hit ID, hit to lead and lead optimization,
00:10:32.04 you've just reached this plateau here and now you look up
00:10:36.00 and there's all of this stuff still to go
00:10:38.12 because that's the drug development stage that you need to get to to finally reach the nirvana.
00:10:43.29 So, now having worked all, done all the work in the drug discovery part,
00:10:49.17 which is basically this area from target ID to, through lead optimization
00:10:55.29 to identify a clinical candidate,
00:10:59.11 that typically takes somewhere around three to four years to go through that process.
00:11:03.01 There's still a whole lot of work to go on.
00:11:06.25 From IND enabling all the way through registration and FDA approval.
00:11:10.10 This process is an even, much more expensive and longer process,
00:11:17.11 about four to eight years to get through that process.
00:11:19.24 And it's worth considering what are the,
00:11:23.23 what are some of the reasons that compounds, we should this as a funnel,
00:11:27.13 because there's a lot of attrition that goes on in this process.
00:11:30.25 In fact, when a compound enters the clinic,
00:11:33.21 here in phase one, only one in ten compounds will actually make it to being a pill.
00:11:38.23 What are some of the reasons that they don't make it all the way to being a pill?
00:11:43.02 About a third of the reasons are due to lack of efficacy.
00:11:48.16 So, this is, when a compound shown to be efficacious in animals
00:11:54.28 and all of the target ID experiments that were done before,
00:12:01.11 this really reflects the fact that we didn't have a validated target
00:12:04.17 to begin with. That there's more going on with the biology than we thought
00:12:07.22 and it's sort of back to the drawing board
00:12:11.00 about what is actually driving that.
00:12:13.11 Other reasons for compound failure
00:12:15.08 is about a third of them die because of toxicity in humans.
00:12:17.17 These are things that we couldn't really model in animals
00:12:22.09 and so they wind up being disqualified as, they may have failed in phase one.
00:12:28.29 Other reasons for failure are pharmacokinetics,
00:12:33.06 they just don't have the right clearance
00:12:35.18 and the right properties to be a once a day or twice a day medication.
00:12:42.25 And, so they can fail for those reasons.
00:12:45.16 There's another, and much less likely failure,
00:12:48.27 which is due to commercialization reasons,
00:12:52.00 and this may be because the company just feels that the market size is not what they thought,
00:12:54.21 the drug is really not going to be worth developing further.
00:13:00.24 And so they will stop that.
00:13:03.14 There are many other reasons that might happen,
00:13:05.11 but these represent the lion's share of the kind of failures that happen in the clinic.
00:13:09.27 So, we really need to get much better at doing this.
00:13:15.02 We know that healthcare costs are spiraling out of control
00:13:19.21 and this represents something like fifty to twenty percent of the GNP
00:13:24.13 and new drugs and new therapeutics can really
00:13:27.16 spare us a lot of money in our economy if we're able to find new compounds that can ameliorate a disease.
00:13:38.01 Because it's much better to treat it early than to have to deal with the symptoms
00:13:43.21 and the ramifications after.
00:13:45.27 This also highlights the fact that we need to get to
00:13:49.14 prevention and cure, versus crisis and symptom,
00:13:52.06 which is the process that we're in now.
00:13:55.11 It's much cheaper to prevent and cure
00:13:57.04 than it is to manage a crisis or just simply manage the symptoms of disease.
00:14:01.10 Now, how are we doing in this process?
00:14:05.02 Well, there's something like twenty thousand gene targets,
00:14:08.03 meaning twenty thousand genes that encode for proteins;
00:14:11.19 if each of those were associated with a disease,
00:14:16.08 there's got to be a one to one correlation
00:14:17.13 but you could imagine that we should be able to find a lot more drugs
00:14:24.04 because currently there's only two hundred drug targets that are addressed.
00:14:27.28 And that means we're a long way from really
00:14:32.13 saturating the genome, or the proteome, as it were, with potential compounds
00:14:37.23 that can affect disease.
00:14:40.05 The other thing is that these days
00:14:43.10 we're only adding, the FDA, while it approves something like
00:14:46.07 thirty to forty new drugs a year,
00:14:48.28 we're only adding three to four new targets addressed per year.
00:14:51.20 That's because many of the drugs that are approved these days
00:14:54.28 are what are called "me too" drugs
00:14:57.04 that are second generation drugs that may have better pharmacokinetics,
00:15:00.09 less toxicity, but they're really addressing a brand new mechanism.
00:15:05.12 We truly need to address new mechanisms if we're going to affect
00:15:09.25 our healthcare.
00:15:12.13 So, in the last little bit here,
00:15:15.18 I wanted to tell you why I think drug discovery is so exciting
00:15:18.00 from a scientific point of view.
00:15:20.16 And I'm going to do this, these next few slides
00:15:23.03 with the example of Gleevec.
00:15:24.11 Gleevec's a new drug to treat CML,
00:15:30.05 and other cancers.
00:15:33.10 And this was discovered from some very careful
00:15:36.21 molecular biology and oncology studies
00:15:41.04 and genome wide association studies, that showed that this target,
00:15:45.17 the ABL kinase, gets hyper activated when its fused in patients with
00:15:52.27 CML because of this fusion with this protein BCR.
00:16:00.28 So, BCR-ABL, when they're fused together
00:16:03.14 at the gene level and then encoding a protein,
00:16:06.23 leads to this hyperactive protein kinase
00:16:09.09 that phosphorylates its substrates much more active than it should
00:16:12.23 and that drives cellular proliferation in chronic myeloginous lukemias.
00:16:19.01 Now, turns out that there's a, this is a kinase that binds ATP
00:16:24.24 and using high through put screening, a group at Novartis was able to identify compounds
00:16:29.17 that would actually displace ATP from BCR-ABL
00:16:32.20 and thereby inactivate the kinase
00:16:35.06 so it that it would no longer drive CML.
00:16:39.11 And this was a spectacular discovery,
00:16:41.26 one of the first really highly-targeted kinase inhibitors
00:16:47.03 that showed dramatic effects on this really life-threatening disease.
00:16:52.00 Another aspect of drug discovery that's so fascinating
00:16:57.14 is the fact that it allows us to understand how small molecules bind to their proteins of interest.
00:17:04.06 These are, they use weak forces to carry out this binding
00:17:12.13 reaction and itâ€™s those weak forces that if we understood them better,
00:17:16.22 we could really make better drugs, discover drugs more rationally,
00:17:19.01 and more completely.
00:17:21.14 And so, by looking at the structures of compounds bound,
00:17:24.29 to here, we're showing Gleevec bound to BCR-ABL,
00:17:28.22 in the spheres here,
00:17:31.26 we can begin to understand the hydrogen bonding interactions
00:17:34.13 between the drug, the ion dipole interactions between the drug and the protein and
00:17:38.26 hydrophobic interactions between the drug and the protein
00:17:41.16 and we can make analogues, chemical analogues, of these drugs and see how that affects potency,
00:17:47.18 so we can really understand the molecular details of how these contacts drive potency
00:17:51.29 and that way we can then understand how we can build better ones
00:17:56.02 because by seeing the values of certain interactions,
00:17:58.27 we can encode those into the small molecule to drive potency.
00:18:02.26 Another aspect that I find particularly fascinating
00:18:07.05 is that proteins themselves are quite dynamic.
00:18:10.15 They're moving in solution, their side chains are rotating on the surface of them.
00:18:15.10 This is our friend BCR-ABL and when this protein
00:18:21.05 binds its substrate, it undergoes conformational changes
00:18:24.21 and when it binds the drug Gleevec, it involves other conformational changes.
00:18:29.26 So, maybe if we could show this movie here,
00:18:33.27 you can see here, as the protein binds the substrate,
00:18:35.29 there in this green region,
00:18:37.07 you can see a conformational change.
00:18:38.24 And then when it binds Gleevec, there's a much larger conformational change that ensues.
00:18:43.17 These conformational changes are very important
00:18:46.17 in understanding how proteins bind small molecules
00:18:52.07 and how they can be controlled
00:18:53.20 because they are moving, dynamic objects that can be trapped
00:19:00.13 in different conformations and drugs allow us
00:19:01.26 a way of studying these kind of phenomena.
00:19:04.29 Now, another thing that's very exciting about drug discovery
00:19:10.08 is it allows one to have, a chemical biologist or biologist, to have tools
00:19:14.01 that can allow you to effect a biology in a cell in a very rapid and dose dependent manner.
00:19:21.24 Small molecules will act, often, on the timescale of seconds
00:19:26.13 to minutes, and this is much faster than
00:19:30.19 such probative technologies like knockout mice,
00:19:34.03 which take the development of a whole organism,
00:19:38.03 or siRNA, which typically takes 72 hours before you've had a complete knockdown
00:19:42.19 of the protein of interest.
00:19:46.14 Another aspect of drug discovery that's very exciting,
00:19:49.28 and this is for the pharmacologists,
00:19:51.28 is that it allows them tools to understand, to relate the physiological
00:19:57.22 effect of inhibiting or activating a particular target,
00:20:02.12 on the animal itself.
00:20:04.19 And in how that relates to the disease.
00:20:07.13 And so, this really goes, provides, important tools for pharmacologists
00:20:14.18 and has for years, for understanding
00:20:17.25 how particular biologies are regulating the phenotypes that we see in animals.
00:20:23.27 And lastly, it provides tools for geneticists and clinicians to relate
00:20:29.15 pharmacogenomics and epidemiology to disease.
00:20:33.29 As I mentioned, the ultimate target validation is a drug itself,
00:20:37.25 an approved drug that is shown to be effective in a particular disease
00:20:43.02 indeed validates that target, that this is important.
00:20:46.16 And additionally, by looking at responders and non-responding patients
00:20:52.12 and the difference in their genomes, we can begin to understand what other factors modulate
00:20:57.17 the disease itself and what makes them sensitive or resistant to certain compounds.
00:21:06.09 And finally, one of the things that's so great about drug discovery and development
00:21:12.03 is that it provides the fundamental basis for a healthier society.
00:21:16.14 Drugs have made a huge impact on our
00:21:22.15 health and will continue to do so.
00:21:26.03 We have a lot more to do in this area, as I've shown you,
00:21:29.04 and this is something that's, I think, is particularly exciting and worthwhile doing.
00:21:36.14 So, I'd like to thank you for listening to this presentation and see you later!

Videos in this Talk

Part 1: Introduction to Drug Discovery
Part 2: From "Hit" to Pill

Speakers: James Wells, Michelle Arkin

Total Duration: 1:06:03

Recorded: June 2009

All Talks in Archive

Talk Overview

The modern drug discovery process integrates our deepest understanding of the molecular basis for disease with fundamental understanding of how potential drug molecules interact with specific disease targets and the whole organism. These two lectures are intended to give a broad and general introduction to the drug discovery process. Part I focuses on the early stages of drug discovery. We describe the basic stages of the drug discovery process, beginning with how disease targets are identified.

We then take you into the laboratory to show you how a popular approach, high throughput screening, is used to identify compounds that can engage the molecular target or pathway of interest. Part II focuses on the later stages of the drug discovery process. After the identification of the first hit compounds, iterative rounds of chemistry, biochemistry, cell biology, pharmacology, and toxicology are used to progressively build and polish compounds to make them ready for clinical trials. The adventure of drug discovery represents one of the most fascinating and beneficial pursuits in science.

Speaker Bio

James A. Wells, Ph.D. Dr. Wells received a B.A. degree in biochemistry from the University of California, Berkeley, and a Ph.D. degree in biochemistry from Washington State University with Dr. Ralph Yount. He is a joint Professor in the Departments of Cellular & Molecular Pharmacology, and Pharmaceutical Chemistry and Chair of the Department.

His current research uses a combination of protein design and small molecule drug discovery to study and modulate cellular processes involved in apoptosis and inflammation. Some of his awards include: the Hans Neurath and Aviv Awards given by the Protein Society, the Pfizer Award given by the American Chemical Society, the du Vignead Award given by the Peptide Society, and in 1999 was elected member of the National Academy of Sciences. See www.wellslab.ucsf.edu for more details.

Speaker Bio

James Wells

Michelle Arkin

Michelle R. Arkin, Ph.D. Dr. Arkin received her B.A. degree in chemistry from Bryn Mawr College and her Ph.D. in chemistry at Caltech with Dr. Jacqueline Barton. In 2005 she became the Associate Director of Cell Biology at Sunesis and led the translational science team for Voreloxin, an anti-cancer agent in phase 2 clinical trials…. Continue Reading