Synthetic biology startup Lygos raises $13M to make cleaner chemicals(techcrunch.com)
techcrunch.com
Synthetic biology startup Lygos raises $13M to make cleaner chemicals
https://techcrunch.com/2016/12/13/synthetic-biology-startup-lygos-raises-13-million-from-os-fund-and-ia-ventures-to-make-cleaner-chemicals/?hn
5 comments
This is why I walked away from my field of chemistry...
We're learning more and more that microbes are essentially radically optimized factories. They've evolved for millions of years to be really good at producing stuff. Typically, though, it's not stuff we want. But if we edit the genetic code of the microbes we can direct that productive capacity towards making things we do want.
That's exactly what Lygos is doing for industrial chemicals. And the beauty of the microbes Lygos is engineering -- specifically yeast -- is that they feed on sugar & water. Feed them sugar & water, and out pops specialty chemicals. This allows Lygos's microbial factories to produce industrial chemicals with sustainable inputs and with no toxic outputs.
I'm incredibly excited by the potential for synthetic biology to reinvent the way we produce all sorts of things we like to eat, wear, or make things with in the decade ahead.
That's exactly what Lygos is doing for industrial chemicals. And the beauty of the microbes Lygos is engineering -- specifically yeast -- is that they feed on sugar & water. Feed them sugar & water, and out pops specialty chemicals. This allows Lygos's microbial factories to produce industrial chemicals with sustainable inputs and with no toxic outputs.
I'm incredibly excited by the potential for synthetic biology to reinvent the way we produce all sorts of things we like to eat, wear, or make things with in the decade ahead.
> ...optimized factories...
> Feed them sugar & water, and out pops specialty chemicals.
Even if your culture contains microbes that produce the specialty chemical you're interested in, natural selection in the reactor favors the populations that use the least sugar and water and multiply the most. In other words, reactors are very good at growing microbes: surprise! They're quite bad at selecting for much else.
Just doing what you described is called the Evolvenator. That discovery described how to build and test the microbe, but not necessarily how to design it.
There's a fix: what if we make the interesting chemical a side effect of the kind of more efficient metabolism that's expressed by the highest-surviving population in your culture? In other words, the microbes that multiply the most make the interesting stuff by accident, usually as a consequence of the media in their reactor and a small tweak to their genetic code.
It turns out that designing a chemical pathway that is metabolically superior to a natural one and as a side effect, produces an economically valuable product, is crazy hard. It is usually harder than designing an industrial process to manufacture the economically valuable product in the first place.
That's why the article mentions that these other biotech firms "specialize" in making certain kinds of chemicals. In reality, they experiment with maybe two targets a year, and they land one that works out of 10.
You'd be wrong to judge them on their hit rate: rather, I'm just pointing out how there isn't some kind of "optimization" magic to discover efficient chemical pathways. It's incredibly hard research! And I'm always surprised (though pretty delighted) that VC finds its way to these people, because VC avoids actual science projects like the plague.
PS: The $13 million will buy things like the Sartorius bio reactor, a $1 million up front + $1-2 million a year machine capable of running automated experiments on yeast cultures in bioreactors. If you're in the business of making pickaxes instead of trying to strike gold, clearly you should make a better bioreactor. If you agree, just e-mail me.
> Feed them sugar & water, and out pops specialty chemicals.
Even if your culture contains microbes that produce the specialty chemical you're interested in, natural selection in the reactor favors the populations that use the least sugar and water and multiply the most. In other words, reactors are very good at growing microbes: surprise! They're quite bad at selecting for much else.
Just doing what you described is called the Evolvenator. That discovery described how to build and test the microbe, but not necessarily how to design it.
There's a fix: what if we make the interesting chemical a side effect of the kind of more efficient metabolism that's expressed by the highest-surviving population in your culture? In other words, the microbes that multiply the most make the interesting stuff by accident, usually as a consequence of the media in their reactor and a small tweak to their genetic code.
It turns out that designing a chemical pathway that is metabolically superior to a natural one and as a side effect, produces an economically valuable product, is crazy hard. It is usually harder than designing an industrial process to manufacture the economically valuable product in the first place.
That's why the article mentions that these other biotech firms "specialize" in making certain kinds of chemicals. In reality, they experiment with maybe two targets a year, and they land one that works out of 10.
You'd be wrong to judge them on their hit rate: rather, I'm just pointing out how there isn't some kind of "optimization" magic to discover efficient chemical pathways. It's incredibly hard research! And I'm always surprised (though pretty delighted) that VC finds its way to these people, because VC avoids actual science projects like the plague.
PS: The $13 million will buy things like the Sartorius bio reactor, a $1 million up front + $1-2 million a year machine capable of running automated experiments on yeast cultures in bioreactors. If you're in the business of making pickaxes instead of trying to strike gold, clearly you should make a better bioreactor. If you agree, just e-mail me.
FWIW, in a past life, I bought (4) 2L Sartorius reactors used for $75k, built out my lab for <$50k and had an operating budget of $450k w/ 2 fully-burdened RAs.
I agree with much else of what you've said. I worked in biosynthetics with some of the Lygos teammates at another company. All-in-all it's not a bad idea, but the development time is _so_ long, especially in yeasts. It's analgous to writing assembly w/o documentation, and debugging takes months per code-change.
I agree with much else of what you've said. I worked in biosynthetics with some of the Lygos teammates at another company. All-in-all it's not a bad idea, but the development time is _so_ long, especially in yeasts. It's analgous to writing assembly w/o documentation, and debugging takes months per code-change.
What if you added a lot of "error-checking" code that kills the host when it doesn't produce enough of the valuable chemical? I'd guess this is very difficult in practice.
Or also add a competing organism that "triggers" when the concentration of byproduct is too low?
Or also add a competing organism that "triggers" when the concentration of byproduct is too low?
For the first idea, you'd have to have some lethal protein or something that's specifically inhibited by the chemical of interest. Coming up with that might be even harder than the original problem. And even if you could, evolution wouldn't favor cells that produced it, since it confers no advantage. There's an energetic cost to making that protein.
For the second idea, again, evolution ruins everything. The competing organism would do just that - compete.
For the second idea, again, evolution ruins everything. The competing organism would do just that - compete.
Because "you get what you select for". Instead of getting more of the chemical, you'll get a host that has hacked and disabled your error checking code.
Why can't you just design organism / insert DNA to make X.
Grow organism in 1,000 little reactors, measure output of X, keep top y% and repeat?
Yield may drop over time, but this way evolution is somewhat on your side.
PS: I don't expect that to work, but I am not sure what the issues are.
Grow organism in 1,000 little reactors, measure output of X, keep top y% and repeat?
Yield may drop over time, but this way evolution is somewhat on your side.
PS: I don't expect that to work, but I am not sure what the issues are.
Let's say you had such an organism.
Naturally, all the yeast individuals in your population have slight genetic variations. Maybe millions of them. Most have your plasmid to make X.
Due to natural selection, the competing priority—let's call it the "Y" product—is reproducing * surviving as much as possible. The yeast genes that make the yeast reproduce as much as possible and survive as well as possible will eventually be the largest population in your reactor.
So if X doesn't also achieve Y, the only thing that will be left in your reactor is a population that achieves Y.
The worst thing (which is true most of the time) is that achieving X is opposed to Y. Processing some chemical in the media that doesn't translate to food for the organism is wasted metabolic resources. So for most applications you can think of, your X population will die or reproduce away its X genes.
There's nothing you can parameterize to fix this! Like Y will happen regardless of what temperatures you use, or what media you use, or how much oxygen you give the reactor. By virtue of growing something, you're optimizing for Y. Either design X to also achieve Y, or fail. Evolution is opposed to you, not on your side!
In practice today, we don't understand the biological systems design well enough to make X also achieve Y. So what happens is you grow the yeast until you have a population with maximum metabolic efficiency (i.e., maximally achieves Y), and THEN modify the yeast to do X. It dies quickly, and your left with a little bit of product. This works, but it's not viable.
Naturally, all the yeast individuals in your population have slight genetic variations. Maybe millions of them. Most have your plasmid to make X.
Due to natural selection, the competing priority—let's call it the "Y" product—is reproducing * surviving as much as possible. The yeast genes that make the yeast reproduce as much as possible and survive as well as possible will eventually be the largest population in your reactor.
So if X doesn't also achieve Y, the only thing that will be left in your reactor is a population that achieves Y.
The worst thing (which is true most of the time) is that achieving X is opposed to Y. Processing some chemical in the media that doesn't translate to food for the organism is wasted metabolic resources. So for most applications you can think of, your X population will die or reproduce away its X genes.
There's nothing you can parameterize to fix this! Like Y will happen regardless of what temperatures you use, or what media you use, or how much oxygen you give the reactor. By virtue of growing something, you're optimizing for Y. Either design X to also achieve Y, or fail. Evolution is opposed to you, not on your side!
In practice today, we don't understand the biological systems design well enough to make X also achieve Y. So what happens is you grow the yeast until you have a population with maximum metabolic efficiency (i.e., maximally achieves Y), and THEN modify the yeast to do X. It dies quickly, and your left with a little bit of product. This works, but it's not viable.
So from my understanding,
bacteria is competing to reproduce the fastest, and the chemical that we want is simply a byproduct of its metabolism or pathway for reproduction
issue is, however, the byproduct has to either be the most efficient factor for reproduction, or else by natural selection it will be removed
my question is, can we use the same techniques found in tree grafting to prevent or delay natural selection, and grow our byproduct?
bacteria is competing to reproduce the fastest, and the chemical that we want is simply a byproduct of its metabolism or pathway for reproduction
issue is, however, the byproduct has to either be the most efficient factor for reproduction, or else by natural selection it will be removed
my question is, can we use the same techniques found in tree grafting to prevent or delay natural selection, and grow our byproduct?
question - so why does agriculture work? why are we able to grow tomatoes year after year.
is the entropy at the cellular level versus at a plant level so different that the processes are "locked" in place.
does it make more sense to try and grow plants that produce malic-acid tomatoes? sure it might be harder to geow one... but once grown, you can scale it infinitely.
is the entropy at the cellular level versus at a plant level so different that the processes are "locked" in place.
does it make more sense to try and grow plants that produce malic-acid tomatoes? sure it might be harder to geow one... but once grown, you can scale it infinitely.
I suspect that it's that you can manually select which tomatoes you want. It's a bit harder to select individual microbes.
So essentially, the difference between the trait you want, X, and the trait that survives, Y, doesn't exist. Every Y is an X, since generations of farmers are the ones manually doing the selection.
So essentially, the difference between the trait you want, X, and the trait that survives, Y, doesn't exist. Every Y is an X, since generations of farmers are the ones manually doing the selection.
that makes a lot of sense. so in case of tomatoes, the survivability of the plant has no correlation with actual survival. it is being engineered by the farmers.
what we really need is a way to select and ensure survival of the yeast that we want.
what we really need is a way to select and ensure survival of the yeast that we want.
Alot of companies do exactly that. Using systems like: http://www.tapbiosystems.com/
But it's not cost effective when producing millions of liters.
But it's not cost effective when producing millions of liters.
TAP Biosystems is the manufacturer of what is now called the Sartorius.
Also, if you look carefully at the Lygos website photos, you'll see the Sartorius.
Also, if you look carefully at the Lygos website photos, you'll see the Sartorius.
@doctorpangloss's comment is 100% right. I'd add the small problem that behavior in small reactor != behavior in large reactor. Reactor scale-up is somewhat unpredictable and often ends in tears.
Here are two great articles on the challenges of biofuels:
1. The Rise And Fall Of The Company That Was Going To Have Us All Using Biofuels: https://www.fastcompany.com/3000040/rise-and-fall-company-wa...
2. A Biofuel Dream Gone Bad: http://fortune.com/kior-vinod-khosla-clean-tech/
1. The Rise And Fall Of The Company That Was Going To Have Us All Using Biofuels: https://www.fastcompany.com/3000040/rise-and-fall-company-wa...
2. A Biofuel Dream Gone Bad: http://fortune.com/kior-vinod-khosla-clean-tech/
just another iteration of biofuel..
So, Dupont has a production-scale microbial process for 1,3-propanediol. What is the market for malonic acid if Dupont can already go directly to 1,3-propanediol?
http://www.dupont.com/products-and-services/fabrics-fibers-n...
http://www.dupont.com/products-and-services/fabrics-fibers-n...
Malonic acid and its derivatives have their own demand as esters and neat acid apart from conversion to 1,3 propanediol. It's possible to catalytically oxidize 1,3 propanediol to malonic acid, so they are somewhat competing in the same space.
The "clean" advantage of these fermentation processes is IMO greater than the "sustainable" part. You could use renewable energy to synthesize all the upstream precursors of traditionally-synthesized malonic acid (e.g. monochloroacetic acid, hydrogen cyanide) from air and seawater and it would be perfectly sustainable. But the intermediate products are hazardous to handle and there's a large waste stream of aqueous salt solutions and organic solvents. Eliminating hazardous intermediates and producing waste streams that can be handled by simple wastewater treatment makes it cheaper to build and operate facilities for these chemicals.
The "clean" advantage of these fermentation processes is IMO greater than the "sustainable" part. You could use renewable energy to synthesize all the upstream precursors of traditionally-synthesized malonic acid (e.g. monochloroacetic acid, hydrogen cyanide) from air and seawater and it would be perfectly sustainable. But the intermediate products are hazardous to handle and there's a large waste stream of aqueous salt solutions and organic solvents. Eliminating hazardous intermediates and producing waste streams that can be handled by simple wastewater treatment makes it cheaper to build and operate facilities for these chemicals.
Agreed. Just pointing out that they're going up against an established DuPont process.
Also that process has been around for at least 10 years:
https://www.ncbi.nlm.nih.gov/pubmed/14580573
I met the guy (Charlie Nakamura) who developed the process. It took a hell of a long time to get everything working at useful efficiency, then a fair amount of development to get it working at scale.
Also see others in this space:
Amyris (older synthetic bio company) Zymergen (newer synthetic bio company that seems to be doing some of the same things Lygos wants to do)
https://www.ncbi.nlm.nih.gov/pubmed/14580573
I met the guy (Charlie Nakamura) who developed the process. It took a hell of a long time to get everything working at useful efficiency, then a fair amount of development to get it working at scale.
Also see others in this space:
Amyris (older synthetic bio company) Zymergen (newer synthetic bio company that seems to be doing some of the same things Lygos wants to do)
Agree- our startup (www.Librede.com) does biosynthesis of cannabinoids in yeast from sugar. Much more environmentally sustainable and economical. Further, the modularity of the cannabinoid synthesis pathways in cannabis allows production of specific cannabinoids by substitution of single enzymes.
I just want to say thank you for making a startup from an idea my friends had in college. For some reason we just never managed to get our shit together to the point of doing this. I can't remember why...
Imagine warehouses filled with growing tomatoes. They wouldn't have leaves or roots. The plants would be directly injected with sugars and nutrients to grow. There wouldn't be any wasted energy due to the inefficiency of photosynthesis, or the need to grow leaves that will just be thrown away when the fruit is harvested. There'd be no pesticides, herbicides, or fertilizer running off into the water ways. No more habitat destruction. We would have delicious tomatoes in our grocery store, picked only a day before, full of nutrients and flavor. I'm aware that Monsanto is a terrible company, but GMO is going to help save the planet.
Lygos is the first step towards the above scenario happen. At its heart, a cell is basically a huge finite state machine. It is programmed both through dna and proteins. Some day (and maybe Lygos already has something like this) we'll have a programming/description language to describe what a cell does/how it works. It would start with the basics. (When you see this molecule, attach this other molecule to it. When concentration of this gets too high, start doing this.) It would also have a basic support structure (nucleus/ribosomes) needed to be a functioning call. On top of this, you could program it to produce what you needed.
I never supported Trump, but I'm severely disappointed in the direction his administration is going. If he wants to make America great again, lets start researching projects like what I described above. There was a pdf posted in yesterdays posting on Germany's fusion reactor, talking about the private companies that were helping to build it. They had to come up with new manufacturing skills and tools, and are now able to use those in other jobs. That is how you make America great again.
Lygos is the first step towards the above scenario happen. At its heart, a cell is basically a huge finite state machine. It is programmed both through dna and proteins. Some day (and maybe Lygos already has something like this) we'll have a programming/description language to describe what a cell does/how it works. It would start with the basics. (When you see this molecule, attach this other molecule to it. When concentration of this gets too high, start doing this.) It would also have a basic support structure (nucleus/ribosomes) needed to be a functioning call. On top of this, you could program it to produce what you needed.
I never supported Trump, but I'm severely disappointed in the direction his administration is going. If he wants to make America great again, lets start researching projects like what I described above. There was a pdf posted in yesterdays posting on Germany's fusion reactor, talking about the private companies that were helping to build it. They had to come up with new manufacturing skills and tools, and are now able to use those in other jobs. That is how you make America great again.
What is up with this and the other lengthy post? Sounds like some marketing spiel. The company may he doing good things but this sounds disingenuous.
Direct injection of sugar to grow plants to circumvent the inefficiency of photosynthesis?
Where do you suppose sugar comes from?
Where do you suppose sugar comes from?
>Imagine warehouses filled with growing tomatoes.
The fact is, even with a perfectly engineered tomato, it's hard to beat the low cost of dirt+sun.
The fact is, even with a perfectly engineered tomato, it's hard to beat the low cost of dirt+sun.
> There wouldn't be any wasted energy due to the inefficiency of photosynthesis...
Honestly, tomatoes are really efficient as is. They grow huge amounts of fruit per acre; they grow in preposterous conditions, like people's windowsills, greenhouses, fields across the world, etc. I don't mean to pick on your particular choice of example, but you really got to get into the weeds to really understand the efficiencies of agriculture.
Indeed, there's consensus that the greatest threat to agricultural efficiency are all the weaknesses of a highly-engineered monoculture. Imagine the devastation of the Florida orange harvest a few years ago due to frost, but due to a single weed.
> a cell is basically a huge finite state machine
There is a huge amount of research to create an "algebra" of synthetic biology, which is really fascinating stuff. You'd love it! You're right that a lot of what you're describing is inevitable, but there's a huge grind to get there, solving engineering problems totally orthogonal to the theoretical biology of it all.
Honestly, tomatoes are really efficient as is. They grow huge amounts of fruit per acre; they grow in preposterous conditions, like people's windowsills, greenhouses, fields across the world, etc. I don't mean to pick on your particular choice of example, but you really got to get into the weeds to really understand the efficiencies of agriculture.
Indeed, there's consensus that the greatest threat to agricultural efficiency are all the weaknesses of a highly-engineered monoculture. Imagine the devastation of the Florida orange harvest a few years ago due to frost, but due to a single weed.
> a cell is basically a huge finite state machine
There is a huge amount of research to create an "algebra" of synthetic biology, which is really fascinating stuff. You'd love it! You're right that a lot of what you're describing is inevitable, but there's a huge grind to get there, solving engineering problems totally orthogonal to the theoretical biology of it all.
> There is a huge amount of research to create an "algebra" of synthetic biology
do you have links to such research?
do you have links to such research?
A cell is absolutely not a finite state machine, and anyone who tells you this either has no clue about biology or has no idea what a finite state machine is. A cell's "state" is governed by a highly parallel set of probabilistic interactions, not logic gates. No two cells given the same inputs can be guaranteed to enter the same state. Populations of cells always exist in a distribution of states.
And those probabilistic interactions perform no logic at all? Obviously a real cell isn't identical to a silicon circuit, but I feel it is a useful abstraction. If you want, look at proteins as being the finite state machine. These people agree:
> We propose a theoretical formalism, molecular finite automata (MFA), to describe individual proteins as rule-based computing machines. (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3070173/)
A textbook I read several years ago on cellular mechanics inspired my thoughts on this, unfortunately I'm not able to find it right now. I was blown away at the time, the chapter on dna transcription strongly reminded me of programming.
> We propose a theoretical formalism, molecular finite automata (MFA), to describe individual proteins as rule-based computing machines. (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3070173/)
A textbook I read several years ago on cellular mechanics inspired my thoughts on this, unfortunately I'm not able to find it right now. I was blown away at the time, the chapter on dna transcription strongly reminded me of programming.
Logic is the wrong way to look at it, cells have no logic. They have an impossible array of equilibria, random interactions, chaotic mixing, and things we don't understand at all, along with others we may not even know exist.
People like to think of blobs interacting and not interacting, but really a cell is more of a crowded mess, see this image for a sense of scale:
http://media.tumblr.com/525082e14ba810bc539c61a646fdba93/tum...
Now picture everything in that image swirling around incredibly fast, with some parts of it (the yellow fibers) more stationary and everything else randomly zooming/bouncing around. There is no equivalent in computation.
A logic may appear to emerge from that jumble, but it's not going to behave very predictably. It takes years of work to be able to predict mechanistically just a tiny aspect of one system, much less the whole cell.
The DNA example you describe is a fascinating area, but it took ~30-50 years to get that understanding, and we still find out new things about it all the time! Now that's a really easy system to study, you can isolate its components, you have clean readouts (new DNA generated, sequencing etc.). How about everything upstream of that? The signaling chain going from the cell surface to the DNA to turn on a gene is quite long, picture at least 3 links with yet another 3 links each, which themselves have separate, independent links back into the DNA.
That's not to say we can't make assumptions and make some progress, see Marcus Colvert's lab:
https://covert.stanford.edu
But it's really, really hard, and an FSA is a quite incomplete way of describing it. Looking at the paper you sent, a useful MFA would have to be insanely large to be useful, and defining all the terms is just an impossibly large task.
People like to think of blobs interacting and not interacting, but really a cell is more of a crowded mess, see this image for a sense of scale:
http://media.tumblr.com/525082e14ba810bc539c61a646fdba93/tum...
Now picture everything in that image swirling around incredibly fast, with some parts of it (the yellow fibers) more stationary and everything else randomly zooming/bouncing around. There is no equivalent in computation.
A logic may appear to emerge from that jumble, but it's not going to behave very predictably. It takes years of work to be able to predict mechanistically just a tiny aspect of one system, much less the whole cell.
The DNA example you describe is a fascinating area, but it took ~30-50 years to get that understanding, and we still find out new things about it all the time! Now that's a really easy system to study, you can isolate its components, you have clean readouts (new DNA generated, sequencing etc.). How about everything upstream of that? The signaling chain going from the cell surface to the DNA to turn on a gene is quite long, picture at least 3 links with yet another 3 links each, which themselves have separate, independent links back into the DNA.
That's not to say we can't make assumptions and make some progress, see Marcus Colvert's lab:
https://covert.stanford.edu
But it's really, really hard, and an FSA is a quite incomplete way of describing it. Looking at the paper you sent, a useful MFA would have to be insanely large to be useful, and defining all the terms is just an impossibly large task.
I'd be pretty impressed if somebody could grow in vivo tomatoes efficiently. It's really hard to beat leaves and roots for nutrient and energy absorption/conversion. I've seen warehouses of plants growing, about the best people can do right now is flowing nutrient solutions directly on the roots.
Photosynthesis is actually fairly efficient.
As for the whole cells-are-state-machiines, while that's technically true, the state vector has a trillion parameters.
Photosynthesis is actually fairly efficient.
As for the whole cells-are-state-machiines, while that's technically true, the state vector has a trillion parameters.
Some other organic acids are already made by bacterial fermentation. It's standard practice for lactic acid, and an uncommon method for succinic acid.
More information: http://www.nnfcc.co.uk/publications/nnfcc-renewable-chemical... http://www.nnfcc.co.uk/publications/nnfcc-renewable-chemical...
More information: http://www.nnfcc.co.uk/publications/nnfcc-renewable-chemical... http://www.nnfcc.co.uk/publications/nnfcc-renewable-chemical...
Heck, Britain almost lost WWII because they couldn't make acetone. Chaim Weizmann invented a technique to make bacterial produce acetone.
The Acetone–butanol–ethanol process:
https://en.wikipedia.org/wiki/Acetone%E2%80%93butanol%E2%80%...
Obsoleted by cheaper petrochemical processes, but there's renewed interest now that environmental considerations are getting more attention, so maybe it will return to commercial use some day.
https://en.wikipedia.org/wiki/Acetone%E2%80%93butanol%E2%80%...
Obsoleted by cheaper petrochemical processes, but there's renewed interest now that environmental considerations are getting more attention, so maybe it will return to commercial use some day.
[0] Yes, this Kristy Hawkins: http://www.bodybuilding.com/contest_media/2542/22122/d/img_4...