GF2045(2013) Proceedings: Dr. George Church
Lincoln Center, New York City, June 2013
Global Future 2045: Towards a New Strategy for Human Evolution, Congress Proceedings
Bionanotech for extending Moore's Law, the BRAIN project I/O & human genome engineering
Dr. George Church
Professor of genetics at Harvard Medical School,
Director of PersonalGenomes.org
Abstract: We have developed a variety of CRISPR devices - protein-RNA-DNA complexes - enabling human genome - and epigenome - engineering with 20-fold higher efficiency and >100-fold easier programming than previous methods. To test these devices, PersonalGenomes.org provides the world's only biobank of human cell lines consented for fully open access sharing - and already outfitted as a sophisticated human synthetic biology chassis. We have designed and tested the first nanorobots made from hybrid materials - DNA, proteins and inorganic - which have sensors, logic & actuators capable of distinguishing subtle differences among various cancer, normal and immune cell types. We are exploring the use of similar hybrid nanostructures for manufacture of ultra-fast and complex electronic, optical and quantum computing and have demonstrated bionano storage a billion times more compact and with lower copying energy than conventional digital media. Such nanodevices offer significant advances in our ability to perform highly parallel input and output in animal and human neural systems.
Transcript
So, I have many people to be thankful to and institutions as well. This is my conflict of interest slide. We will be talking about bio nano tech and comparing it to many other technologies. So, from today and yesterday, one gets the impression that this is not really an audience. This is more like a congregation. And we were talking about the church of testable hypotheses, so if you open your hymnal to page 241, there is this chart of augmentation past and current, where we clearly are no longer natural species; we go much faster in many ways, we've extended our senses and our abilities, and in particular, our memory extends more than just a few minutes, in some animals to an entire lifetime, to thousands of years, and all of written human knowledge available.
But there are perils to the kind of exponential extrapolation that many of us here are addicted to. This is two curves, actually, the electronics industry, the so-called Moore's law in dotted lines; and the green curve is the molecular and genomics curve. And it was predicted that it would take about six decades, going on a very aggressive Moore's law type of exponential to get to an affordable human genome from the $3 billion that I helped initiate in 1984. That could've flatlined when the genome project money disappeared. And the lesson that I'm taking is not so much that it's hard to predict the future, it's more about how we connect the dots. Not just how we predict the dots but how do we actually bring these exponentials to occur.
Now fortunately, this curve did not flatline, as you've seen in one of Ray's slides. It accelerated. And instead of six decades, this arrived in six years, so I may be one of the few people on the planet that calls Ray Kurzweil a pessimist. And furthermore there is a tendency for us to put great faith in electronics. Ray correctly mentioned molecular as the next wave for it is possibly replacing electronics. We need to get down to smaller and smaller features and that's where molecules come in. And as part of this revolution I'm going to show you two slides where the molecular technology is already six logs ahead of the electronics and as you can see from that previous slide it's accelerating, it's going much faster than Moore's law – tenfold per year on occasion, many years in a row, rather than 1.5 fold. And here's an example where the density is higher, and in the next slide I'll show you where the energy efficiency is six logs higher.
Here we took a book, a neurogenesis book, and encoded it into digital DNA and this was about six logs better than any of the commercially available consumer storage devices, including flash memory, and disk drives, and so forth. That's density and you can see me down there in the lower left the actually putting on gloves and doing experiments. That's the fun part of my field, is you can still do it. Now, this is a progress curve for energy efficiency and you can see we are making progress at roughly the same rate they were making progress on, density of large-scale integrated circuits, but still there is more than 1 million-fold gap between what the best energy efficiency is for the circuits and the already high efficiency of molecular devices, you could call these nano robots of you will, but these molecular devices like DNA polymerase. And I'll show you a couple of of examples of what we can do with these molecular machines like DNA polymerase, in just a moment, but keep this in mind that we're six logs, at least, ahead, and it's going faster.
Now here's an example of a molecular computer - exactly what this person looks like is not quite so relevant as what he did with a molecular computer inside of his brain. And it's not quite so important what he did in such a short period time: speed was not the issue, it was the qualitative nature of what he did. This was a huge number of completely paradigm-changing physics papers that were written, five papers in little over half a year, and the two bell curves there represent clearly in some regards, he is off way on the edge of the bell curve in certain computational senses, human computation.
What if he were in the middle of the bell curve, there were a new bell curve made up of of human beings, or molecular computers if you will, where he was average? What would the extreme look like in that case? This is a very interesting thing to reflect on and you can see the Jeopardy computer, Watson, is nowhere near this particular individual, and it may be qualitatively different rather than merely being able to remember more, I mean the fact Einstein did not have access to 200 million webpages, but he still did a pretty good job as a 26-year-old.
When we look for exceptional individuals, it's not just about computing, our ability to think, it's also our ability to live a long time because it takes us a long time to accumulate these important memories. Here are some individuals, and you shouldn't take home from this slide that the way to live over 110 years is to smoke and drink, but the point is that no matter what the environment is, no matter how many infectious agents they've been exposed to, and other risks that epidemiologist tell us all humans reveal, they have potentially protective genetic and environmental components, and we need to define these, and we want to be able to share this information and possibly as therapeutics.
Now those super-centenarians, centenarians that are over 100, 110 years old are being sequenced right as we speak. We don't yet know what rare protective alleles they may have, but we do have some examples of rare protective alleles that are already understood in some detail. For example, myostatin, MSTN, double nulls, meaning knocking out both copies of what would be from your mother and your father, results in people with very large lean muscle mass and very low probability of atherosclerosis.
These are very rare. Maybe a handful of people on the planet are known to have this, and they're healthy as far as I know in every other regard. LRP5, if you ever seen Bruce Willis in Unbreakable, that's actually based on some rare individuals that have this heterozygous state where they have extremely unbreakable bones. They are eight standard deviations out in bone density.
PCSK9 is a mutation that's inspired the drug industry to develop a new type of anti-coronary artery disease. Because these individuals have such incredibly low levels of bad LDL cholesterol that some physiologists were surprised that they were alive, but on the contrary, they had extremely low coronary artery disease. CCR5 will get its own slide, but double nulls in CCR5 and double nulls in NTT2 are resistant to some very serious viruses, HIV and norovirus, for example. And certain versions of APP put you at risk for Alzheimer's, but the reason it's on this slide is there are certain versions that can delay Alzheimer's by a factor of 10 to 15 years.
So, all of these are rare protective alleles. They're are not the rare deleterious alleles that we normally hear about in genetics. Now we selectively breed organisms like dogs. There can be 1000 fold difference in the weight of adult dogs. We also selectively bred ourselves and so there's these big parts of the world where there's malaria resistance. We have persistence of the enzyme that allows us to drink milk well into adulthood. These are selectively bred, somewhat natural.
There are genetically engineered organisms is well, like these glowfish, and Timothy Brown is one of the first genetically engineered human beings. He got a stem cell transplant for the CCR5, which is a receptor for the AIDS virus. He had the unfortunate situation of both having leukemia and HIV. Treating his leukemia resulted in this treatment for AIDS, which is now generalized from that somewhat serendipitous situation to a molecular machine called zinc finger nuclease, that will go on and target both copies of your CCR5 gene so that your T cells – the cells the HIV virus normally infects in your immune system – don't have the HIV receptor and so they selectively survived this and you become cured. Not something that multidrug resistance and vaccine trials in HIV have not proven quite as effective as we'd hoped, but this is something where you literally remove the receptor. This is in phase 2 clinical trials, which tells us it's pretty far along, and it is curing AIDS patients and it's interesting in that not only is it more precise than any previous therapy, but it's aimed at, in a certain sense, augmentation.
It's not curing some rare disease, rare deleterious thing like cystic fibrosis or sickle cell, it's aimed at all of us, who are in a certain sense susceptible to the AIDS virus and other viruses. The next step in this is CRISPR. I'm not going to spend a lot of time on this but listen. It's like you know in the [graduate plastics?] for the word CRISPR. It's another way of doing this, it's about 100 times easier and about 10 times more efficient than the method that's already in phase 2 clinical trial, so this is going to allow human genome engineering on an unprecedented scale.
We can also make tissues. This is an example where in a mere four days we can take human skin derived stem cells. This is actually from my skin. They're turned into embryonic-like cells, but they're adult-derived, and then they're turned into these bipolar-like neurons going through the regular pathway. Now I should mention that those stem cells are derived from the Personal Genome Project. This is the world's only open access source for genomes, environments, and trait data, but not just data, interpretation software and the cell lines. These cell lines are broadly distributed internationally. The NIST – National Institute Standard Technology, and the Food and Drug Administration, have a unique new collaboration to produce standards in this field called Genome in a Bottle, and you can go to genomeinabottle.org to learn more about this; and they looked all over the world for suitable samples, genomes and cells, and they found this Personal Genome Project was the only one that was properly consenting individuals for this kind of open project where you could share it and make international standards.
This is an example of the annual meeting we have for the Personal Genome Project volunteers. I know many in the audience are in this. They're invited to come, and unlike most medical studies where anonymity is promised but not necessarily delivered, here they can wear name tags and say, “Hi, I'm PGP”. And it produces a much more participatory environment where people can share their information.
So, this is the genomes environments and traits. An example of a trait that's relevant to this conversation is functional magnetic resonance imaging. It's unfortunately a low resolution, but it is a whole brain and it is functional. It gives you information about what you're thinking in a way that you don't get from any other methods, and here's an example of one of the PGP volunteers in that previous slide with an MRI scan, and you might worry that these are not actual slices. This is my brain so thankfully they are virtual slices.
These are cartoons of actual slices of the mouse brain, which is part of the Allen Brain Atlas project. In this project, unfortunately, they wanted to visualize all of the molecular identification of all the cells in the brain, but they can only do one at a time so they'd have to do thousands of brains, each one getting one of them at a time and then try to integrate that data, but each brain is a unique story and so they don't really align that well. It's still a useful atlas, but wouldn't it be great if you could do these thousands of different molecular IDs or RNAs in a single brain and that way they're integrated, every voxel, every pixel of every section can be aligned and you know where you are. And not just for the RNAs and other molecular identifiers, but what the brain was thinking during the activity and we'll get to that in just a moment here.
So, there are hundreds of thousands of individuals in the world who have electrodes implanted in their brain already so this BRAIN initiative that we're embarked on, that Obama helped get going. is already in the clinics. This is a case of an epileptic who also contributes to experiments. In this case, on the top of the slide you see they happen to - with a small number of electrodes, such as the ones that Ed just described - they happened to find, by showing lots and lots of images, one set of neurons that corresponded to Jennifer Aniston images, and not any other images in the control down below.
Now this is simultaneously very inspiring and very frustrating in the sense that we now know that there's a very high level one. Not the lower-level ones that we've been talking about so far today, but a very high level cognition going on here, but it's frustrating we don't know how many other neurons were involved, what other images would produce, and how the whole thing fits together. For that we really need to measure every neuron and at appropriate speeds and long periods of time.
That's what the BRAIN initiative is about. It's not just about measuring them, it's also about stimulating them because we need, as Ted mentioned, we need to be able to deal with pathological situations, and we need, inevitably, to augment our capabilities of healthy individuals. Here's this long list similar to the one that Ed showed, and John Donohue's group has helped this woman who was tetraplegic for decades to now move a robotic arm and John and his colleagues hope to now make it so that she can move her own arm, and at high speed and in much less awkwardness than is currently possible.
This is from from Ed Boyden and his coworkers that he just showed you, where you can have arrays, but these arrays are not at the density of measuring all 80 billion neurons in in the human brain, or roughly 1000th that that number in a mouse brain. To do that we want to leverage many different methods, including the ones I just showed, but in particular the molecular ones that allow you to do this at - remember at the high density and the high energy efficiency – and energy efficiency, this is not about saving petroleum or trees, this is about the amount of energy that you can allow to transact in a brain, a factor of 1 million is a big deal because the brain only has 20 W of energy dissipation, so 1 million times that would not be a good thing.
Here's an example where one can look at the in-situ sequencing, we'd normally take it out of the context and sequence it, but here you can do it in-situ and so you can learn about the RNAs, other molecular identifiers in great detail, and we have on the order of - we're not limited to the number of colors as we reuse these colors in this new method called fluorescent in-situ sequencing. So, you get effectively 4 to the 60th power of colors. And this is really no different from next gen sequencing, except that it's in-situ.
And here are some examples of some human cells, where we look at all the RNAs and we're seeing all four colors, all at a particular cycle and if you look at a particular place in a particular messenger RNA in each of the spindle shaped cells, you can see one color. The next next-generation sequencing devices may well be these molecular-electronics hybrids with the so-called nanopore sequencing where we have DNA polymerase, it helps couple to ion channels and you basically convert DNA information into ionic information into computer information.
So, that's going from DNA to ions, but for the brain we're interested in the opposite, going from ions to DNA, and we now have molecular nano devices that can do this, they can go from light to DNA and they can go from ATP to other small molecules of DNA. So ions to DNA – and these are some of the graduate students and postdocs involved in this project – ions to DNA, we have the ion channels that are going through the spiking or in some cases non-spiking behavior that Ted and Ed had talked about, but here it's influencing the fidelity of the DNA polymerase, so that the time series of the calcium, say, turns into a spatial series of changes in the DNA sequence which you can then later read out, or compute on.
This is again a million times denser and a million times higher energy efficiency so it fits in the brain. Here's some data taken from a much bulkier paper showing the effects of calcium in the micromolar range which is physiological. And we're extending this to new polymerases and we're, as part of the BRAIN initiative thinking about the Rosetta brain. We're not picking one particular winning technology, we're trying to stimulate new innovative technologies and to integrate them all into one brain and do that over and over again in different brains, in different pathological settings in humans and other organisms.
And here's a laundry list of the types of bytes that we need to store this information. If you're interested in brain uploading and downloading, these are the kind of absolute minimal numbers that you need to be considering, but it's not out of range, it's not necessarily far off in the future.
So, we're done. Just in summary, we already augment ourselves extensively, get used to it. Extrapolation can be off by five decades or more, either in a positive or the negative direction. Molecular technology is probably where we're going with electronics and it's already here, again it's six logs better and it's improving faster in certain cases. Remember the word CRISPR even though I haven't properly explained it. And “innovative neurotechnologies” is part of the acronym BRAIN that Obama has gotten us all thinking about, and it's about a plurality of technologies, including but not limited to the polymerase technologies and the fluorescent in-situ sequencing I just talked about.
