Ep: 4 The Future of Medicine

Eric Schmidt
September 15, 2020
Listen to this podcast on your favorite platform!
September 15, 2020

Ep: 4 The Future of Medicine

How will advances in biology open up a new era of medicine? Two trail-blazing scientists, Aviv Regev and Jim Collins, discuss their breakthrough work in cell biology, inventing new viral tests, and harnessing advanced computing to unlock new frontiers in the fight against disease.

Our guests are two renowned scientists working on research ushering in the next era of revolutionary medicine. They're making advances that won't just aid our response to COVID-19, but fundamentally change the way we approach health.

Episode Transcript

Jim Collins (00:05):

Our mission for the [inaudible 00:00:06] products is to develop these learning platforms to enable us to discover or design seven new class of antibiotics against seven of the deadliest [inaudible 00:00:17] pathogens over the next seven years. Just prior to the pandemic, we actually showed that this was feasible.

Eric Schmidt (00:25):

The coronavirus pandemic is a global tragedy, but it's also an opportunity to rethink the world, to make it better faster for more people than ever before. I'm Eric Schmidt, former CEO of Google and now co-founder of Schmidt Futures, and this is Reimagine, a podcast where trailblazing leaders imagine how we can build back better. In this episode of Reimagine, we're going to discuss innovation in the world of health sciences. Looking back 100 years, we can see how far technology has come, but the tools we've been using to fight this pandemic, especially early on, have been largely unchanged.

Eric Schmidt (01:21):

Our guests are two renowned scientists working on research ushering in the next era of revolutionary medicine. They're making advances that won't just aid our response to COVID-19, but fundamentally change the way we approach health.

Eric Schmidt (01:35):

We'll start the episode with Aviv Regev, who will help us understand the disease of COVID-19 and the role vaccines will play in overcoming it.

Eric Schmidt (01:44):

Then we'll talk with Jim Collins, a pioneer in the field of synthetic biology, who teaches at MIT, who will share with us some of the most exciting and innovative testing methods in synthetic biology.

Eric Schmidt (01:55):

Both Professor Regev and Professor Collins work could be game changing, not only in the fight against the novel coronavirus, but for all future battles against disease.

Eric Schmidt (02:06):

Our guest, Professor Aviv Regev is a legend in the field of computational biology of Aviv is a professor of biology at MIT and the director of the Klarman Cell Observatory and Cell Circuits Program at the Broad Institute at MIT and Harvard. She's also the newly named Head of Research and Early Development at the biotechnology company, Genentech.

Eric Schmidt (02:26):

But interestingly, over the last few months, instead of working on that, she mobilized the entire Broad team under her direction to switch gears and address the coronavirus. They use their knowledge and networks and so forth to rapidly learn about this new virus. Professor Regev, welcome.

Aviv Regev (02:43):

Thank you, Eric.

Eric Schmidt (02:45):

You've done fundamental research in how the disease works. What's the truth of this virus? How dangerous it is. What does it do to people?

Aviv Regev (02:55):

SARS-CoV-2, which is the virus that causes the disease COVID-19, varies in how it's dangerous for different people, Eric. Some people will have a very mild disease and possibly will never show any symptoms. Whereas other people, for reasons that we don't fully understand yet, will develop a much more severe disease. In some cases, they will just feel really bad. They would be at home. It could take them weeks to recover. But at the end of the day, they will become healthy again.

Aviv Regev (03:21):

In other cases, the disease will deteriorate substantially, and then things happen in all sorts of ways. Some patients will develop a lung condition known as ARDS, because the virus will get deeper into their lungs and not just into their nose and airways. In other individuals, actually, there will be a failure in other types of organs, it can be the heart, it can be the kidney or the liver in some patients, or even neurological symptoms.

Aviv Regev (03:48):

There's all sorts of additional things that happen in the body. We now understand that there could be events involving blood flows in our blood vessels. We don't really know the full ways in which the virus can wreak havoc on the body. That's one of the mysteries of this virus, that it seems to reach very far and wide, but not in everyone, and the disease course actually differentiates between different individuals. There's a lot of scientific and biomedical mysteries for us to resolve so that we can treat patients better.

Eric Schmidt (04:20):

A reasonable presumption is that this disease will be with us for the next year or two. There's some-

Aviv Regev (04:26):

It will be with us forever.

Eric Schmidt (04:28):

I see.

Aviv Regev (04:30):

Even if we have a vaccine, it will not necessarily lead to the disease going away, a vaccine will simply allow us to control it. For most of the diseases for which there are vaccines now for many decades, the disease itself is not gone or rather the virus is not gone. The virus is in the population, but we have a way of controlling it and avoiding at least the difficult manifestations of disease. When we're particularly lucky, we have a way of avoiding disease altogether.

Eric Schmidt (04:58):

But these viruses have existed for a very long time. As a computational biologist, why haven't you been able to figure out a way to eliminate all viruses?

Aviv Regev (05:07):

Well, you're fighting evolution and that's really a losing battle. I'm not entirely joking about that. There are several diseases where you are fighting evolution. One example is in infectious disease, because there's a very rapid evolutionary process that happens to both viruses and to bacteria, which are another type of infectious agent. It's also true in cancer. That's another disease where you fight evolution because the genome keeps changing for cancer cells and so whichever medicine we bring in, there's so many different cancer cells, so many viruses, so many bacteria, that of course they can mutate and change, and in this way, avoid the therapy.

Eric Schmidt (05:48):

A quick thought for you to consider. Let's imagine a future pandemic, sorry to be the bearer of bad news, but there will certainly be one in the future, and let's imagine a global competitor, China, not only invents the solution, but keeps it to themselves. How would we feel if leadership in something as fundamental as biology, disease, and medicine was transferred to another company, another country, what have you? It's crucial that America invest in the kind of innovation that our guests have talked about here, because without it, we don't drive our industries and we don't drive our healthcare.

Eric Schmidt (06:26):

What's amazing to me is with all of the investment of hundreds, of years of health and biology and medicine, we still can't build a full digital model of a cell. We still don't have a full model of a brain. Even of a microorganism. Biology is really hard.

Eric Schmidt (06:44):

In the case of synthetic biology, there've been a number of discoveries where people were just thinking, well, maybe there's some promising thing here. What we need to do is marry American and scientific curiosity with a funding source and a structure where experimentation can occur.

Eric Schmidt (07:00):

The Broad, of which I'm a board member, was founded almost 20 years ago in order to do fundamental research in initially the human genome and now disease progression and how these cellular things interact. From my perspective, the most important thing that Broad has done is it's established a mechanism of funding through private philanthropy, in combination with a set of healthcare systems and two universities where the faculty have a reasonably high likelihood that they're going to get the funding to do the kind of experiments that will yield to real discoveries in genetics, disease, and disease processes.

Eric Schmidt (07:38):

The question, if you're a computer scientist is, what is the next big thing that's going to happen? The information platforms are now largely understood, but the biggest area of unknowns is biology. We can see a certain amount into how biology works, but we have orders of magnitudes of interactions that we don't today see. The only way we can see them is with incredibly accurate diagnostic tools and the use of sampling and machine learning.

Eric Schmidt (08:07):

In this next part of the conversation we pulled back from COVID specifically and spoke about some of Professor Regev's incredibly innovative work. It has repercussions, not just for the current pandemic, but healthcare overall. Now back to the conversation.

Eric Schmidt (08:22):

Before all of this happened, you led, or was a co-leader, of the Human Cell Atlas Project, which was the next great information platform for biology. I was surprised to think that we didn't already have one. It would seem to me that we have humans, we have cells, we should know how many we have. Seems like an obvious question. How many cells do we have? What are they like? Why is the Human Cell Atlas so important?

Aviv Regev (08:54):

We have about a 37.2 trillion cells in an adult human body. That's an estimate. It actually was a contentious estimate, but it seems to have stabilized around that number. That's a very big number. Of course, they're each, in one way or another right now, different from each other, but by and large, they come in different categories that we call cell types.

Aviv Regev (09:16):

If you open something like, I don't know, Wikipedia, it might tell you that there's 300 different cell types and it's not wrong, it's just not very granular. It's not very high resolution. These are things like you have neurons versus you have T cells and B cells, which are kind of immune cells, versus you have adipocytes, which are kind of a fat cell. Each of these is a cell type, but there's a lot of additional diversity amongst our neurons, probably thousands of different kinds of neurons, so just saying that there's neurons is a little bit crude.

Aviv Regev (09:50):

You're right that we should have had something by now, but there's a good reason why we have. First of all cells [inaudible 00:09:57] discovered in the 1600s and they're a result of a technological revolution known as the microscope.

Aviv Regev (10:03):

In the 1600s, Hooke, looking under the microscope, really realized he was looking at plant cells at first and they actually look like squares, which is why he called them cells He was reflecting on the cells in which monks in the monastery actually live.

Eric Schmidt (10:18):

Very interesting.

Aviv Regev (10:20):

[crosstalk 00:10:20] of the word. But also was fascinated with them ever since. I tend to tell people that this was an endeavor, at least for the last 150 years, since our microscopy roles and our ability to stain and look at cells under the microscope improved to map the cells of the human body and the cells in biology in general. It's always been driven by technology.

Aviv Regev (10:42):

The challenge that we had, is that we didn't have a good system of coordinates. There wasn't a unifying view to look at cells. One person might describe them by their molecules, another person might say here is where they are, another one might talk about the shapes that they have, and in another case, we might talk about their functions. These were all legitimate descriptions, but they weren't in the same set...

Aviv Regev (11:03):

And these were all legitimate descriptions, but they weren't in the same set of coordinates. Each time it was in a different way. And then on top of that, we didn't have great ways that were both scaled and data driven to define them because our way of hooking at them couldn't always look at them one cell at a time. It might need to look at a million or a billion of them all mushed up together before you could actually, for example, look at which molecules they are using or which genes they express out of our genome.

Aviv Regev (11:33):

And so that was the time of another technological advance, which is about, I don't know, seven, eight years old now, maybe a little older, which is called single cell genomics that allowed us for the first time to look at individual cells and for each individual cell to look at all the RNA molecules, the parts of the genome that the genome actually uses it in a particular cell to do its business. So it has the gene, the gene expresses RNA, from the RNA come proteins and the proteins are the business end of the cell. So knowing which genes are active, all of a sudden became possible at the level of individual cells.

Eric Schmidt (12:11):

So, will this atlas allow us to cure some diseases that have bedeviled us for thousands of years?

Aviv Regev (12:17):

So we definitely hope so. And this is because the [inaudible 00:12:21] happen in tissues that are made of many different kinds of cells. And I'll give you maybe a couple of examples of this. So one example would be in cancer. Cancer is made of many different kinds of cells in a tumor. So some of the cells that make up the tumor mass are actually cancer cells. They're malignant, they're mutated, and they're not all identical. They're different from each other. And often when we have a therapy, it might be targeting one of these kinds of cells, but it won't be targeting the others. And that gives the tumor an escape route. So the cells that are susceptible to the therapy will die, but the other ones can proliferate and grow and take over. And as a result, the tumor will come back.

Aviv Regev (13:05):

At the same time, in our tumors are many other kinds of cells. They're not cancer cells are cells of our healthy body that are either coming and trying to eradicate the tumor or are lured actually by the tumor cells to come in and help them. So a tumor needs blood vessels, it needs oxygen and nutrients and blood vessels coming in order to treat the tumor inadvertently. They don't want to do it, just the tumor lures them in. There are immune cells that come in in order to kill cancer cells, but cancer cells try to kill them and exhaust them and make them tired and ineffective so that they won't be able to kill it.

Aviv Regev (13:41):

In the past, when we tried to look at cancers, we looked at all of those cells at once. Put them into the blender, got some average out. Now that we can see the individual cells in the tumor, we're increasingly able to say, "Oh, here are things that we should be targeting and we didn't even know were there in the patient." Or we can say, "This patient's tumor and that patient's tumor are actually not the same." And we should be targeting that tumor with one drug and the other tumor with a different drug. Because on average they might look quite similar, but in the detail, in the high resolution view, they're different. So that's one example of something that has befuddled for a long time and where single cell genomics and access can be extremely useful.

Eric Schmidt (14:23):

In my final question, Aviv, not deterred by the extraordinary problems and difficulty of building an atlas of us, you've now embarked on a new idea. And you have a new idea to look at the language of how biology works. Can you explain what a language of biology would look like?

Aviv Regev (14:44):

Yeah. Well, that's a challenge for the next 20 years or so to fully decipher it. But I'll try and give people a little bit of an idea of why it's difficult and at the same time, why we're so close too. So one of the things that really characterizes biology is that the whole is more than the sum of the parts. Because we sometimes say it's not additive, or it's not linear. It means that if I know what, say, one gene does and so I can predict that if I take away that gene, maybe something would happen to a cell or an organism. And I know what another gene does and I can predict if I take that gene, something else would happen. I actually can predict what would happen when I take both of them out at once.

Aviv Regev (15:31):

And to know the language of any [inaudible 00:15:35], we need to understand that. That's really the semantics of the system. It's the explanation of what is actually going on. And so we want to be able to reach a point where, by looking at our genome and knowing the genes that are there and looking at our environment and knowing the factors that are there, we're able to say, "This is how a cell responds to it." And not just this is the chain of events, but we'd be able to say, "And we can predict what the cell will say next. We understand what the cell is talking about by metaphor." And that means things like when a cell is challenged by a virus like COVID-19 or when I just had some coffee, or when I am walking around or exercising, this signal is perceived by cells through receptors. In response to it, we want to be able to say the cell activates certain types of programs, which are made of different patterns of genes that it uses.

Aviv Regev (16:36):

And these programs are like the words or the building blocks or the sentences of the cell. It puts them together in particular ways. And we can understand what they mean because we can predict what they do. Cells are all about doing, and our bodies are all about doing. And we can understand that when a cell wants to do something, it has to use this program. And when it uses this program together with another program, such and such thing will happen to it. It would divide, it would move to another place. It would release some pieces of information that another cell will pick up and would respond to in particular ways, activating other programs and propagating downstream from that.

Aviv Regev (17:20):

Now, what we know how to do today is how to observe the system really nicely. We can measure a lot of things about our cells and about our tissues and see exactly what they're doing at certain moments in time. But those are just description. We can't actually stay by this, "This is what it is going to do next." We we can say, "This is what it actually does." And what we want to be able to do is to take enough of these observations and using computational and algorithmic approaches, for example, from machine learning, move to a level where we say, "And this is what it means and we can predict what would come next." Or, if we wanted to do a certain thing, we want our cells to divide or to die or to change in any way, we understand its language, we can talk back to it. We can intervene in the cell, take something out, put something in and make it do what we wanted it to do.

Aviv Regev (18:12):

So why is that such a difficult problem? So it sounds like we should have been able to solve it by now, just like with the atlas of the cell. Well, the reason that it's such a difficult problem is that when things are not additive, you can't just add them up, we started at least, we thought we had to measure all of them. So if you have 20,000 genes in the genome and every combination of them might do something different, and you had to test all of these combinations, that's never going to happen. That's more than the number of not just cells on the planet, it's actually more than the number of atoms in the universe.

Aviv Regev (18:48):

So you will never be able to do all of those experiments. You have to find a way of looking at enough examples and predicting the rest. Looking at enough examples and learning the language, just like a child listens to enough examples. They haven't heard all possible books, they haven't heard or possible sentences that people would say, but they would still understand the next sentence just by listening to a large enough number of examples.

Aviv Regev (19:15):

And that's what two things that have happened in the world will allow us to do. The first is the amazing advances in machine learning and our ability to infer from data. And the second is the great advance in experimental biology that allows us both to intervene in biological systems, for example, by using CRISPR to knock out genes and to measure biological systems, for example, by [inaudible 00:19:41] many millions and millions of cells. This would give us enough examples from which we can learn this language. We can find the sentences and the words and so on. And as a result, we'd be able to predict the next sentence or be able to actually speak to the cell in its own language.

Eric Schmidt (19:57):

In your 15 years at the Broad, you built the Broad into the world's premier research institution in biology. As our listeners can hear, you're also one of the most acclaimed teachers of biology that I know. Thank you so much, Aviv.

Aviv Regev (20:12):

Thank you, Eric.

Eric Schmidt (20:12):

While Professor Regev is helping advance science by giving us a look at the building blocks of life in more detail than ever before, our next guest is one of the leaders applying that knowledge in exciting new technologies. Jim Collins leads research that sometimes has the feeling of a scifi thriller. Professor Collins is a professor of bio engineering at MIT, spending his career pushing the boundaries of biology. He joins me today to talk about his quest to discover new antibiotics, to synthesize better tests and to reinvent the future of medicine.

Eric Schmidt (20:47):

Thank you, Professor Collins for being here.

Jim Collins (20:50):

Thanks for having me on your podcast.

Eric Schmidt (20:52):

You followed the path of a Rhodes scholar athlete trying to make the world a better place. How did you end up in synthetic biology? And maybe we should start by synthetic biology is in fact what?

Jim Collins (21:07):

So synthetic biology is still a relatively new field. It's bringing together engineers with molecular biologists to use engineering principles, to model design and build synthetic gene circuits and other molecular components and to then use these circuits and components to rewire and reprogram living cells, to endow them with novel functions to impact various aspects of the economy and broader humankind around the world.

Eric Schmidt (21:38):

Now it's possible to build synthetic life. What are some examples of synthetic life that people are building that are either good or scary or both?

Aviv Regev (21:48):

Yeah, there are many, many good examples. So, I'm involved with a company that was spun out of my lab and Tim Lu's lab called Synlogic. And Synlogic is building upon some of these very early efforts in synthetic biology to take engineered genes-

Jim Collins (22:03):

To study biology, to take engineered gene circuits and engineered pathways, introducing them into bacteria and reprogramming those bacteria to serve as living medicines. So the company has multiple programs underway, including engineering bacteria to address rare genetic metabolic disorders. So many kids are, for example, born without enzymes to break down certain metabolic toxins, Synlogic is engineering bacteria using some of our technology so that it would break down those toxins, replacing lost enzymatic function.

Eric Schmidt (22:34):

So you use the bacteria as a sort of a manufacturing plant?

Jim Collins (22:38):

In this case, it would be a processing plant, in that it would be kind of a waste processing plant. So it would take otherwise waste products out of metabolism that a healthy individual would break down. These would be introduced into kids to replace their loss function. So they don't have the enzyme to break down these toxic molecules that can build up and cause real harm, in some cases death. But they are also producing manufacturing plants, kind of, drug manufacturing delivery units, for example, to address inflammatory bowel diseases, such as Crohn's Disease, and also colitis where they're engineering E. coli to introduce immunomodulatory molecules at the site of inflammation in the GI tract.

Jim Collins (23:23):

So it's opening up a whole new class of medicines that are really quite exciting, that could address in, in some cases, some very, very complex diseases.

Eric Schmidt (23:32):

I know quite a few people who will benefit from your discovery, that's fantastic. At some point in this process, you started looking at Ebola, sort of the last frightening thing before COVID, what did you do in Ebola that was different?

Jim Collins (23:45):

What we did there was really building on remarkable creative lead made by then one of my postdocs Keith Pardee. So I had encouraged Keith to begin looking at cell-free extracts. What these are is that you can basically go inside a living cell, take the machinery of that living cell outside of the living cell and play with that machinery, which could be DNA or RNA, ribosomes, and other molecular machines, as well as ATP and amino acids and play with these in Petri dishes and test tubes. This has been done for decades in molecular biology and cell-free extracts, for example were used to work out the RNA code in the 1960s.

Jim Collins (24:23):

I had encouraged Keith to look to see could we engineer liposomes or vesicles that could encapsulate cell-free extracts along with synthetic gene circuits to create synthetic sentinels, that could be placed into a person's body to serve as living diagnostics and living therapeutics. For reasons I still don't understand why he did it he decided to see what would happen if he spotted cell-free extracts along with synthetic gene circuits onto paper, freeze drying them. What he found was that he could freeze dry cell reactions along with synthetic gene circuits, and then sometime later rehydrate them with water or blood or urine and they would become reactivated and they would function as if they were inside a test tube or inside the living cell.

Jim Collins (25:06):

When he showed me that I got very excited and realized this would open up so many new real world applications, including low cost diagnostics. This was in 2014 and we were initially focusing on diagnostics for antibiotic resistance, given a very big focus of our lab is around discovering new antibiotics and trying to better understand how resistance arises. But in mid August as we were preparing and finalizing the paper for submission, Keith Pardee, along with another postdoc, Alex Green came running into my office and they said they wanted to hold up our publication. I challenged them why and they said, we want to see if we can develop paper-based diagnostics for Ebola.

Jim Collins (25:45):

What they did was remarkable, in 12 hours they designed, built, tested, and validated 24 different senses for Ebola. 12 for Sudan strain, 12 for the Zaire strain. They could differentiate those strains and get an output in roughly an hour with sensors that in the end only cost 2 cents per task.

Eric Schmidt (26:05):

That's incredible.

Jim Collins (26:07):

So we are tremendously excited about what this opened up and really the key advances that we are then enabled to harness the power and diversity of biology outside of the laboratory, without needing to refrigerate biological material. By freeze drying these, you could then stick them in your pocket and back down, we can still travel you can travel around the world, take it out and use it or send it around in just a regular envelope and even at any site in the world store it properly in your guest drawer and then take it out anywhere for a year or two later, and it still would be functional. So it opened up a whole new class of inexpensive.

Eric Schmidt (26:41):

Let's move to the COVID pandemic, which is getting worse, unfortunately, pretty rapidly. I was reading something where you said what we need for any outbreak are data, data on who's been exposed, who's infected, so you can take measures to isolate those individuals and reduce the spread. This virus is insidious in part because you can be infectious several days before you're symptomatic. How do you propose to address this? We can't see it. We don't know we're infected. We don't want to get other people sick and people are dying.

Jim Collins (27:12):

We have to address it on multiple fronts. I do think science and technology is going to help get us out of this current pandemic and is going to better prepare us to the next one. So I do think synthetic biology has great offerings, particularly on the diagnostic front, and we are working as hard as we can to address this on several counts.

Jim Collins (27:33):

So we have built on our cell-free freeze dried efforts in synthetic biology to create a range of different diagnostic tests, including applying it to CRISPR. CRISPR was initially developed for gene editing and it was applied in the biomedical space initially for therapeutics or in the context of gene therapy modifying genetic diseases. We teamed up with Feng Zhang from the Broad Institute at MIT and created a platform that we call Sherlock that enables one to use CRISPR as a diagnostic platform.

Jim Collins (28:06):

Feng and I, along with several of our colleagues launched a company called Sherlock Biosciences now going back about two years ago. Beginning days of the pandemic, Sherlock, to its credit, pivoted its entire company towards addressing challenges with the pandemic and actually have an FDA EPA so that the approved CRISPR test that's on the market, a partnership with IDT to manufacture to a million of these tests. Quite soon, they're going to announce a partnership to have a point of care test out there.

Jim Collins (28:36):

So here are the tests that can give you an output in under an hour, costs about $25 per test. In each case, we need to do better, we've got to get faster. We need to get a cheaper and within my lab at MIT and Harvard, we're working on both of these challenges, really with two different efforts, but using the same technology.

Jim Collins (28:57):

One is around a face mask diagnostic. So we are working and about to finish a prototype where we've embedded freeze dried cell-free extracts along with freeze dried synthetic [inaudible 00:29:10] CRISPR components into inserts that can go into a face mask. The idea is that you could wear for an hour or two, when you breath, when you talk, when you cough, when you sneeze, if you're infected you give off a good number of viral particles and this system would be set up to produce a fluorescent signal that could be detected with a inexpensive handheld perimeter from our expensive test if it's a point of care, quite a bit more expensive device that can give you an output very, very quickly. In really an at home setting, if you're infected.

Jim Collins (29:44):

Related, we're using basically the same technology to try to produce inexpensive and rapid at home saliva based tests, both to indicate viral particle as well as to indicate antibody tests. So to get after more data, we need quicker tests, we need cheaper tests, mean more testing. I think we've been flying blind since the beginning of the pandemic, and only now are starting to get after some of the measures that could be used.

Eric Schmidt (30:12):

We're fighting a war where we can't see the bullets. We talked a little bit about your research, and you've also started a fairly large machine learning group essentially. For our listeners machine learning is basically where the computer learns patterns and it can see patterns that are deeper than humans' can. So you can basically see something that's a needle in a haystack, much easier if you're a computer machine learning rather than a human doing the same thing. What are you using machine learning for?

Jim Collins (30:40):

So we, at MIT, have just launched what we've called the Antibiotics AI Project. This is an effort in collaboration with Regina Barzilay, one of my professor colleagues at MIT and a world leader in applying AI to healthcare. Prior the pandemic for Regina and I had collaborated along with Tommy Jacqueline and other MIT faculty to create a deep learning platform that would enable us to address the antibiotic resistance crisis, which is one of the existential crises facing humankind.

Eric Schmidt (31:09):

So just, we are more exposed to existing things because the antibiotics don't work anymore.

Jim Collins (31:16):

That's right. We've had antibiotics, or penicillin was discovered just a little more than 90 years ago and then put into manufacturing about 15 years after that. So we've had antibiotics now for many, many decades, and we've lived in the age of antibiotics, but due to the overuse of what we have resistance has grown and thus these bugs now are no longer susceptible to the antibiotics.

Jim Collins (31:40):

Unfortunately the antibiotic market is broken in that either the antibiotics are priced at too low an amount to recover the R and D efforts and or they're shelved in order to be protected and as a result, biotech and pharma have gone out of business because it costs just as much development antibiotic, as it does say, a blood pressure medicine, which you could sell to somebody for the rest of their life, well antibiotics you take for five to seven days. So our pipeline has been drying up.

Jim Collins (32:09):

So Regina and I took this on and thought about applying deep learning to take this on. Our mission for the Antibiotics AI Project is to develop deep learning platforms to enable us to discover and design seven new classes of antibiotics against seven at the deadly spectro pathogens open the next seven years.

Jim Collins (32:31):

Just prior to the pandemic, we actually showed that this was feasible. We developed a deep learning neural net against training data. We collected applied it to the Broad Institute Drug Repurposing library, and identified a molecule that looked quite different from any existing antibiotic, that we termed halicin. That is remarkably poned against multidrug resistance, extensively drug resistant and pan resistance bacteria, including C Diff a nasty gut pathogen, as well as Acinetobacter baumannii, which is known as the-

Jim Collins (33:03):

... Nasty gut pathogen, as well as acinetobacter baumannii which is known as the Iraqi bug and is it the top of the WHO's list of deadly bacterial pathogens, which we need new antibiotics and it was infective against TB. And so we are very excited at where this can go. And we are doing two things in the context of the pandemic around this platform. One is, and what may not be appreciated by many, including the listeners of this podcast is that bacterial infections play a major role in this current pandemic.

Jim Collins (33:33):

One out of seven COVID-19 patients requiring hospitalization have a bacterial co-infection. 50% of those who die have a bacterial co-infection.

Eric Schmidt (33:42):

Mm-hmm (affirmative).

Jim Collins (33:43):

When you look back on the Spanish flu in 1918, which we hear a lot about in the midst of the current pandemic, it was very deadly, largely because of bacterial co-infections. And those were the days before we had antibiotics. And so we are harnessing our platform, our deep learning platform to go after narrow spectrum antibiotics that can treat lung infections. And we're now repurposing the platform, which was initially set up as a screening tool to look at large libraries to now be a design tool. So we're now using the deep learning model to learn features about molecules that could make good antibiotics to design, to Nova new molecules.

Eric Schmidt (34:27):

You started with the data that was already existing in the Brode system. And you came up with halicin. Now you figured out a way to take those tools and turn them into a way of trying other candidates.

Jim Collins (34:33):

That's right. So we actually did two things. One is we took that initial screening tool and went even beyond the Brode data set to look at the zinc database. And the zinc database has 1.5 billion molecules, just enormous. Couldn't empirically screen that many. But we were able to screen 10% of those that we thought could make for good antibiotics in just three to four days, on a computer and identified several hundred candidates that were promising. Tested a couple of dozen and identified eight potential new antibiotics. We've since now modified the platform from a screening tool to now a generative tool, where you can make effectively any molecule in principle. And we are now synthesizing a number of promising candidates to go after COVID-19, rather than lung infections.

Jim Collins (35:19):

The second rotation we did on the platform was that this deep learning approach is not specific antibiotics. It's truly indication agnostic. And prior to the pandemic, we were initiating efforts to go after cancer approaches and see if we can develop oncology drugs. But in the midst of the pandemic, we then pivoted toward antivirals. And we are now applying the platform towards screening data that have been collected by other groups around the world, and we've also launched our own screening efforts. And in most cases, people are screening the same molecules. Everybody's looking at repurposing existing drugs to try to get to faster approval. And that's been suboptimal from a diversity standpoint.

Jim Collins (36:03):

We are now looking to see how you could use deep learning approaches to get after combination therapies. And I think that's going to be the real contribution for antivirals. In many cases, single shot molecule wonders are very hard to come by and in antiviral therapy, HIV, for example, the most successful therapies are cocktails. And I think the same is going to hold true for SARS COVI II.

Eric Schmidt (36:29):

You know, Professor Collins, this has been the most optimistic thing I've heard in weeks. Not only are you going to, hopefully by the end of the year, have serious progress in terms of rapid testing. And in fact, the possibility of detecting that one has become dangerous enough to be able to unfortunately infect others. But this notion of seven classes of antibiotics and seven classes of the world's deadly bacterial pathogens over the next seven years is a truly audacious goal. I really wish you the greatest of success. And I know the world counts on you. Thank you very much.

Jim Collins (37:02):

Thank you, Eric. Really appreciate it.

Eric Schmidt (37:06):

So where are we now? Re-imagining health care means creatively using the remarkable new tools and technology. This pandemic has engaged the creativity and tenacity of our medical community to the fullest extent. They are doing heroic work. I'm inspired thinking about all the lives that could be saved by these new discoveries.

Eric Schmidt (37:28):

Problems in science are hard. Diseases are hard. We celebrate the lone inventor, the Edison, if you will. But there were hundreds of people whose work he used to invent the light bulb. Example after example of, it's a community sport. In the case of Aviv Regev, she is one of the first people who combined both digital and computer science skills with biological skills in a seamless way. She saw that if you took biology, which is an analog, I call it squishy world. And you converted that to a digital world, you could use the power of computers to see things that the scientists with microscopes could not. With Jim Collins, what he did famously aside from being an inventor of synthetic biology, was he thought that maybe we can use artificial intelligence to suggest new kinds of antibiotics in new ways that we, as humans have not thought about.

Eric Schmidt (38:21):

And he and his team trained such a model. That's a breakthrough, not just for antibiotics because who wants to get sick? But more importantly, it's a breakthrough in the application of AI to complicated systems. Science funding in the United States is at the lowest point it's been since 1957. 0.7% of our GDP is science funding. That's the number before Sputnik. It's been as high as two to two and a half percent. We are not investing in our future in the way we should be. Philanthropists, corporations are to make up the gap, but there's no alternative for longterm stable, federal and state funding for our universities.

Eric Schmidt (39:05):

Everything that you listen to, every thing you use, everything you take for granted was ultimately started by some kind of a research grant way back when that ultimately went to a set of typically young people who were trying to do an experiment, or they had a strange question, or it just came to them that they should experiment. Just glimpsing a sliver of these new approaches with our two guests has been a heartening and motivating experience to me. It reflects the spirit of what I'm trying to do with SMID Futures, betting early on exceptional people making the world better. And that work will result in a more resilient and precise system that will catch or disarm a future pandemic before it spreads.

Eric Schmidt (39:45):

As a reminder, don't forget that we're still accepting submissions for our Reimagine Challenge 2020, which includes up to $1 million in scholarships and prizes. It's a chance for students around the world to share how they would reimagine the future. The deadline is September 25th, 2020. For more information, visit re-imaginepod.org/challenge. Next week, I'm really excited to speak with JD Vance, investor and author of the acclaimed memoir, Hillbilly Elegy, about rural America and re-imagining the US economy to increase shared prosperity.