WEBVTT 1 00:00:00.420 --> 00:00:05.300 The Untapped Therapeutic Frontier There are nine point five million possible drug-disease combinations. 2 00:00:05.800 --> 00:00:12.360 You've tested thirty-two thousand five hundred. That's like searching for your keys in zero point three four percent of 3 00:00:12.360 --> 00:00:20.140 your house and declaring them lost. The other ninety-nine point seven percent is sitting right there, untested, because 4 00:00:20.140 --> 00:00:27.440 the molecules weren't profitable enough to bother. How You Accidentally Cured Three Things and Then Stopped. 5 00:00:28.720 --> 00:00:36.580 In nineteen ninety-six, a team of Pfizer scientists sat in a room looking at clinical trial data for a chest pain drug called 6 00:00:36.580 --> 00:00:40.880 sildenafil. The drug had failed. It did not fix chest pain. 7 00:00:41.740 --> 00:00:49.380 Nobody's chest felt better. The trial was, by every metric that mattered to the people who paid for it, a complete waste 8 00:00:49.380 --> 00:00:55.960 of money. Then the test subjects refused to return their leftover pills. 9 00:00:56.660 --> 00:01:01.000 This had never happened before. People don't hoard failed medicine. 10 00:01:01.640 --> 00:01:09.140 The scientists investigated. What they discovered was that sildenafil had a side effect so commercially valuable it would 11 00:01:09.140 --> 00:01:14.160 generate one point nine billion dollars per year for the next two decades. 12 00:01:15.100 --> 00:01:23.680 You know it as Viagra. The cure for erectile dysfunction was hiding inside a failed heart drug, and the only reason anyone 13 00:01:23.680 --> 00:01:28.900 found it was because a group of men did something unusual with a feedback form. 14 00:01:29.880 --> 00:01:34.410 Minoxidil has a similar origin story. Developed for high blood pressure. 15 00:01:34.920 --> 00:01:38.870 Failed at that too. Side effect: hair growth. It became Rogaine. 16 00:01:40.000 --> 00:01:49.020 Then sildenafil got repurposed *again* for pulmonary hypertension, making it three cures from two failed drugs, none of 17 00:01:49.020 --> 00:01:56.620 them intentional. On Wishonia, when we heard about this, we asked the obvious question: "So then you tested all the other 18 00:01:56.620 --> 00:02:00.180 drugs for all the other side effects, right?". You did not. 19 00:02:00.960 --> 00:02:09.539 You found three cures by accident, said "huh, neat," and went back to testing the same twelve biological targets that pharmaceutical 20 00:02:09.539 --> 00:02:17.200 companies can patent. This is like finding buried treasure in your backyard, selling the treasure, and then never digging 21 00:02:17.200 --> 00:02:21.430 in your backyard again because the first hole was supposed to be for a fence post. 22 00:02:22.340 --> 00:02:27.850 I have been observing your species for eighty years. I have seen you do many confusing things. 23 00:02:28.320 --> 00:02:32.340 But this one keeps me up at night, which is saying something because I don't sleep. 24 00:02:33.820 --> 00:02:37.860 Your Medicine Cabinet (After five thousand years of Civilization). 25 00:02:39.120 --> 00:02:42.380 Prescriptions: twenty thousand products approved in the U.S. 26 00:02:42.980 --> 00:02:47.060 Sounds impressive until you realize most are variations of each other. 27 00:02:47.440 --> 00:02:54.680 Your pharmaceutical industry has invented three hundred ways to sell the same painkiller in different-colored packaging. 28 00:02:55.220 --> 00:02:59.620 On Wishonia, this would be considered a graphic design achievement, not a medical one. 29 00:03:00.580 --> 00:03:06.440 Unique Ingredients: These products contain only one thousand six hundred and fifty compounds. 30 00:03:06.800 --> 00:03:10.580 Fewer unique molecules than a French bakery uses ingredients. 31 00:03:11.320 --> 00:03:16.960 Investigational Drugs: seven thousand five hundred compounds have passed Phase One globally. 32 00:03:17.460 --> 00:03:23.240 These are compounds you already know won't kill people, sitting in filing cabinets, doing nothing. 33 00:03:24.480 --> 00:03:29.580 Safe Stuff: The F.D.A. generally recognized as safe, or G.R.A.S. 34 00:03:29.600 --> 00:03:35.140 list contains six hundred and thirty-five substances. Things you could eat right now without dying. 35 00:03:35.760 --> 00:03:43.620 Low bar. Still useful. Total unique compounds humanity knows are safe-ish: nine thousand five hundred compounds. 36 00:03:44.280 --> 00:03:49.600 That's it. Five millennia of civilization, and this is your entire toolkit. 37 00:03:50.500 --> 00:03:54.500 On Wishonia, we have more compounds in a single bathroom cabinet. 38 00:03:55.180 --> 00:03:59.900 Your bathroom cabinet has shampoo and expired cough syrup from twenty nineteen. 39 00:04:01.280 --> 00:04:06.860 Ways Your Body Breaks (The To-Do List Nobody's Doing). Codes: The I.C.D.-ten lists fourteen thousand ways your meat can 40 00:04:06.860 --> 00:04:12.120 malfunction. This is your species' most comprehensive to-do list. 41 00:04:12.280 --> 00:04:21.240 You are not doing it. Real Targets: After consolidation, one thousand trial-relevant diseases are actually worth testing 42 00:04:21.240 --> 00:04:27.190 drugs against. The Multiplication Problem. Here is where it gets uncomfortable. 43 00:04:28.800 --> 00:04:34.280 Take your nine thousand five hundred safe compounds. Multiply by one thousand diseases. 44 00:04:34.780 --> 00:04:41.080 This gives you the total number of things you *could* test tomorrow morning if you weren't busy filling out forms about 45 00:04:41.100 --> 00:04:49.040 whether you're allowed to test them. Nine thousand five hundred compounds times one thousand diseases equals nine point 46 00:04:49.040 --> 00:04:55.180 five million possible drug-disease combinations. This is not a hard calculation. 47 00:04:55.840 --> 00:05:03.080 A child could do it. Several children have. The children were then told to stop asking uncomfortable questions and go play 48 00:05:03.080 --> 00:05:13.920 outside. What You've Actually Tried. Approved uses: one thousand seven hundred and fifty unique approved drug-disease pairings. 49 00:05:14.080 --> 00:05:19.570 This is your entire medical canon. Five thousand years of civilization. 50 00:05:20.380 --> 00:05:24.700 Libraries of medical journals. Nobel Prizes. White coats. Stethoscopes. 51 00:05:25.520 --> 00:05:33.720 All of it produced this number. Repurposed uses: thirty percent of drugs gain at least one new indication after initial 52 00:05:33.720 --> 00:05:42.300 approval. The rest never get a second date. Failed trials: Let's be generous and say you tested ten times as many things 53 00:05:42.300 --> 00:05:49.120 that failed. You learned nothing from most of them because your system throws away negative results like a restaurant that 54 00:05:49.120 --> 00:05:56.510 only writes down the recipes that taste good. Total tested: thirty-two thousand five hundred relationships. 55 00:05:57.700 --> 00:05:59.800 The Number That Should Keep You Awake. Thirty-two thousand five hundred relationships tested out of nine point five million 56 00:05:59.800 --> 00:06:10.400 possibilities equals zero point three four two percent. where: **You have explored one-third of one percent of the things 57 00:06:10.400 --> 00:06:16.780 you could test using molecules you already have.** I want to sit with that for a moment. 58 00:06:17.420 --> 00:06:23.760 You have safe compounds. You have sick people. You have the ability to introduce them to each other. 59 00:06:24.400 --> 00:06:28.760 You have done this zero point three four two percent of the time. 60 00:06:29.280 --> 00:06:34.460 The other ninety-nine point six six percent of introductions have never been made. 61 00:06:34.880 --> 00:06:39.440 Not because they wouldn't work. Because the molecule didn't have a marketing budget. 62 00:06:40.460 --> 00:06:47.560 On Wishonia, if a doctor explored zero point three four two percent of treatment options and declared the patient incurable, 63 00:06:48.220 --> 00:06:55.700 we would gently suggest the doctor continue looking. On Earth, you publish a paper about the zero point three four two percent 64 00:06:55.700 --> 00:07:02.080 and apply for a grant to re-study the same zero point three four two percent you studied last time. 65 00:07:03.180 --> 00:07:07.160 "Maybe the Other ninety-nine point six six percent Is Useless". 66 00:07:08.140 --> 00:07:13.320 This is the objection I hear most often from humans who would prefer not to feel bad about this. 67 00:07:14.120 --> 00:07:19.660 Let me address it by introducing you to your own biology, which you have mostly not met. 68 00:07:20.300 --> 00:07:26.260 You're ignoring eighty-eight percent of your own biology. Mapping three hundred and fifty thousand plus clinical trials 69 00:07:26.260 --> 00:07:31.080 showed that only twelve percent of the human interactome has ever been targeted by drugs. 70 00:07:31.700 --> 00:07:35.120 You have a one hundred-room mansion. You live in the bathroom. 71 00:07:35.700 --> 00:07:40.230 You have decorated the bathroom very nicely. You have published papers about the bathroom. 72 00:07:40.840 --> 00:07:47.440 The other eighty-eight rooms are locked, and when someone asks about them, you say "we're focused on bathroom excellence 73 00:07:47.440 --> 00:07:51.920 at this time.". Your drugs are promiscuous and you're pretending they're not. 74 00:07:52.800 --> 00:08:00.120 F.D.A.-approved drugs bind to things you didn't intend. This means "side effects" are frequently "cures for something else" 75 00:08:00.140 --> 00:08:04.480 that nobody bothered to notice. Remember Viagra? That was a side effect. 76 00:08:05.420 --> 00:08:16.220 Rogaine? Side effect. Sildenafil got repurposed *again* for pulmonary hypertension: three cures from two drugs, none intentional. 77 00:08:16.400 --> 00:08:21.140 Your entire drug discovery strategy is "wander around and hope you trip over something." 78 00:08:21.760 --> 00:08:28.550 It works surprisingly often, which is less a compliment to the strategy than an indictment of having no strategy at all. 79 00:08:29.060 --> 00:08:35.179 Your diseases share wiring you haven't traced. Diseases cluster on shared biological networks. 80 00:08:35.860 --> 00:08:44.039 Arthritis and Alzheimer's share pathways. Lithium was prescribed for gout in the eighteen hundreds; decades later, it turned 81 00:08:44.039 --> 00:08:47.230 out to be the gold standard treatment for bipolar disorder. 82 00:08:47.820 --> 00:08:51.840 Nobody connected the dots because nobody was drawing lines between diseases. 83 00:08:52.640 --> 00:08:58.020 The drug that treats your grandfather's inflammation might fix your grandmother's neurodegeneration. 84 00:08:58.720 --> 00:09:04.660 You won't find out, because testing it would require a one hundred million dollar clinical trial that takes longer than 85 00:09:04.660 --> 00:09:09.340 Grandma has left. Beyond Your Cupboard (Where It Gets Embarrassing). 86 00:09:10.340 --> 00:09:13.260 Everything above only counted molecules you *already have*. 87 00:09:13.680 --> 00:09:18.900 The safe stuff. The things sitting in your filing cabinets right now, gathering dust and expiration dates. 88 00:09:19.440 --> 00:09:27.680 If you look at what's chemically possible, it gets worse. Chemical Space: Estimated ten to the twenty-third power to ten 89 00:09:27.680 --> 00:09:33.640 to the sixtieth power drug-like molecules. More candidates than atoms in your body. 90 00:09:33.740 --> 00:09:40.640 More than stars in your galaxy. More than the number of times your regulatory agencies have said "we need more data.". 91 00:09:41.740 --> 00:09:47.300 Virtual Libraries: ZINC twenty-two contains tens of billions of purchasable virtual molecules. 92 00:09:47.680 --> 00:09:53.000 You can order them online. Like shoes, but for not dying. You have ordered more shoes. 93 00:09:54.080 --> 00:09:59.980 Synthesized: You have made less than ten to the negative fortieth power of possible molecules. 94 00:10:00.720 --> 00:10:08.040 Tested divided by Possible equals approximately ten to the negative twentieth power to ten to the negative fiftieth power. 95 00:10:08.240 --> 00:10:12.940 These are numbers so small that the notation system itself starts to feel embarrassed. 96 00:10:13.940 --> 00:10:19.840 Ten to the negative fiftieth power is not a fraction. It's a rounding error's rounding error. 97 00:10:20.360 --> 00:10:25.820 It is the mathematical equivalent of having not started. Because you have not started. 98 00:10:26.720 --> 00:10:35.860 Why You Are Still Sick (A Diagnosis). I want to be very clear: you are not sick because curing disease is impossible. 99 00:10:36.400 --> 00:10:44.660 You are sick because you built a system that charges one hundred million dollars per introduction between a molecule and 100 00:10:44.660 --> 00:10:51.780 a disease, and then gave all the money to a different system that introduces warheads to cities. 101 00:10:52.600 --> 00:10:58.640 The warheads work fine. The disease introduction system has four design flaws. 102 00:10:59.220 --> 00:11:02.880 All of them are about money. None of them are about science. 103 00:11:03.840 --> 00:11:13.340 The ticket price is insane. A typical Phase Two to Three trial costs thirty to one hundred million dollars (median cost 104 00:11:13.340 --> 00:11:21.720 per patient forty-one thousand dollars). At that price, you only test things pharmaceutical companies can patent and sell 105 00:11:21.720 --> 00:11:32.300 at enormous markups. Everything else sits untested. Imagine a dating app where each date costs one hundred million dollars 106 00:11:32.300 --> 00:11:38.300 and only billionaires are allowed to swipe. That's your drug approval system. 107 00:11:38.460 --> 00:11:44.200 The molecules without money die alone. Everyone goes to France. 108 00:11:44.220 --> 00:11:51.760 Trials overwhelmingly test the same few biological targets due to preferential attachment dynamics. 109 00:11:52.720 --> 00:11:57.500 Everyone studies what everyone else studies because that's how you get grants. 110 00:11:58.120 --> 00:12:05.600 It's a herd behavior that would embarrass actual herds. Cows at least spread out across the field. 111 00:12:05.940 --> 00:12:13.340 Your researchers cluster around twelve targets because those are the targets with grant funding, and the grant funding goes 112 00:12:13.340 --> 00:12:17.250 to those targets because those are the targets with researchers. 113 00:12:18.080 --> 00:12:24.440 This is a circle. Not the good kind. Cheap molecules can't play. 114 00:12:24.620 --> 00:12:30.920 Most molecules have no patent. No patent means no one hundred million dollar trial. 115 00:12:30.980 --> 00:12:39.740 No trial means no approval. Metformin, a diabetes drug from the nineteen-fifties, shows persistent signals for reducing 116 00:12:39.740 --> 00:12:45.820 cancer risk and extending lifespan. It costs pennies. It's off-patent. 117 00:12:46.700 --> 00:12:55.120 The trial to prove it works for aging (TAME) spent years begging for funding because no pharmaceutical company will bankroll 118 00:12:55.120 --> 00:13:02.980 a drug they can't monopolize. The molecule that could save your life doesn't make enough money to be worth testing. 119 00:13:03.540 --> 00:13:11.310 You die. Quarterly earnings go up. The system is working exactly as designed, which is the problem. 120 00:13:12.160 --> 00:13:19.660 Unprofitable diseases don't exist. If you have a rare disease with no market, you don't get a drug. 121 00:13:20.300 --> 00:13:29.480 Your cells malfunction in an unprofitable way. The invisible hand of the market has weighed your life against shareholder 122 00:13:29.480 --> 00:13:35.930 returns and found you wanting. On Wishonia, we don't have an invisible hand. 123 00:13:36.400 --> 00:13:43.730 We have a visible calculator. It counts deaths. When the number goes up, it allocates resources. 124 00:13:44.520 --> 00:13:51.540 This is apparently "radical" on your planet. The Timeline (How Long Until You Finish Looking). 125 00:13:52.480 --> 00:14:00.280 At your current pace of three point three thousand trials per year, I calculated how long it would take to systematically 126 00:14:00.280 --> 00:14:08.300 test what you have. The results are, and I say this with genuine affection for your species, not great. 127 00:14:09.080 --> 00:14:18.880 Just the safe compounds you already own: Nine point five million combinations divided by three point three thousand trials 128 00:14:18.880 --> 00:14:27.080 per year equals two point eight eight thousand years. Two thousand eight hundred and eighty years. 129 00:14:27.700 --> 00:14:32.680 Using compounds *already confirmed not to kill people*. Sitting in cabinets. 130 00:14:33.220 --> 00:14:39.940 With labels on them. You won't finish until roughly the year five thousand, at which point the people who needed the drugs 131 00:14:39.940 --> 00:14:44.820 will have been dead for three thousand years, which somewhat reduces the clinical urgency. 132 00:14:45.760 --> 00:14:54.680 Combination therapies (which is how modern medicine actually works): Oncology, H.I.V., cardiology, psychiatry: all routinely 133 00:14:54.680 --> 00:15:00.180 use multi-drug cocktails. The real question isn't "does this one drug work?" 134 00:15:00.580 --> 00:15:09.260 but "which combinations work?" Testing pairwise combinations: Forty-five point one billion combinations divided by three 135 00:15:09.260 --> 00:15:15.420 thousand three hundred trials per year equals thirteen point seven million years. 136 00:15:16.340 --> 00:15:19.640 That's longer than *Homo sapiens* has existed as a species. 137 00:15:20.480 --> 00:15:26.680 You would need to have started testing drug combinations before your ancestors climbed down from the trees, and you still 138 00:15:26.680 --> 00:15:32.120 wouldn't be done. And that's only pairs. If you want triplets, you'll need to outlive the sun. 139 00:15:32.920 --> 00:15:41.460 The full chemical universe: Beyond your cupboard, the chemical space contains ten to the twenty-third to ten to the sixtieth 140 00:15:41.460 --> 00:15:49.880 power drug-like molecules. At current trial capacity, exploring even the synthesizable fraction would take longer than the 141 00:15:49.880 --> 00:15:55.760 universe has existed. The universe is thirteen point eight billion years old. 142 00:15:56.480 --> 00:16:02.430 You need more time than that. To test pills. But Won't We Run Out of Easy Discoveries? 143 00:16:03.400 --> 00:16:09.740 Some of your economists (the ones who've never looked through a microscope) will argue "diminishing returns." 144 00:16:09.820 --> 00:16:16.320 The more you search, the harder each new find becomes. This assumes you've already picked the low-hanging fruit. 145 00:16:17.160 --> 00:16:24.880 You are not in the orchard. You are standing outside the orchard filling out a forty-seven-page application to enter the 146 00:16:24.880 --> 00:16:32.060 orchard. The application requires three letters of recommendation, an institutional review board approval, and a fourteen-month 147 00:16:32.060 --> 00:16:36.700 waiting period. The fruit is rotting. You can see it rotting. 148 00:16:37.480 --> 00:16:45.120 You're on page twelve of the application. Diminishing returns kicks in after you've explored a meaningful fraction of the 149 00:16:45.120 --> 00:16:52.760 space. At less than one percent explored, every trial teaches you something new, improving your ability to predict what 150 00:16:52.760 --> 00:17:01.760 works next. The math: even with weak learning effects, diminishing returns only dominates after exploring roughly thirty-seven 151 00:17:01.760 --> 00:17:09.540 percent of therapeutic space. For combination therapies, that crossover is over five million years away at current pace 152 00:17:09.540 --> 00:17:16.099 (thirty-seven percent of forty-five point one billion combinations at three thousand three hundred trials per year). 153 00:17:16.700 --> 00:17:20.940 You cannot have diminishing returns when you haven't had any returns. 154 00:17:21.819 --> 00:17:27.300 See the formal analysis for the equations. What Happens When You Actually Try. 155 00:17:27.380 --> 00:17:34.540 The Oxford RECOVERY Trial during COVID-nineteen proved something your pharmaceutical industry would prefer you didn't know: 156 00:17:35.580 --> 00:17:43.020 large-scale pragmatic trials can run for roughly five hundred dollars per patient and deliver results in less than one hundred 157 00:17:43.020 --> 00:17:48.040 days. Five hundred dollars versus forty-one thousand dollars. 158 00:17:48.740 --> 00:17:52.480 Read those two numbers again. One of them is the cost of finding out if a drug works. 159 00:17:53.180 --> 00:17:59.020 The other is the cost of finding out if a drug works while also paying for marble lobbies, compliance departments, and a 160 00:17:59.020 --> 00:18:04.200 CEO's third yacht. The medicine is the same. The patients are the same. 161 00:18:04.820 --> 00:18:13.520 The yacht is new. With a decentralized F.D.A., trial capacity increases by twelve point three times: Nine point five million 162 00:18:13.520 --> 00:18:20.700 combinations divided by forty thousand seven hundred trials per year equals two hundred and thirty-four years. 163 00:18:21.440 --> 00:18:27.460 Still not fast. But at least it's a number your grandchildren might live to see the end of, rather than a number that requires 164 00:18:27.460 --> 00:18:35.640 your species to outlive its own star. The first few decades would capture the highest-value discoveries, the cures that 165 00:18:35.640 --> 00:18:40.300 current trial scarcity forces you to ignore while people die in the waiting room. 166 00:18:40.960 --> 00:18:47.060 Your decentralized F.D.A. builds permanent infrastructure for systematic therapeutic exploration. 167 00:18:47.640 --> 00:18:54.020 The goal isn't to "finish." The goal is to stop leaving cures on the table while you form subcommittees to discuss whether 168 00:18:54.020 --> 00:18:57.960 to form committees to review the proposal to consider looking at them. 169 00:18:59.160 --> 00:19:08.400 The Bottom Line. You have tested zero point three four two percent of the drug-disease combinations testable with compounds 170 00:19:08.400 --> 00:19:16.640 you already own. If you include the larger chemical universe: roughly one-hundred-quintillionth to one-hundred-tredecillionth 171 00:19:16.720 --> 00:19:21.260 of what's possible. The reason is not scientific. The reason is financial. 172 00:19:21.680 --> 00:19:29.360 Trials cost too much and the money is busy being missiles. A decentralized, automated pragmatic trial system is the only 173 00:19:29.360 --> 00:19:35.660 route to systematically explore the remaining ninety-nine point nine nine nine percent of medical space. 174 00:19:36.700 --> 00:19:39.820 The cures exist. The molecules are sitting in your cabinets. 175 00:19:40.300 --> 00:19:46.680 The patients are willing. The only thing between them is a system that charges one hundred million dollars to introduce 176 00:19:46.680 --> 00:19:54.720 them and takes seventeen years to make the introduction. On Wishonia, we have a word for holding sick people hostage until 177 00:19:54.720 --> 00:19:59.410 someone pays a ransom. We call it crime. On Earth, you call it "regulation." 178 00:20:00.420 --> 00:20:01.840 You seem very proud of it.