Lab of the Future | Importance of End-to-End Workflows with Robert Williams, Ph.D. (Sponsored by Molecular Devices)

We continue our series on the "Lab of the Future"  by delving into the importance of end-to-end workflows with Agilent Director of Digital Lab Innovation, Informatics Division Robert Williams, Ph.D.

Williams discusses how Agilent views the lab of the future and why end-to-end workflows are so critical. He explains the benefits of these workflows and how they can streamline processes and save time.

Episode transcript available on Buzzsprout.

Key Learning Points: 

• The challenges and benefits of implementing end-to-end workflows
• How to better utilize data to improve automation processes
• How to retrofit these workflows into an already existing lab
• How these workflows can inform front-end automation. 

Our Sponsor for this Episode
Molecular Devices makes scientific breakthroughs possible for academia, biopharma and government customers. Dedicated to enabling life science labs of the future, where innovative technology and novel research meet, Molecular Devices empowers scientists to advance discovery, driving earlier diagnoses and safer therapeutics for patients. Spanning cell line development, 3D biology and drug screening, their automated, end-to-end solutions streamline and scale complex workflows while integrated machine learning-enabled analytics allow researchers to mine data easily for insights.

Molecular Devices is the innovation partner that empowers scientists with next-generation technology to advance discoveries, improving the quality of life everywhere.

Learn more about Molecular Devices by visiting:
https://www.moleculardevices.com/

Stay connected with SLAS

• Online at www.slas.org
• Facebook
• Twitter @SLAS_Org
• LinkedIn
• Instagram @slas_org
• YouTube

Upcoming SLAS Events:

SLAS Europe 2025 Conference and Exhibition

• 20-22 May 2025
• Hamburg, Germany

View the full events calendar

Transcript

Hannah Rosen: Hello everyone and welcome to New Matter, the SLAS podcast where we interview life science luminaries. I'm your host, Hannah Rosen, and today we'll be continuing our series focusing on the lab of the future with Robert Williams of Agilent Technologies. Agilent was one of our SLAS 2023 Lab of the Future companies, and Robert is joining us today to discuss the importance of end-to-end workflows in the lab of the future. Welcome to the podcast Robert! 

Robert Williams: Thank you very much, pleasure to be here. 

Hannah Rosen: Oh, it's our pleasure to have you. So, to start us off with, I would love it if you could just kind of provide us with a little bit of your professional background. 

Robert Williams: OK, I'm the Director of Digital Lab Innovation and Excellence Informatics Vision. My background is in software development, and I've been with the with Agilent for over 8 years working on various aspects of software, informatics and laboratory digitalization as either a developer or as a manager. 

Hannah Rosen: Oh wow. What got you interested in the field? 

Robert Williams: It's kind of... from the very beginning I worked at a very small business in Irvine, CA, coming straight out of graduate school, and one of the things that I was amazed at back then was just the sheer extent to which good software packages could sell instruments. So, we had a... a very large mass spectrometry system, that was... that was often custom... custom made. And a lot of times it was the software that edged us out over the competition. So that's sort of how my passion for software and software innovation came about, as just seeing how just... just the impact that good, user friendly, flexible software could have on, you know, selling instrumentation. 

Hannah Rosen: That's really interesting. And I think that's probably something that we'll come back to a little later on in the podcast like, this, you know, this seems like more and more people now are realizing just how crucial that software element really is, you know, as we get these more advanced automation hardware. So, I'd love to hear, you know, your perspective as a speaker for Agilent, you know, how is Agilent viewing this concept of the lab of the future? 

Robert Williams: Well, in... in many ways it's already here. I've attended several conferences over the last year and I've just been amazed at what people have... have been able to achieve in terms of, you know, connectivity across the lab automation. So, the lab of the future is... is here today and... and it's actually... we're... we're glad to be a big part of it as a vendor. You know, as you kind of alluded to, it's not just about instrument specifications anymore, it's not just about requirements, it's really about how well your instrumentation can connect and integrate with, not just your own equipment, with other equipment in the lab. So, it's about connectivity or automated intelligent workflows, but then also automating the data management. There's a lot of components to this, and again it... it's been interesting seeing, you know, some companies are quite sophisticated now in that in this area. So... so as a vendor, we try to partner with our customers to ensure that we're providing them with the right digital tools and capabilities so they can enable their vision of an intuitive user friendly, you know, digital lab in the future. So, it... it really is a partnership between us as a vendor and... and the, you know, companies we supply to work together to bring this about. 

Hannah Rosen: Can you describe for us what, you know, we mean when we say end-to-end workflows? What are... what does that mean and why does Agilent feel that they are so important to the lab of the future? 

Robert Williams: An end-to-end workflow, at least by my definition, is really about seamlessly integrating all of the components in the laboratory operation, you know, from experimental design all the way to data collection and through to where data is aggregated, and making that process as seamless and user-friendly as possible. So, it really is from one end to the other, we want to make sure that it's automated, it's seamless. Yeah, we've... we've taken as many of the manual operations... manual transcriptions out of it, and we're agile, it fits in all this. We need to make sure that our systems are connectable because most companies already have something in place that they're using today, and we need to be able to integrate to that. So, we really have to meet people where they are today and make sure that we can integrate into that. So then, again, by connecting all of these, you know, we can really increase the productivity of the lab because it comes down to, you know, how can I reduce errors and common errors or things like, I put the wrong sample in the wrong place, you know, I wrote down the wrong number, and that's kind of what this end-to-end automation is really all about. It's just eliminating all those possible errors and... and really automating as much of the connecting and automating through two different things, but related connecting and automating, you know, all these processes in the lab. 

Hannah Rosen: It seems to me, you know, I'm wondering what the challenges of... cause, you know, a lot of this equipment that people are buying is really expensive. And so, I can imagine that people probably may not want to, if they are interested in end-to-end workflows, connecting... you know, having to buy all equipment from one provider to make sure it's all connected if they already have equipment from separate providers. So, I was wondering if you could speak a little bit to... to how is that kind of changing the game as a provider for, you know, having to... are you working more with other providers to make sure your technologies can talk to each other? How does that work? 

Robert Williams: Well, you're absolutely correct in that it's a very heterogeneous laboratory environment. So, in cases, for example, with small laboratory devices and so forth, we may partner with or collaborate with those particular vendors to ensure that our software can communicate with theirs. In... in some instances where we're working with true competitors, a lot of times you have to provide the end user with the tools that they need to connect those things and try to make that easy as possible. So, a lot of times that's, you know, providing the appropriate software development kits or the appropriate application programming interfaces so that their developers can go and connect things up on their own. And again, a lot of these labs are different. So, it's kind of hard to write a standardized application that will fit all these environments, especially again with that homogeneity of system. So, we have to make sure that we provide the appropriate tools to help them bring all this together. 

Hannah Rosen: So, you know, when we're talking about end-to-end, I feel like what most people think of is, you know, start to finish. But what I understand is, more and more we're trying to create more of a... a closed loop system. And so, I would love it if you could talk a little bit about how can people use the end data that they're getting, maybe at the end of an experiment, to help inform the front end of their automation. 

Robert Williams: Yeah, this is a big... this has been a big push over the last few years, and frankly, this is one of the areas where you're starting to see a significant amount of your AI/ML, you know, applications is, I've collected, you know, I've started the front end, lt's say I have a... either a sample prep or a method development thing I'm working on. I have a set of front-end automation components that I want to use to... to develop that method. I... I do a... first I analyze the data with my, you know, analytical instrument setup and all of that's automatic. Samples, they have to be walked over to the instruments, but ultimately... but then again, we got people who are transferring the robot, so bringing all those samples over to the instruments. I collect that data and then I put that data in some type of data repository or data lake and then I use artificial intelligence or machine learning to tell me how to iterate what I was doing on the front end, and then go through that process again. So it becomes this sort of iterative closed loop process as you describe where you're using the data that you collected in some type of, you know, AI/ML and that's usually again, you know, depending on whether you're in pharmaceuticals or applied, you know, a lot of times, there's a lot of IP and what they develop in terms of those algorithms, we're... we're closing that. So yeah, most definitely, you're starting to see that it's not really end to end anymore where you start and finish. It's really a... a loop of... of going back and forth to optimize something. 

Hannah Rosen: And I would imagine then that, you know, because this is something that researchers have always done, you know, it's kind of part of the... the scientific method is, you know, you get to the end of your experiment and you use those results to inform, you know, if you have to repeat the experiment or build off of that experiment. And so, you know, do you look at this as, we're just kind of enhancing the researchers ability to build off the experiments, or are we kind of entering into an entirely new way that experiments will be run? 

Robert Williams: It depends on the context. Many contexts we're augmenting the scientists in the laboratory. Because again, we're removing all of the, I would say, tedium sometimes associated with a given set of experiments. So instead of having to sit there and manually pipetting, where errors could occur, instead of having, you know, instead of again writing things down, manually performing measurements, in some cases, you know, depending on the type of system somebody has, there could be a... I create an experiment in my laboratory notebook, and then I spend an hour creating sample lists or creating various things to be able to analyze that. And by automating all that, we can really keep the science focused on the innovation, and again less focused on that tedium sometimes associated with the actual conduct of the experiment. So yes, it really is, you know, an enhancement of what the scientist is trying to achieve. 

Hannah Rosen: What are some of the biggest difficulties or challenges in developing end-to-end workflows? 

Robert Williams: I think the biggest thing, at least at the outset, and you kind of alluded to this earlier, is just the sheer number of vendors, different types of equipment, different data systems, different software packages, and if you were just getting started, I mean it could be a pretty overwhelming, you know, and sometimes just identifying where to start can be difficult. You know, do I want to focus on the data management first or do I want to focus on the experimental design? And where we've had a lot of success in addressing this is, and it's going to sound simple and obvious, but it's basically really breaking things down and trying to identify areas in an existing workflow. So, if we go in and we map that whole process, pick areas where we can have the most impact on the productivity, you know, in a relatively short period of time, and then going in and focusing there, and then once we're successful there, then kind of, you know, using that as a starting point then build off of, that and slowly kind of iteratively build out, you know, a... a fully digitalized lab. It's really figuring out where to start. And a lot of times it's mapping that process out and say, you know what, we waste an awful lot of time here, so I'm going to... I'm going to focus there and start to work in that area. 

Hannah Rosen: I wonder, how often do you have it... I mean, obviously, you know, your customers are coming to you, so they're the ones who are motivated to do the change. But I wonder, in your, you know, perception, how... is it very common that, you know, people see all of this end-to-end workflows, you know, that maybe other people are doing, or they're hearing about it, and they think, I want to do this and I need to figure out where to start, versus being kind of forced into it one way or another. And I'm sure that you probably see both in the same lab sometimes. 

Robert Williams: Yeah, we... we see a lot of both. I mean, it's sometimes... it's... it starts as simple as, my company is making us take all the data and put it in the cloud, you know, or... or put it, you know, so they want me to put all my data... they want to aggregate all of my data in the cloud. And then there's a discovery of, now that I have this data, you know, that... that I've removed some of the vendor specificity, I've got it in the cloud, what can I do with it? Is there something I can learn from it? Can I mine it for things? So, you know, sometimes, yeah, it starts with a push from, you know, an IT department, you know, to kind of say it... for people sometimes to realize the value of it. But then other times it just, you know, if you look at, for example, drug discovery, it's not only for the patient, but for the business. Time to market is in... it is incredibly important. You know, the faster I can get through certain parts of the discovery and development process that, you know, the faster, you know, obviously I'll get to market and, you know, so forth. And that's where a lot of times they ask themselves, what can I automate? What, you know, steps can I facilitate in this process to compress that timeline? And... and sometimes that's where it originates. So, it... it can come from a couple of different areas and again, you know, sometimes they come to us saying, hey, you know, I've got a pretty large sample prep backlog that, you know, I need some assistance working through, and then we can provide automation tools, you know, to help them get there. 

Hannah Rosen: So, how often does it happen where... cause I could just think back to my own days in the lab where I just remember producing these massive quantities of data and just knowing, like, there must be something here, I just don't know where to start or how to approach it. It's not relevant to the topic I'm asking right now, so I'm going to save the data because who knows, but now it's just sitting on a hard drive somewhere, and nobody's probably ever gonna look at it again. And, you know, when... now that we're sort of integrating AI more into these end-to-end data processing is, are we starting to see that, you know, is I guess what I'm asking, is the AI kind of capable of telling us, hey, here's all this data sitting here that you're not using and here's something that you can do with it to help make your processes more efficient? 

Robert Williams: It... it really starts with... and it's funny because one of the... one of the fundamental things, you know, associated with data science is... and oftentimes knowing what questions you want to ask, then sometimes going back and saying OK, what data do I need to answer those questions and do I have it? But to your point, it's funny about the archive is that we've had, you know, I've had conversations with a number of customers who have this massive archive of data that they've collected over 20 years. And this could be, you know, pharmaceutical data, it could be toxicology data, it could be all sorts of stuff where they might be able to glean learning without even having to do any experiment if they could extract that data. So a lot of time what people are able to do is really focused on, OK, how do I get at that data in the first place and... and if it was taken 20 years ago, are there tools available that I can use to extract that data out? You know, get it into a form where I can use it and consume it and try to get some of the answers I'm looking for. It all comes back to that. How... how can I get the data out of these old archive files and into a place where I can actually use it? And then going forward, how do I want to collect my data so that it's ready from the get go for that purpose going forward? 

Hannah Rosen: And how often is it feasible to kind of pull that old data that was taken. 

Robert Williams: It kind of depends. I mean, there... there are some instrument vendors and so forth that, you know, the software packages, for example, you know, that were in use 50 years ago are still available today and those can be used to... to extract and pull data into a better neutral format. But then there are other companies out there. I mean, I've bumped into a couple of small companies that specialize in this sort of like, you know, going out and OK, we know you have this massive data archive, we can help you extract it into something that you can use, but nowadays it's... it's really, it's... you said it's... it's kind of the, I want to make sure I'm collecting all of the data in the right formats so I can go and do those... do those things. 

Hannah Rosen: Well, so speaking to that, you know, how difficult is it to go in and like, essentially retrofit these processes into already existing labs that have all of their, you know, old processes in place that may not be optimally set up for these sorts of end-to-end workflows? 

Robert Williams: It's... well, there is... there’s a... it's kind of funny. There are a couple of... of levels to that, you know, levels to that question. I think the first part, when you go in and you try to... to automate a laboratory that's traditionally been very, very paper based or... or... or very, very manual based is... frankly, it's cultural. It is really convincing the individuals to place their trust in, you know, a pretty, you know, high level of automation, you know, and that really doesn't happen overnight. It's a... it's an adaptive process and... and working with people and so forth. The... the other, you know, challenge in all, but is there's often a concern about the roles of individuals in the lab of the future, you know. OK, as we automate these things, what's my role going to be going forward? And I... I think it's important to bear in mind that automation systems are really focused on augmenting the scientist capabilities in the lab. So, allowing them to focus more on the science, more on the innovation, and less so about instrument maintenance, paperwork, you know, and kind of the other routine, tedious tasks. I mean, I had a... a colleague once who used to manage a... a rather large discovery and development lab, and he said the most painful thing for him to see was when one of his top-notch scientists was standing in front of a photocopier, because he was just thinking about all of that... all of that knowledge, all of that know how sitting there making copies, you know. So, and the idea is, can we... can we prevent those sorts of things, you know, through automation? And again, the other thing that I think plays a lot, you know, labs of where I think again the folks in the lab can find it liberating is by reducing errors. You can also remove a lot of rework, you know, when you have to go. And I mean, because there's nothing more tedious than just repeating the same experiment over and over again because something went wrong. So again, by reducing that, we can reduce the amount of rework. So, from a cultural perspective, it's like getting the people who are working in the lab to see the value of bringing this in and... and you know and embracing that value because it... it does, again like I said, it could be quite liberating in terms of being able to keep people focused on the science. So... but, you know, there can always be a bit of an uphill battle with this, rightly so. People might be concerned about, what is my role if we bring in all these robots and all this stuff in the lab? 

Hannah Rosen: Yeah, absolutely. What about smaller companies? You know, how feasible is it for smaller companies or smaller labs with maybe just fewer resources to implement these, sort of, end-to-end workflows? 

Robert Williams: It used to be, especially if, you know, if you look at some of the larger companies, because the vendors didn't necessarily provide connectivity related tools, they would have to hire software developers, you know, and IT people to help them connect all this stuff. And as you can imagine, that's pretty expensive. So therefore, it was just out of reach for a lot of smaller companies to, you know, justify that level of expense just if you're a startup or something, it's, you know, how can I really justify all this automation... automation related expense, when I'm, you know, trying to hit my drug targets and so forth, but now a lot of the vendors are putting tool kits out. They are putting, you know, software packages. I mean, some of the small lab device manufacturers have software apps that allow you to route data coming off of these... these systems to... to other... to other systems to try to make it easier for people to... to move that data around. And by making it easier and providing these tools, the... since... since we're talking chemistry, the activation barrier we're actually deploying, some of these things is getting lower. And... and now they're, you know... you know, the... you're starting to see, you know, packages being sold, you know, to... to do these sorts of things, if not even included in some of the... the instrument software. So, it's getting better. It used to be pretty... pretty difficult and somewhat cost prohibitive to try to build this out. And if you look at some of the, you know, more successful companies doing this early on at, to your point, they're usually quite large. And they usually have a fair number of software developers, but that's becoming less and less of a requirement now, as... as the companies put out the appropriate tools to be able to do this with less... less of that type of resource. 

Hannah Rosen: So, when we talk about, you know, people utilizing this data that they get kind of at the end of their... their experiment processes and using that to kind of inform the front-end development, you know, what are some of the ways that people are using this? Is it just figuring out ways to improve your experimental protocol, or are we also talking about ways to more efficiently use the equipment itself? 

Robert Williams: It's both. First of all, you can... there's a whole side, and... and Edgewood has a division whole devision that works on this sort of thing, is really understanding, yeah, laboratory utilization, what instruments are being, you know, what instruments are being overused as it were, being taxed, and what instruments aren’t being used, to help you optimize, you know, the laboratory, you know, operation. Also spotting trends, you know, being able to spot trends in data collection. You know, you can predict when... everything from when an instrument might go down, and then you might need to swap the column out. You know, by have all this data available to look for trends, to look for things that might be happening in the lab that, you know, you or I may not notice. So, there's a lot of, you know, the... the... there's a lot of different uses that, you know, not only just scientific, but, you know, there's a whole lot operational component, there's a whole maintenance component, so there's a lot that can be done with the scientific data that, you know, and other telemetry data from the instruments for that matter, that essentially, if you can... and again, it all comes back to... and I get it in the right form and in the right plate, then I can really start to do a lot with it. So yeah, but there's... there's a lot of different uses for the data and it's not just... not just AI/ML. Sometimes it's just simple trend analysis and so forth. You know, very simple algorithms just to determine, you know, whether things are moving in the right direction or in the wrong direction. 

Hannah Rosen: That's nice, and that's nice to like, remember, because I feel like especially, you know, now more than ever, AI and ML are getting so, so much attention and all the focus is on them, and it's important to kind of reinforce that they're great and they're important, but you can do so much with your laboratory automation without them even. 

Robert Williams: Well it... it's know... it... it really comes down to knowing the role of... of AI/ML. You know, where... where do I need it and where do I not need it. And then what type of AI or ML do I need, cause again it's... I mean, I remember going to a conference a few years ago, and AI/ML was all the rage, but people didn't exactly know what they were going to do with it yet or how they were going to use it, and there were a lot of, I would say to be nice, visionary kind of things that people thought they were going to be able to do with AI/ML, and what people are discovering is... is that they're discovering its limitation and now I would say doing a better job of figuring out when and where to apply it to get the best result, you know, so that... but... but again, there's a lot of things you can do about it, but we still have to have that data. 

Hannah Rosen: So, and you know, can you speak a little bit more about, what are the sort of ways to speak about these limitations of the AI and ML? What are you referring to in this setting and do you see that changing over the next several years? Do you think that we will advance AI and ML to the point where those limitations no longer exist? 

Robert Williams: Well, I think... I think the big... I think the biggest limitation that came in was sort of this expectation, and you alluded to this earlier, you know, are we going to be able to replace the scientist, you know, in... in the loop. And I think people are getting their head wrapped around the idea that no, you're really not going to be able to do that. You're not going to be able to start with an experiment design and now, never say never, but you're not going to be able to do that, at least not right away. That you're going to... that AI/ML, it's best suited again augment that scientists role, you know, because you have this massive data out there, right? And... and it's more than, you know, you and I could... probably even handle ourselves, and the idea is by using those tools, I can actually make use of that... of... of that data. And then again, the scientists can go back and focus on the particular innovation challenges that they're trying to solve. They see the tools getting better and our ability to manage more and more, you know, more and more data sets, but also more and more types of data where not only it's just scientific data coming off the instrument, it could be literature, it could be information that’s available on the web, and aggregating all of that in a presentable form. And that's another challenge with all this is, how do I... how do I design the interfaces and so forth such that the scientists now can go in and see this data in a way that's useful to them. So, visualization is a key, you know, component of this as well. And that I would say is something where I... I've seen a lot of innovations over the last year is how to visualize all of this data. You know, once you've had the algorithms crunch on it, how do I present so scientist can... can derive some meaning from it so. 

Hannah Rosen: Well, you know, unfortunately, we're almost to the end of our time. But before we say goodbye, I just wanted to give you an opportunity, Robert, if there is anything that we didn't get a chance to discuss here today that you think is really important for people to know about these end-to-end workflows or anything else related to the lab of the future, I'd love to give you the opportunity. 

Robert Williams: Oh, no, no. I'm... first of all, appreciate you having... having me on and giving me the opportunity to talk... uh, talk a little bit about this. I... I mean, I think it's an exciting space to be in. I remember not that long ago when everything was being written down and... and... and stuffed into CSV files and so forth, and in some places it still is. But it's really exciting to see all of this come together. And the other thing it's... it's exciting to see is, you know, there's a lot more openness and sharing, I think, of information as a... as a result. Now you see company sharing, you know, data that they believe might be helpful to the industries as a whole. So that's been another, I think, exciting thing is these technologies that, you know, become available is... is you're seeing much more sharing amongst even sometimes what would have been competitors because they realize that it... it advances the industry so. 

Hannah Rosen: Definitely exciting times. Well, Robert, thank you so much for joining me here today. It's been a really interesting discussion and we really... we really look forward to seeing Agilent at our future SLAS events. 

Robert Williams: Thank you. Sounds good. Thank you.