Meta-Flux: The AI Biologist Supercharging Drug Development
An interview with Lee Sherlock and Brendan Martin of Meta-Flux, by Joe Gorman NDRC Programme Manager.
Lee Sherlock (left) and Brendan Martin (right), founders of Meta-Flux.
Finding a co-founder is usually a painful process. You have to go to networking events to sell your idea, prostrate yourself on LinkedIn, and attend a whole plethora of online and offline meetups which might yield no return, no interest, and no dream partner to build your company with.
For the lucky few though, serendipity intervenes. Your ideal co-founder might just be the sound guy at work.
“Myself and Brendan were hired on the same day for the same job, in the Science Gallery. They interviewed and brought us on as a pair, ” says Lee Sherlock, the CEO and co-founder of Meta-Flux. “Soon after that, we realised we’d get along. I thought right, I'm going to skive off most of my shifts and spend time just talking to this dude about whatever subject crosses my mind.”
Brendan Martin is the technical architect behind Meta-Flux, a startup designed to rapidly accelerate the progress of drug development using AI. Sherlock’s favourite aspect of the job at the Science Gallery was a physics-based exhibition all about terraforming Mars. Martin preferred “the mad variety of different shows, from physics to biology to more tech-oriented stuff.” The two undergrads were in good company: “Every day you would learn something new. The whole team there was so smart, with most of them going into PhDs. A learning environment like that is really fun, but it also showed me and Lee how we would work well together. I’m tech-oriented, Lee was very science-oriented, so we bounced off each other and both learned something.”
I’m beginning to think of the kinds of conversations I have with my colleagues at lunch (How much would you need to be paid to swim in the Liffey? What’s the biggest animal you think you could take in a fight?). I make a mental note to bring up terraforming Mars the next time I’m at the water cooler, then ask how two young Science Gallery employees decided to delve into such a complex niche.
Sherlock, the Trinity biochemistry graduate, had the initial idea: “Part of my undergrad involved writing on the concept of personalised medicine, and I was trying to incorporate some forms like AI into it to figure out if all of the excess of personalised medicine could be trimmed, honed, and improved. I was doing a lot of work on protein sequencing in college then going into work in the Science Gallery every weekend and annoying Brendan, asking him how AI worked. He was good enough to spend time explaining it and breaking it down, and in the back of my mind I think Meta-Flux was forming.”
Through his previous work in laboratory settings, Sherlock had an inkling that there was the potential for AI to vastly improve the rate of drug development. He had seen the inefficiencies which meant it could be “up to six months” before he had a workable data set to analyse with his Principle Investigator. A good PI would usually have questions that would require reanalysing and restructuring the data, always examining one variable at a time. The “secret sauce” of Meta-Flux is that it combines linear and nonlinear data, allowing researchers to skip several steps and have a multidimensional approach to data analysis, examining multiple variables at once.
This methodology took a while to take shape. On their lunch break, walking laps of St. Stephen’s Green, the pair bounced ideas around, trying to figure out how they could work together. Then, circumstance intervened once again.
Meta-Flux founders Lee & Brendan with industry veteran Brian Caulfield during NDRC Accelerator mentoring session.
“We got hit with the perfect storm,” says Sherlock. “We got made redundant from the Gallery because of COVID. We were getting paid by the government to basically stay at home and do nothing. Everyone was ringing their mates and just trying to stay connected with each other.”
“Me and Brendan were on FaceTime one of the days and I said “Remember that idea that we had ages ago? I'm gonna teach myself how to code, and build it.” And with that, Brendan just got pissed off, and just said: “Move over you.”
Martin laughs but recalls it slightly differently: “I distinctly remember within two or three days of Lee trying to learn how to code, he said: I need your help, or I'm done with this. That was pre-ChatGPT so nowadays it might be a different story…”.
The nucleus of Meta-Flux may have started with Sherlock’s experiences in lectures and labs, but the motivation to build the company comes from deeper within. Sherlock lost his grandfather to lung cancer a few years ago. Soon after, Brendan Martin’s father passed away after a battle with the same disease. Adversity did not stall their progress - it spurred them on.
The two founders cobbled together an MVP that they admit was pretty rough around the edges in terms of look and feel. Still, they managed to test it on 120 lung cancer patients through a double blind study and returned with near-perfect predictions of patients’ age, sex, lung cancer diagnosis and overall prognosis. The project started to gather steam when the two founders had a conversation with Sherlock’s mentor from Trinity, Dr. Ken Hun Mok, who looked at their results and assured them that they had something of real value.
“He’s been my crutch for navigating the science side pretty much the whole way through,” confesses Martin. “He’s one of the top ten experts on metabolomics globally but will sit down with me and point out why some things make sense and some don’t. Getting the data to the right point is an art form.”
Over time, these sessions with Dr. Mok crystallised into a quantitative computational process. I make the mistake of asking exactly how this works, despite my lack of experience with quantitative computational processes. Seeing the quizzical, vacant look of the village fool after a minute or so, Sherlock starts again:
“Modern science today is quite one-dimensional. It's linear. When you're kind of working on these data sets, you identify the protein and the potential mechanism that your drug is acting on. But that’s missing the point of what’s happening biologically. Drugs that you’re testing have a three-dimensional relationship in the molecular landscape of proteomics and metabolomics. The end user needs to understand the complexity of those relationships to understand where to go next. We are developing Meta-Flux to get you closer to causation - to make the leap to the answer of “what do we do next?” a lot shorter.”
Martin is keen to point out that a lot of the processes in data analysis and research in this realm are antiquated. “It’s a lot of manual data manipulation: swapping around columns and cells on Excel. It’s laborious and time-consuming and AI is incredibly good at reproducing that work at huge speed. The infrastructure of AI has exploded, so all that needs to be fed in is good data. So the data was one of the big blockers for this industry, but we’ve developed tools that are able to take multiple data sets, funnel them in, and allow for that exponential scaling.”
Lee & Brendan presenting Meta-Flux on NDRC Accelerator Kick-off.
“We want to be at the R&D point of drug development, delivering detailed answers that will give deep insights into future patient outcomes,” says Sherlock. “Meta-Flux will help you test multiple hypotheses for multiple applications of new drugs, simultaneously. If we can deliver answers at the intervention point, we can either give a new medication the best chance of success or kill it fast. We want to help the people developing these drugs to fail faster, and fail earlier.”
“To get a therapy from discovery to market is a ten to fifteen year process, if you’re really good at it. You’re looking at two to three billion dollars in that timeframe, and half of that is at the R&D stage. There are about 21,000 drugs in R&D at any given point, and most of those are wasted as just 0.001% get from the discovery stage to the open market. We’re helping people get to the answers faster - is this the right application of the drug, or not? Saving time and money as a result of that is a byproduct of what we do.”
Brendan Martin jumps in: “We did a demo of the product with the Chief Scientific Officer of an early-stage R&D company and he said “Jesus, this is so powerful, it’s exactly what we need.” Seeing that excitement is nice. We’re saving them a lot of the legwork at the early stage, giving them space to think and freeing them of some of the financial constraints by testing multiple ideas simultaneously.”
Martin and Sherlock are generous with their time in breaking down the intricacies of how new medicines hit the market. You learn a lot in a short space of time around them. Still, there’s one question left before I close the door on our conversation. Where does ‘Meta-Flux’ come from?
“It stems from good old metabolism fluctuations” says Sherlock, smiling. “I didn’t want to waste too much time on the name - I’d rather just start building it and get going. We’re hoping to get the best drugs to market to make a major impact on people’s lives. The best rock bands in the world have the shittiest names. So we’re not focused on the name. We’re focused on the music.”
NDRC Accelerator is now open with rolling applications and monthly intakes. Apply Now