Founder Stories: Sajith Wickramasekara - Co-Founder and CEO of Benchling

Cool. Good evening, everyone. My name is Saji, and I work at BenchLink. We make tools for life science research, so think pharmaceutical and biotech companies. Scientists use our tools to build drugs faster. We're a bit of an oddball at this event actually though, we're not exactly a data company. We are a company that helps transform and frame the complex problems of modern science as solvable data problems for scientists. Compared to a lot of software out there, what we're doing isn't very sexy honestly. There's going to be no AI for diagnosing and curing diseases in my presentation, at least not yet, fingers crossed. Instead, I want to show you how you can leverage actually quite standard data practices, stuff that's gonna seem table stakes to many of the folks here, to make a big impact in an industry that's outside of sort of the traditional lenses of technology. So when you think of drug development today, you probably think of really impressive science that modulates the body in all kinds of impressive ways. Avastin is a good example of this. It's a drug that came out in 2004. It's what's called a chemotherapy add-on. It's an antibody therapy, and what that means is for an antibody therapy, an antibody is injected directly into your bloodstream, and it's kind of like a homing missile for a really specific biological molecule. In the case of Avastin, the molecule is called blood vessel growth factor. Blood vessel growth factor is exactly what it sounds like. It is something that promotes the growth of new blood vessels. When a cancer patient has tumors, tumors depend on blood vessel growth to get bigger and to spread through the body. And so you now have these homing missiles that destroy the ability to create new blood vessels, and it makes chemotherapy more effective and take less time because life is just that much harder for the tumors. When it came out, Avastin was a big step forward in chemotherapy. It's approved for multiple cancers and still used today, and it's improved the survival rates for some of them by quite a lot. You would think that such deliberate advanced science would have a large amount of software and advanced data behind it as well, but that's actually not the case. Just to step back for a second, science is generating more data than ever, but we're not doing anything to really support it. We're not doing what we need to be doing. So my background's in computer science, but I spent many years as a bench researcher in parallel. My plan was to go get a PhD and go work in drug discovery as a scientist sort of doing this. But I kind of got frustrated having one foot in the world of software and one foot in the world of science. In software, it's, an absolute pleasure to build a product with a team of people who could be all around the world because you have the tools to collaborate around code and data very effectively. In science, you have really amazing scientific tools. DNA sequencing has gotten incredibly cheap and efficient. We have things like CRISPR now for really precise gene editing. But the way people work together on research hasn't fundamentally changed since I started and probably long before that. This is an industry that runs on paper, email, and Excel, and it's something that touches everyone. So as I said before, we're not exactly a data company. We're a software company that intensely leverages very standard data practices in some cases to make a huge impact on a field that touches everyone. So today I'm gonna talk about how that applies to the challenges of making an antibody, much like Avastin, and sort of how we approach that problem. So making an antibody is fundamentally an iterative, high-volume process. It generates lots of information and it's kind of a data problem. Creating something like a vaccine actually involves starting with tens of thousands of candidate antibodies in the beginning. After a few rounds of screening, scientists eliminate most of them and they're down to a few dozen candidates. From there, they have to optimize, characterize, and eliminate more until they have just a handful where they gather enough information to make decisions on which ones are promising enough to take to the clinic. This is a really complex process. It takes coordination between a lot of different teams. There's a lot of inputs and outputs. This is where modern data and software engineering can really help. Scientists today have to solve data problems like these, but they don't have the tools they need to structure their work in such a way that it's actually a solvable data problem. So as a quick interlude into, into what we do, we're a single application for the entire research workflow.. And what that looks like is a lot of normally disparate tools all integrated together very tightly. This sort of platform-based approach to aggregating and managing data isn't anything novel, but giving scientists full access to their data and their colleagues' data allows them to approach data problems in a completely different way, and it's just underserved in the life sciences. So with that context, let's talk concretely about a case with one of these companies doing antibody discovery that was able to really change the way they did research using technology. This is a company that is working on about 10 different drugs in parallel. They've got multiple teams at multiple sites all around the world all working together as well. So at a very high level, their original data platform consisted of Microsoft Word for documenting experiments and high-level reports, Excel and FileMaker Pro for tracking the actual antibodies themselves and the data associated with them. Each of these research sites all around the world has slightly different ways of tracking and documenting the experiments and their data as well. They're doing cutting-edge science, but there's something really wrong with this picture. They're literally trying to cure cancer using Excel. This is the equivalent of one of the companies you work at building software without version control. It's honestly probably a little worse, actually. So the The resulting workflow is beyond complex and convoluted. It involves a lot of backtracking and doubt. Beyond making life difficult for the individual scientists who have to spend, I don't know how many years, training just to get to that point to do this research, it prevents the company from having a bird's eye view of the problems that they're actually trying to solve. The data is hidden by the system that it's being stored in., and that in turn makes it very difficult to frame any of these problems as solvable data problems. So how do we help? Step 1, the primary way is by enabling collaboration around data. That's kind of part 1. Before we stepped in, scientists would have to spend time, often hours, sifting through records trying to find individual candidate antibodies and the data related to them. If they needed a specific sequence and that was owned by a colleague who might be somewhere across the world, they'd have to email them and wait a couple days in some cases to get the data back so they could continue their research. Took a pretty obvious modern approach to this which was to combine one, the tools for actually doing the biological design, the tools for creating the antibodies with the tools for documenting them. So it's kind of like CAD for biology plus digital lab notebook all linked together. When a scientist is documenting and designing at the same time, the system is able to create references between the two. So that if anyone ever looks at this again, both pieces of data are surfaced. Remotely accessing your colleagues' data using the internet is something that's 10 years ago for everyone here in this room, but we're still operating in a world where the folks building our cutting-edge treatments for all these diseases often rely on email and flash drives. Our customers need to communicate the results of tens of thousands of antibody experiments. Across sites. So when we were first in contact with them, they did this using a slew of Excel spreadsheets. They'd have to update these, email them back and forth, sometimes have a central kind of shared drive where these were kept, sometimes not. At one point, they even had spreadsheets to track the spreadsheets. This became troublesome because for any individual experiment, you're generating hundreds of antibody candidates at once and data associated with them. And the information gets scattered across these spreadsheets, and you never really know which ones are the most up-to-date. So what we did was we built what in, in science words is called a biological registration system, which is a fancy word for database. But it's a database with biologically aware schemas that the customer can configure. It's capable of handling numerous data types from antibodies to DNA sequences, proteins, and even biological samples. And it's the kind of system needed to handle the complex cataloging to discover, characterize, and optimize these antibodies. So by combining this awareness of the biological data with the searchability of a database, it's much easier for scientists to compare results among thousands of antibodies. Instead of trying to figure out where the most up-to-date copy of, you know, a spreadsheet is or requesting data from a lab in Europe, the scientist just writes a query. To power this bioregistry, we use really boring old technology. Postgres, 20 years old, RDS, also old. But what we've brought that's new to this is a fresh perspective on how to organize, model, and index the data on top of it such that we can make a big impact in the life sciences. So for most antibody research, the discovery process itself is a work in process. There's no formula for spitting out a drug by going through some set of steps that's going to work. You have to iterate on the process itself while you're doing the experiments, so it's kind of meta. Because we've now got this place where all the data and documentation for these experiments is living, we can abstract one layer and actually get data on the process itself so we can help scientists improve how they do their work. They currently, customers before BenchSync would keep track of their keep track of this through spreadsheets, they would hire project managers, ad hoc task tracking systems, you name it. This is a huge challenge for, for resource allocation. Teams never know when requisitions are going to be coming in. You have to hire dedicated people to manage projects and to understand where they're tracking and blocking. And when you need to surface something that happened 6 months ago and all the associated context, it's nearly impossible. With the workflow integrated into the rest of the system, scientists can do their work and figure out how to do their work better. Science can sometimes seem like an art, and it is to some degree, but a lot of it can be broken down into a process. Life science is very complex, but it— and there's a lot of moving parts, but fundamentally you've got steps, you've got teams, you've got inputs and outputs that map to data that falls into a specific part of the system. Instead of scientists needing to know where a certain file is or worrying about who's doing what, all of that should be taken care of automatically. By the system so that you can analyze it later to make improvements. Tracking the completion of these experimental steps is really where our tool helps with antibody discovery being a work in progress. Since the science is so intensive and since there are tens of thousands of candidates for any given study that need to be assessed and refined and characterized, it's necessary that scientists iterate on the process. That's really hard to do when your infrastructure is Word and Excel Excel and email and so forth. You don't have an integrated view over all of your work. The files and data aren't linked together. And so without that information, process optimization is just really challenging. So standardizing through centralization and optimization is the theme here. There's an aging but persistent trend in software to automate these processes, whether it's monitoring, deployments, recovery, etc. You gain productivity when you encode these processes in software. You have better assurance of consistency, the data can be automatically tracked, and you get deeper insights into your operations. That's all been set up software operations, and we're now bringing it to, to the life sciences. So just to kind of synthesize what I talked about, we see drug discovery moving from a paper, email, Excel world to kind of a single end-to-end application, a system that enables scientists to capture more data than ever before and enable, you know, applications that previously weren't possible. Having companies being able to understand a bird's-eye view of their research so they can actually optimize the process, giving scientists the ability to search their data and their colleagues' data for the first time. This kind of optimization is so critical to science, and a lack of it really slows it down. So if there's one thing I want you to take away from this talk, it's that using quite standard data practices, we were able to make a big dent in a really high-impact field that touches everyone. And it doesn't require a ton of domain expertise. Most of our team actually has no background in the life sciences at all. There are definitely a few of us that have been there, but many of the folks are just interested in making a difference with their tools. So if you're interested in a mission like that, and obviously a huge shout out to our team and investors who helped us build such a great product, we're hiring.

Thank you. Thanks very much. So one or two questions from me and then I'll open up to people. Okay, that was quick. What's your— so having been in the trenches of selling this product to the life sciences industry, what's, what's the level of readiness of people?

Are people evolving towards, you know, this type of products? Yes and no. When we first started, there was a strong aversion to any cloud-based software. The ideas of collaboration weren't really resonating, but in the last few years since, there's definitely been an uptake. I think it's normal to expect the life science industry, and other industries honestly, to lag kind of one or two cycles behind. So I think about tech, like things like Slack and whatnot are only coming to maturity now, so like you kind of got to, you got to look back a few years when you set your expectations.

And the early adopters are younger folks or smaller companies or?

Started with the younger folks, so started with folks in grad school who are gonna adopt a tool out of their own volition just because it makes their lives better. And you go to the small biotechs, medium biotechs, and eventually the large companies, the pharmaceutical giants start noticing. Great.

We have one question, but I don't think Andrew is here, so we only have one mic I think for now.

All right. Yeah, thank you. Um, um, so since you're dealing with, uh, quite sensitive data in terms of business aspects and that, um, I have two questions. Um, first of all, the observation that, um, this data is shared by email— can you tell me whether this is encrypted in any way, or it's just going clear text out, or what that is? And the second part of that, um, what do you think, um, or what do you have as experience, um, what people actually see as security problems having this not on-site but somewhere else in the cloud, which is someone else's computer, which you probably see as a problem.

[Speaker] Yeah, so that's a good question. There's a lot of, I don't want to say all of it's theater, but there's a lot of paranoia because it is really important IP. The fact is that a lot of these companies, they do have systems that exist, but they're quite quite clunky and not integrated, so the scientists resort to using email, and this is whatever email systems they have set up, sometimes their personal Gmails as well. So it's kind of like a shadow IT situation where inadvertently scientists are being forced to use these other technologies, and then it's based on whatever email they're using, and so it's only as strong as the weakest link in that case. So I think a lot of concerns are actually better focused on how do you prevent that than how do you make sure that your system The system is ultra secure. I think your second question was just about, do you mind repeating?

[Speaker] Yeah, how do you sell the idea of having someone in the cloud, which as an IT department in such a big company, I would be very nervous about.

[Speaker] Yeah, when we first started, everyone refused. Now people are starting to, I think, just become more open-minded about it because of technology that ends up drifting into their organization anyway.

Is there any evidence that they use into actually validating the security of that?

Do they try that? [Speaker] So we have to, from our large customers, we have to undergo security audits on-site. We have pretty rigorous policies around how we develop our code as well, just to abide by regulations. I guess the one other piece I was going to say is that the pharma industry has typically been wildly profitable, and it still is, but it's under cost pressure for the first time. Just because of patents expiring and that sort of thing. And therefore, they're, as a result of economics, becoming more amenable to cloud software as well.

Great, thank you. It's actually an interesting theme across all these events. It's the fact that data is finally truly moving to the cloud, which even last year felt like people would never do. And it markedly over the last 12, 18 months, you start hearing folks being much more comfortable.

Question over there. Hi. So the Vention platform seems pretty interesting in that, you know, obviously it's a pretty efficient aggregator of data, you know, be able to put the data in a single view of it all for the first time outside of these kind of homegrown and Excel-based tools. But is it simply just a repository, or does it actually afford users and customers, you know, business analytical tools, scientific analytical tools, to analyze that data as well, or do users actually have to then build that custom on top of the platform and/or attach other tooling to it to actually make it valuable for them?

That's a great question. Uh, and something that's really important to be kind of useful beyond just a one giant silo for all the data. So the way we handle is that we have a lot of those tools built in, but then we can, from these sort of like registration capabilities I was talking about earlier, where the customers are able to, in a very simple way, define their own schemas for the things they care about. They can plug that into any of their analytical tools. You basically run it through some kind of ETL process and they can plug it into Jump, Spotfire, Tableau, and do kind of more downstream analysis from there. That's the most common mode of operation, I guess, we see for people to really get into the data from there.

We have time for one or two more. I thought there was another one there somewhere. Otherwise, there's one here.

I made the—

okay, all right. Yeah, so I guess I have another kind of boundary question. Whenever you're doing your own software, question is where do you end and something else begins? And I was wondering about the email component. So obviously the scientists all around the world are having to communicate with each other as well. Is that done through an external email client or is it wrapped up into the software? And so for instance, if I need a report from somebody, does that generate an email request or is it all self-contained?

Sorry, I might have miscommunicated a little bit earlier. When I was using email, I was trying to refer to sort of the state of the world where people are resorting to emailing files back and forth, and email is the system of record rather than a tool like BenchLink that's designed for it. I think we obviously do generate email notifications and so forth, but the data does primarily live in BenchLink. Sorry if I miscommunicated that earlier.

Does that answer the question? Don't mind me, I'm just taking a pic.

Last question. Sounds like you have a pretty established product that has— that you have created from your own experience and those on your product team. How have you found the product evolution to differ from where you started originally from your own experience and then finding that it might not fit where some other labs are going? And are you, do you see yourself like pivoting in certain situations and where are the trends?

Yeah, we started by scratching an itch that was a personal frustration. So we focused on a core use case that we knew very well. And we gave a basic version of our tool out for free all throughout grad schools everywhere and we've got a lot of scientists using it to refine those core use cases. I think what we didn't anticipate is when we started working with larger companies, big biotechs, pharmaceuticals, there's a lot of changes that come with scale and complexity. It's not necessarily changes to the science, but it's things around security, around compliance and federal regulations, around coordination between multiple teams, whereas kind of an academic lab you only ever have like 5 to 10 people, but sometimes you have 500 people all working on the same project, all with different sets of responsibilities, and it just adds— it does add more complexity to the product. And so we've had to spend a lot of time with our customers listening, obviously, in order to be able to deal with those challenges. But it's more of a layer on top of the product, and the core is still pretty much the same. Um, just primarily, like, we found a good compass to be, can we build a tool that scientists will love to spend their entire day job in, and therefore all the data will get accumulated there, and we can enable the customer to have really good downstream access to that.

Is it mainly for antibody research, or do you find that the projects differ?

We have folks doing everything from engineering custom microbes to spit out chemicals, to engineering immune cells to fight cancer, developing antibodies, biofuels. If it touches biology, it's probably a good fit.

Very cool, thank you so much.