Sustainable Chemistry through Syntethic Biology
Dr. Chance Elliott, Amyris

Talking Biotech Podcast
Hosted by Dr. Kevin Folta
===

Kevin Folta: [00:00:00] Hi everybody. And welcome to this week's talking biotech podcast by Colabra. We are literally surrounded by useful chemistry and from food to cosmetics to plastics, you name it. It's important that we have a command on how different molecules can either be isolated or assembled into different structures.

And that's been a huge accomplishment over the last century. So many consumer products contain compounds that may be a little bit rare or possibly come from a non sustainable source. Others are assembled from chemical backbones that also have a non sustainable origin and can even have ecological impacts.

So what if the most coveted chemicals could be created using synthetic biology? And this would be where you would... Take a microbe like yeast and turn that into a factory that could produce important chemicals, maybe from renewable substrates. What kind of work is being done in that space? And what's that business [00:01:00] environment look like?

That's the topic of today's interview. I'm speaking with Dr. Chance Elliott. He's the VP of R and D at Amyris. Welcome to the podcast, Dr. Elliott.

Chance Elliott:  Hi, Kevin. Thanks a lot for having me. Yeah, thank 

Kevin Folta: you very much for having a support. It's kind of a new twist on the series where we're really focusing a little bit more on how the different firms and biotechnology different companies face some of the modern challenges and quickly adapting to a new environment and new technology.

And so really what we have to start out with is what is Amherst do I took a look at the website, it looks like a broad approach to a series of questions using biotechnology, a lot of consumer facing brands. So what exactly do you do. 

Chance Elliott: Yeah, so, um, you know, essentially we reengineer the central metabolism of microbes to produce high performing ingredients, um, from carbon sources like sugar.

Um, our, our, our sort of end goal is to disrupt conventional production [00:02:00] systems. That rely on destructive and non sustainable practices and replace them with highly scalable. Being manufacturing powered biology powered by biology. Um, you know, I think 1 of the things that that sets us apart in the synthetic biology space is that we really take this process all the way from design ideation through scale up.

Process development, manufacturing, and even formulation. So our ingredients. Go into a lot of consumer brands through business to business collaborations, but we also formulate our ingredients into our own brands as well. Bioscience, pipette, pure cane, rose ink, 4U by TIA, those are just a few of our own brands that I can name off the top of my head.

Kevin Folta: And all of the ingredients, or is it most of the ingredients, or just key ingredients are coming through synthetic biology using microbial systems to produce them. 

Chance Elliott: These are like the key ingredients, the differentiating ingredients for a lot of these cosmetic [00:03:00] brands. Squalene is the big ingredient that goes in there.

Yeah, so are we 

Kevin Folta: talking mostly about bacterial systems or bacterial and fungal 

Chance Elliott: systems? Uh, we engineer in a variety of hosts, but our, our main, uh, fermentation production host is yeast. That's what we're really good at engineering. 

Kevin Folta: Yeah, cool. So, so you've been with the company for 13 years and over that time, how has the synthetic biology landscape change, particularly, particularly with respect to engineering and computing side of research and development?

Chance Elliott: Yeah, great question. So, um, yeah. As you said, uh, I, I joined 13 years ago and I think, you know, computing 13 years ago looked a lot different than. It does today, um, and, uh, I think actually we did a great job early on, um, adopting kind of the latest and greatest technologies. Um, but we were also really kind of focused on [00:04:00] a singular.

Uh, target that was financing at the time. Um, and so all the systems that we built, we're really, we're really dedicated to addressing that 1. Single problem and everyone in R and D was really focused on that 1 single problem as well. Um, we had a relatively small team. Um, so, you know, we had a lot of what I would call, uh, hero efforts where you might have, like, a single developer who works really, really hard and and really get something across the finish line, which is great.

But at the same time, they're the only people who know anything about what they built. Um, and then, you know, if they go on vacation, they leave the company, then then you're kind of out of luck. Um, and that sort of resulted in, you know, a fair number of, uh, siloed systems that we've had to deal with. Over time, so, you know, fast forward 13 years, you know, this entire cloud computing effort is is totally mainstream.

And that's something that we've had to [00:05:00] really kind of work hard to re, architect our systems to take advantage of. It's something that we kind of missed the boat on. Originally, um, and once you sunk that amount of, uh, effort into building up your tooling, it's, it's pretty hard to, to change course mid slot, mid stride.

Now, when you go 

Kevin Folta: back 13 years, you mentioned the main metabolite you're interested in was pharmacy. Where would that be found in consumer brands? Or where is that of utility to downstream processes? 

Chance Elliott: Uh, well, originally farnesine was, um, it was a diesel fuel. Um, so it was, uh, you know, a really high performing diesel fuel.

When I first joined Amaris, you know, peak oil was a thing. And, um, our goal was to, was to give the world a sustainable source of, of, uh, renewable fuels. Um. You know, that we've, we've obviously pivoted from there, but farnesine has turned out to be kind of an amazing molecule that you can do a lot of other [00:06:00] things with.

It goes into, um, you know, uh, plastics, rubbers, um, you can dimerize it, you can, you can do a lot of different types of, uh, easy chemistries with it to create, um, new materials. Um, squalene is one of those, is one of those, uh, ingredients. 

Kevin Folta: Yeah, it's all of this makes a lot of sense that you're able to start out with a few core concepts and then expand based upon this backbone.

And as engineering and computing have changed, that's all been a real positive. But what are some of the major barriers that ended up in place just from the company standpoint that forced you to create new changes to keep pace with this rapidly evolving discipline? Yeah, 

Chance Elliott: so, um, I think 1 of the, you know, 1 of the main challenges that we faced over time was just this, you know, the pivot that the company made originally from the single molecule, um, approach to what we do now, which is, you know, many, many [00:07:00] different molecules in development.

At any given time, um. And and so from an R and D perspective back then, you know, we had the all of R and D was focused on 1 thing. Um, and, uh, you know, that was an easy thing to wrap your head around is easy. You know, the, the, the amount of projects that were going on. We're not, we're not crazy. The number of assays that were being run, you know, we're manageable.

This was a manageable problem when you have a team working on 1 single thing. Um, but when you expand that, and now are working on lots and lots of different molecules and different host organisms, all of a sudden, the complexity that's running through the R and D pipelines becomes really, really difficult to manage.

Um, and while we had done a great job, I think, you know, we were probably even ahead of our time in terms of the. The digital infrastructure that we built for R and D originally. Um, that those systems didn't port well to dealing [00:08:00] with lots of different molecules. They weren't really designed. Um, with that in mind, so so we were kind of stuck in this position of having to, you know, both retool our R and D systems.

Um, so our, our pipelines and that's where we kind of got away from big teams who do 1 thing, um, in conjunction with a, um, a molecule project. And pivoted more to this sort of pipeline approach where we have. You know, basically capabilities within R and D that researchers can tap into. So, if you want strains designed, you know, our biologists don't go in the lab and, you know.

Uh, you cook up some DNA and transform it into an organism. They submit an order into our automated strain design system. Um, and they get, they get samples out of that system. Um, we have a core screening platform that scientists submit their samples into and they get their screening data back. We've got a.

A fermentation team who runs all of our [00:09:00] fermentations, you know, process development scale up. All of these are capabilities. That are kind of standalone units that we can use that a molecule program can use. Um, essentially to tie in all their work together and get to the, uh, end goal that they needed. Um, at the same time, you know, we're kind of rebuilding the data infrastructure, um, in flight.

And that was, uh, it's definitely been a real challenge. Over the years and of course, you know, cloud computing took off and, and, uh, you know, we've had to, you know, make a lot of changes in order to, uh, take advantage of that. 

Kevin Folta: Well, in that same timeframe, since you've been with the company, we've also seen the emergence of gene editing tools.

So whether it's CRISPR Cas9 or whatever the other gene editing tools might be, and how much has adaptation to new biological tools really driven company's objectives? 

Chance Elliott: Um, I don't know that has changed the objectives, but we definitely spend a lot of time [00:10:00] adapting these new tools into the pipeline. So, so really, it just makes our pipelines faster and more flexible.

So, um, you know, crisper as a, as a, you know, as a new nuclear nucleus. Um, obviously has been, uh, super valuable. There's there are others. Um, that are more kind of open source that we use as well. And we've got, um, you know, we had designer nucleases, um, from the beginning as well. But, you know, 1, I think kind of 1 of the core aspects of organizing our R and D and the way that we have.

Is that, you know, these teams are are tooled in such a way that they can take advantage of new technologies and then kind of expose that functionality out to, um. Our end users, I kind of think of it a little bit in terms of, like, uh, you know, service oriented architecture and and, um. Software where, you know, as long as our as long as our interfaces don't change, we can change a lot of what's going on inside and really give the [00:11:00] customer, you know, a much better experience without them having to be too aware of what's happening under the hood.

Yeah, so 

Kevin Folta: when you say your customer, are you talking about the end user who buys 1 of your products over the counter or through a website? Or are you talking about other, uh, companies or formulators along that pipeline that are using your products? Or is 

Chance Elliott: it both? Customer for me is our internal customers.

This would be our scientists. That's okay. Okay, 

Kevin Folta: I'm thinking of your product lines and all that stuff, too. Right? Yeah. But so what about collaboration in this environment? Do you have work with other companies to kind of leverage their synthetic 

Chance Elliott: biology strengths? Um, we have not in terms of, like, working with with other synthetic biology companies, we have done some work in the context of, um, government grants, like, uh, like, DARPA grants that we've we've worked with others in the space.

Collaborations [00:12:00] for us are mostly going to be, you know, partners. And the ingredient space where they want a molecule developed. Um, we develop those molecules, and then maybe they have either a formulation or. Um, you know, kind of, uh, a product. A pathway that that we wouldn't necessarily have. So, you know, they get they get the molecule that we need or that they need.

Um, we retain the rights to to some of the back end there, and then we manufacture that molecule for them as well. But in terms of direct. Uh, collaborations within synthetic biology. We haven't done a ton of that. I think, you know, I think this is where there's a notion and, uh, software called domain driven design.

And I think this is 1 of 1 of the things that I've definitely learned over the years is. While it really from the outset, it might look like 2 companies are doing the same thing [00:13:00] or, um, you know, or are really well aligned. And there could be a lot of synergies there. Once you kind of. Really dig into things, mostly everybody's optimizing for something different and it's that optimization.

That becomes really, really critical and that's the difference between, you know, you being able to work well with a partner and not work very well with a partner. 

Kevin Folta: Well, one of the areas that I believe is within your expertise in particular is ways to improve data collection, computational power, organizational digitation.

Can you give me some hints as to how these kinds of things play into the R& D 

Chance Elliott: side? So, I mean, I think probably the first obvious example there is, you know, is our cloud migration. That's one of the first things that we took on. And, you know, we really want to be able to take advantage of these modern tools and architectures, especially for our software systems.

So, you know, the [00:14:00] modernization of our data center was the sort of the 1st step in that, you know, modern digital evolution. Um, that's an effort. That's that's ongoing. But but, you know, really just getting our systems onto the cloud in the 1st place is a was a huge deal. And it really allowed us to focus on.

More strategic deliverables, like what's the best way to organize our data? What's the best way to access our data? What's the best way to expose data to our data scientists and to our end users being able to focus on that rather than, you know, what we were doing before, honestly, is like load balancing servers and making sure that the applications are up and running and and all of the, you know, grunt work that's required to keep a data center, uh, functioning.

It might sound like kind of a small thing, but it's a huge. That's a huge shift to be able to kind of take a step away from, you know, the tactical keep things running mode and to be able to think more strategically. So, you know, that was sort of [00:15:00] the 1st step. And then from there, you know, you know, I talked about managing complexity.

We, we really just took a look across our pipelines. Um, and across our molecule programs and said, you know, uh, you know, what does it look like? Can we track a sample from design ideation, um, through to manufacturing? For example, what gaps are there, um, in, in that process? Um, what metadata do we need to have to do a given analysis?

Is that metadata available? If it's not available, Um, why isn't it available? Um, and then we really began working on addressing those gaps. Um, and I think 1 of the interesting things there is that, you know, people tend to think, you know, you just need more tools, you know, you just need more software and you can fix all these problems.

But, but, you know, 1 of the things that we learned. Um, was that a lot of these issues actually had to do with our pipelines and our workflows and how they worked. Um, and a lot of [00:16:00] the work that we needed to do was not just building tools. But was thinking about how we operated our pipelines, thinking about how our users.

Were accessing and using our pipelines and thinking about how we needed to do things differently really from, like, the lab side of things. In order to be in a position to take advantage of the tools that we could that we could bring to bear. So, you know, getting people to standardize their work and really just getting people to understand the importance of being.

A good digital citizen, you know, I know that you're aware of kind of fair data practices. We have, you know, kind of a term that we've used around amorous is, you know, this notion of single use data where where people are doing an experiment or something and really just trying to get that. You know, answer 1 simple question, um, and how that is just not the way that we want to be able to do it.

We want all the data that we have that we produce to be able to use for lots of other different experiments and lots of other different analysis, not just try to [00:17:00] solve. Um, 1 problem at 

Kevin Folta: a time. Yeah, that makes a lot of sense. What else is really surprisingly worked well, or maybe hasn't worked to your surprise.

Chance Elliott: Um, yeah, so I can tell you, um, you know, 1 of the things I would say that hasn't worked, I think, you know, there's kind of an instinct amongst technologists that, you know, if you build it, they will come. Um, and I think this is sort of like a. It's like a Steve Jobsian mentality that people are going to be clamoring to access your great idea if someone will just let you go out and build it.

And the problem is that I think that, you know, and generally we're not selling people, you know, an iPhone and we're probably not nearly as charismatic as someone like Steve Jobs. Um, and so in the end, if you really want. To do something new, and if you really want to bring others along in that journey, and you want to make impactful change, you're going to have to study up on change management.

And that process [00:18:00] requires. A great deal of work, so when I think of examples where we've sort of failed to do what we set out to do, I can't really think of examples where. We failed from a technical aspect, but I can think of examples where we failed as a result of for change management. Um, and I think the big takeaway from my experiences that, you know, you really have to have your target audience.

Engaged early and often in this process. Um, and you might even consider having them drive the process so that they really feel like they have ownership over it. Um, I can give you a couple of examples if that's. If that's helpful. Sure. Sure. So, um, you know, 1, uh, kind of early in my tenure, leading the computing efforts, we, we developed this new version of of an automated strain design tool.

Um, it was sort of the 2nd generation of something we call the automated [00:19:00] scientist. And just as a side note here, if you really want to sow some distrust amongst amongst your scientist, client based name, your tool, the automated scientist, they will certainly appreciate that. That's a small lesson learned there.

Um, but essentially, you know, we did this internally. We looked at, we looked at, um, we looked at the tool. Um, we, we thought about the things that we wanted to fix, um, in that tool. We went to work, um, and fixed it and, you know, proudly presented it to our strain engineers and they absolutely, um, ripped it to shreds and it never gained any traction.

Um, and it was also really demoralizing for the people who. Thank you. Who worked on it, so, you know, fast forward a bit and kind of in the, in the context of plugging the gaps and, and, you know, sort of our data chain of custody that I was talking about earlier. Um, the same project came up again. Um, and so we really needed, we knew that we needed to get scientists to use this [00:20:00] tool, so that we had, you know, a clean data stream starting from, uh, design ideation.

But this time, you know, we approached it a little bit differently. We approached the scientists directly and asked them actually to lead the project, um, and this allowed them to, to shape the tool that the, you know, the way that they wanted to, um, it allowed them to prioritize science. Some of their own pain points, um, as we were building it out.

Um, and as an, as an added benefit, they really started to understand the importance of creating this clean data stream and they themselves then became advocates. For that process on our behalf, so it changed from us building something and then going out there and saying, hey, you need to use this thing because your data is a mess to them really owning the tool and having the scientists really evangelize to their peers why it was important and what the benefits of using that tool would be.

Kevin Folta: Oh, very good. So we're speaking with Chance Elliott. [00:21:00] He's the vice president for R& D at Amris. This is Collabra's Talking Biotech podcast, and we'll be back in just a moment. And now we're back on Collabra's Talking Biotech podcast. We're speaking with Chance Elliott. He's the vice president for research and development at Amris and.

We're talking about the innovation space around consumer facing products using synthetic biology and change the way companies have had to deal with current problems and applying new technologies to make their workflows work better. Maybe the best way to start on the back side of the break is, can you talk about some of the consumer facing products and maybe some of the ingredients that have gone into them?

From the synthetic 

Chance Elliott: biology standpoint, we're really focused right now on, on clean beauty and health and wellness. And so we have another number of brands. And this is an example of where we, we've taken ingredients that, you know, came from either difficult to sort that either [00:22:00] they were difficult to source.

Um, so they didn't, so those, uh, sources didn't scale well, or they came from non sustainable, um, and, and, or, you know, dirty, uh, uh, synthesis processes. And now we've given, um, these, these consumers and the ability to, to incorporate these ingredients. Um, in a way that just, just wasn't possible, uh, without synthetic biology.

Um, I think, uh, squalene is, is sort of the hero ingredient across our, our clean beauty space. And that's something that is, you know, it's something that your body produces naturally. You produce less of it as you get older. Um, and this is something, you know, where you, there's not really a, another good natural source for it.

There's, you can get it from olive oils. Um, but the quality is pretty low and so this is an ingredient that, you know, really disappeared for a long time because the original source was shark livers. Right? Um, [00:23:00] and they had been killing, you know, millions of sharks in order to get squalene and thankfully that practice disappeared.

But as that practice disappeared, this ingredient disappeared as well. Um, and this is an example where we were able to Take that ingredient, we're able to put that engineering into yeast, produce that ingredient from in a sustainable manner. And now you've not only got an ingredient that comes from a sustainable source, but it's also a natural ingredient, something that your body produces already.

Kevin Folta: Yeah, that's a really great example. So when you said, you know, clean and dirty. So you mean that there, these are somewhat either derived from maybe sourcing some. You know, rare plant that's in the Amazon and instead of grabbing all of it and taking that away from the environment, you're able to produce these rare ingredients in vitro or in a big fermenter and then be able to still have them for cosmetic use without endangering 

Chance Elliott: the environment.

Exactly. 

Kevin Folta: All right, cool. That's, that's really cool. So what does [00:24:00] the work look like today in this area. 

Chance Elliott: Um, yeah, so I mean, I think that, um, you know where we are at. Today, uh, I would say that things run pretty smoothly. Um, it's, it's not perfect. We're always iterating. We're always looking for, you know, new ways to accelerate our pipeline.

And especially from the engineering standpoint, we're always looking for better ways to, um, to build. Software better ways to integrate data science and automation, but I think, you know, we're in a, we're in a good place. We're in kind of continuous operation mode. And we're in a, in a, in a place where we can think a little bit more, um, strategically about, um, our technology portfolio.

Um, I won't talk as much about the customer, the consumer portfolio, because that's not really my focus, but from a From the technology portfolio standpoint, you know, we're in a position where [00:25:00] we can really start to think deeply about how we want to apply automation and computing. For example, these are very, very expensive resources to use.

Um, and we want to make sure that we use them in a way that maximizes their impact. So, um, we're in a position now where we work really, really closely with the rest of leadership and R and D really to define, you know, what is it that R and D is optimizing for? Um, and then let's create a clear strategy that's aligned to that optimization.

Um, and then it's about, you know, your quarterly road mapping processes. Um, defining where the project what the projects are going to be allocating resources, um, and then executing against that plan and managing the inevitable fact that. You know, reality has the inconvenient tendency to deviate from our plan.

So it's the, you know. It's the day to day operations, which I think is, you know, some people might think is. [00:26:00] Not super sexy, but I think it's all this, you know, to me, it's the difference between, you know, having a pipeline that functions and having a prototype, 

Kevin Folta: you know, you're talking about automation and computing.

But when I think about synthetic biology in in the. Parlance of maybe creating a compound that is required for some of these products. I think about it that well, we already know a lot of these metabolic pathways and we understand how this stuff is made. So why do you need to use high computational power and automation if we kind of know how to make these things already?

Is it just finding better ways to do it? 

Chance Elliott: Well, I think there's a, you know, there's a few things there. First of all, I think that, um, the assumption that we understand how biology works is, is perhaps a little bit misleading. Um, I think we, you know, we can look across, um, you know, the portfolio of genes and different organisms [00:27:00] and say, okay, we understand that this gene, metabolite

or, or intermediate that we're looking for. That we're interested in, um, it's a not at all a, um, straightforward step to go from that to let's put this gene in a new organism, um, and create, you know, a molecule, um, that this organism does not produce normally that process is, um, incredibly iterative. And requires a ton of basically experimentation, you know, we've got to run massively parallelized experiments across a huge number of designs to find 1 that produces it, you know, at the cost targets that we need to produce it.

Right? It's not just about being able to produce a molecule. That's sort of the, you know, you can produce a molecule. Great proof of concept. That's not what we're interested in. We're interested in creating an ingredient that we can scale and manufacture. And that's a much [00:28:00] different optimization. That means we've got to get to a point where the, uh, you know, we're producing that ingredient at a, at a cost that's viable.

Um, and we have to worry about, you know, all of these things like, uh, impurities and downstream separation and, and, and all of those things that are, um, the difference between, you know, again. Saying I created something in the lab and having an ingredient in a product that you use every day. Oh, very good.

Oh, can 

Kevin Folta: you describe how computing engineering or digitization? Can't say it can you describe how computing engineering or digitization improvements have really contributed to these processes? 

Chance Elliott: Yeah, I mean, I can give you a couple of examples. Um. You know, I talked earlier about, you know, how are in the early days.

Yeah, We had sort of these siloed systems that didn't [00:29:00] talk to one another. And I think one of the, one of the side effects. Of systems like that, and the side effects that the, of the fact that we were a single molecule company, um, was that we kind of missed out on some of the things that, you know, become really, really important when you're doing these huge, you know, sort of parallelized experiments.

And 1 of those things, you know, almost embarrassingly for me. Was that while we, um, you know, while we stored the sequence for design, so, you know, the, uh, a strain design, you know, at the end of the day is, uh, you know, base pair resolution sequence that gets stored. Um, we didn't have a way to annotate those designs at the time, and it wasn't important when you were working on 1 molecule.

Um, but it becomes really important when you're working on lots of them. So, for example, if you just wanted to say. You know, look across all the strains they've built or even 1 strain and say, hey. Um, you know, what genes did I put into this strain? You [00:30:00] know, there was no way to actually do that from a, uh, from a computational manner.

That was, you would have to look at the sequence and you would have to then kind of calculate what was put into there. Um, so that's not something that at the time was, um, feasible and that then gets in the way of us being able to use, you know, advanced machine learning, for example, to to do genotype to phenotype prediction.

So that was a. Um, 1, 1 area that was, um, kind of a, a, a big impact to us, um, and a big barrier to, uh, for for us to kind of move forward. And it's also a little bit of a cautionary tale because. You know, we didn't do that originally. So now, you know, we have basically this back catalog of genes that are stored, you know, just as sequences or strains that are stored as sequences.

Um, but now we need to know what's in them. Um, and had we from the outset said, okay, you know, when you, when you [00:31:00] design a strain, you know, you're going to use a standard set of components. Um, you're going to annotate those components as they go in the strain. So we know what gene you're using. We know what promoter you're using.

We know what, uh. Terminator we're using, we know what the open reading frame is. It's very straightforward and easy to do that from first principles if you're, if you're starting out. But now that problem becomes, I've got to go look back at these at this legacy data and I've got to calculate all of that information and that's a really difficult challenge.

And the reason that it's a difficult challenge is that You know, if you talk to 10 biologists, you know, and show them a sequence and say, Hey, where's the, you know, where's the promoter start and end, you're going to probably get, you know, 20 different answers to that question. So it becomes a really difficult problem to solve from, from a computing algorithmic standpoint.

And that's sort of, you know, because of the decisions that we'd made early on. Now, we're in the position of [00:32:00] really having to solve this problem and really do it the hard way. And so that's a little bit of a cautionary tale, but that's something that we were able to. Thank you. Overcome and now that's something, you know, we can look across all of our, all of our designs and say, hey, what, you know, what genes are, do we commonly use?

What, what are they, how often do we overexpress them? What do we knock out? Um, what promoters are we using? What terminators are we using? Those sorts of just, uh, pretty basic questions that become really important when you're working on, um, a large number of, uh, of molecules. 

Kevin Folta: Yeah, maybe along a related question is what is possible today?

That really wasn't possible in the very recent future. I'm very, very, I'm sure. Let me say it again. What is possible today? That wasn't possible even just a few years ago. 

Chance Elliott: Right? So, um, I mean, I think that's, that's 1 really important example. Um, and, and just that little, that little, uh, that little barrier to being able to [00:33:00] kind of interrogate a strain programmatically and say, Hey, what's In this, that unlocks all of these, um, other benefits.

For example, now, you know, we can, we can look across, um, you know, a set of bioanalytics experiments and say, Hey, you know, what genes are causal in here? Um, you know, what genes are predictive of performance? At different scales. These were things that were, um, were totally impossible for us to do before. Um, we've also kind of leveraged the use of, uh, ELNs as a, as a way of.

Of aggregating experimental data. So we have all of our limbs that that runs, um, our core pipelines. Um, but then we've kind of layered on an on top of that, which. Which allows a molecule program to aggregate, you know, all of their different experiments, um, through the lifetime of a molecule program. So you're able to kind of act ask, ask these deeper, um, genetic questions and [00:34:00] and and get, um.

You know, useful insights out of that. You're also able to, you know, kind of just look across the portfolio of experiments and understand what worked, what didn't work, what hypotheses were tried, why they were tried and what the outcomes were and not just within a molecule program, but you can do that across molecule programs and and look for, you know, opportunities to generalize.

These molecule programs, um, you know, kind of at a at a higher level. 

Kevin Folta: What might you recommend to other scientific teams that are pursuing greater digitization and data augmentation? 

Chance Elliott: Ah, great question. Um, so 1st off, um. You know, I've, I've worked in a lot of scientific organizations over my career. Um, and I think the, you know, the 1st kind of piece of advice that I have is.

Um, that, you know, you really got to accept this problem for what it is, and it's an engineering problem. [00:35:00] And to solve an engineering problem, you're going to need some engineers. Um, and I think 1 of the mistakes that I've seen made kind of consistently over the years is that, you know, a small, you know, startup stage company, they're, they're really good on the, on the science side of things.

And maybe later on in the game, you know, a couple of years down the road, maybe they hire 1 or 2 software engineers. Um, and and they call it done and and and they get stuck pretty early on. Um, and I'm not suggesting that you need to have. You know, a team of 70 engineers, but I think you do need to think hard about staffing certain core engineering roles.

Uh, very early on roles like systems engineering, data engineering, DevOps, and in addition to your software and and data science roles, even if those aren't like full. They need to be roles that you are aware of that that that people are actively supporting and thinking about so that you've really got the solid [00:36:00] foundation to build on because if you don't get this right at the beginning, you know, you're going to pay for it.

100 times, you know, it's going to be 100 times more expensive. Later on down the road, um, also, I think, you know, there's kind of a common misconception that to solve digitization and automation problems. You just need to buy a bunch of tools. Um, and if you get nothing else out of this today, I would say, you know, just hear this.

You cannot solve process problems with software or automation. And in my experience, 90% of the time, the problem that needs to be solved is a process problem, not just a tooling problem. And if you throw tools. At a bad process, you're just going to take a small mess and turn it into a really big mess. Um, along those same lines, I'm, I'm personally kind of a, a big fan of industrial engineering.

I geek out on that a lot. Um, I think we don't [00:37:00] use industrial engineering and life sciences as as much as we should. And I think if everyone had. An operations research team, and some industrial engineers, I think our collective rate of progress, um, would be a lot faster. Um, and I think the, the last thing would be, you know, I'm personally a big fan of empowerment.

Um. You know, computing resources are, are really expensive. Software engineers are really expensive. They're hard to find. Um, I think scientists can build a lot of the tools themselves if they have sort of the right. Support ecosystems, so I think are a great place to start there. Um, but there are lots of other low code tools that you can leverage as well.

Um, and of course, you know, with with chat and of entry there.

Is is lowered to quite a bit. Um, and and all of that said, I think that the, the way to make that work. [00:38:00] Um, is you've got to have, um, governance. Um, so there's got to be some engineers on the team who are responsible for governments, how these tools are used, um, creating some basic rules and some basic oversight.

And also be there to mentor, um, the, the scientists that are building tools and ensure that they're being created. And supported, um, in a way that they can be useful in the future, because, you know, the, it's so common that, you know, some person in the lab somewhere, you know, code something up and then that ends up getting dropped on the computing team, you know, later on when that person leaves and they look at it and go, what is this?

There's like, there's nothing I can do with this. So. So I think there's a, an opportunity to engage scientists on that front. Um, and the right way to do it is, is through empowerment, but also through mentorship and governance. No, 

Kevin Folta: very good along that line. If somebody listening was, say, in college now or grad [00:39:00] school, and they wanted to really find a place in synthetic biology on the cutting edge of company like Amherst, what kind of classes and what kind of training 

Chance Elliott: is best?

Great question. I mean, I think it, I think it depends on what it is that you're, you're looking to do, even on the engineering side of things. I think it's really important to be conversant and. Um, in molecular biology, this is, you know, this is what we do for a living. And if you don't understand kind of the basics of molecular biology, um, it's going to be really hard for you to be productive.

I think, um, 1 of the. One of the aspects that's important in the way that we've kind of, uh, uh, organized our engineering team is that the engineers work just very, very closely with the scientists. And that means they've got to understand the science to a, to a certain degree. Um, they don't have to be experts in it, but they've, they've got to, they got to know it.

They can't be, you know, [00:40:00] completely lost the whole time. So I think, um, I would encourage. Anyone with kind of the engineering bent to to really focus on, um, you know, getting those core science classes under their belt. Um, and, you know, it's kind of funny that you asked me that now, because I think, uh, I had a similar question with a, with a panel of, uh.

Grad students, a few years back, and I was like, you know, 1 of the things I think we really neglect is, is writing. Um, and you should really think about getting good at writing because, you know, that that, you know, communication medium is so important and then chat GPT happened. And I kind of went, huh? Well, I don't know how.

I don't know how important, um, that stuff is now, but I, I think, you know, that is just to say, like, technology advances so quickly and it's so hard to understand where, um, where these sorts of tools are going to be impactful. So, I think it's, it's just about, um, [00:41:00] whatever you can do to keep, to be flexible and be nimble, um, and, and be curious and continue to just look, uh, for new problems and, and look to be, uh, you know.

Engaged with new, you know, new areas of study. I think it's in the future is going to be all about really flexibility. Like, we need people who are going to be really flexible because we have no idea what technology is going to look like in 10 years. Very true. 

Kevin Folta: If people wanted to learn more about amorous, where would they 

Chance Elliott: look online?

Uh, amorous. com, I think we have a, we have a ton of information there. We've got, um, um, we've got, uh. Some, uh, sorry, Lauren, what did we, we have kind of a tech page that I can't think of the, yeah, it's not common. It's just easy. That's usually what I tell everyone. So when they're listening, it's just the easiest thing for them rather than embers dot com [00:42:00] slash science or tech or anything.

Okay. We have a, we have basically a tech blog on there as well. That's, that's, um, got some pretty interesting material. Tell you what, could you try, could we 

Kevin Folta: start with that one again and just say, uh, go to amaris. com and spell it and then say, we have a tech blog that talks about a lot of the, you know, just go with that.

Okay. 

Chance Elliott: Okay. Uh, yeah. So, you know, just go to amaris. com, A M Y R I S. com. You know, we have a tech blog there. It's got a lot of really interesting content. Um, tells a lot about like what we're doing and the, and the technologies that we're doing and the impact that those technologies are having. So I would definitely start there.

Very good. So, 

Kevin Folta: Chance Elliott, thank you very much for joining me today on the Talking Biotech podcast. And let me know next time if there's a big breakthrough you'd love to talk about, let's talk 

Chance Elliott: about it here. Awesome. Thanks, Kevin. 

Kevin Folta: And is there any other 

Chance Elliott: question I should have asked you? Uh, no, I mean, I think, you know, I think we covered 

Kevin Folta: it pretty good.

All right. Let me put a lid on this here. Stay [00:43:00] with me for a minute.

And for everyone else, thank you for listening to the Talking Biotech podcast by Calabra. Think about innovation and ways in which technology is moving. Some of the new ways in which synthetic biology can make our products better and make them better on the environment. So this is a Talking Biotech podcast and we'll talk to you again next week.

And that's it. So let me hit the stop recording button here, wherever my mouse went. Yeah, that's good. It worked out. Well, I'll have to do a little edit, but that's not too bad. Um, let me stop recording. Remember that is I only use zoom every day in class. Um, why am I not? Oh, there it is. All right. Good.