The Analytical Scientist Innovation Awards 2024: #5
Welcome to the 5th ranked Innovation, Pyxis – introduced here by Matterworks co-founder Jack Geremia
| 4 min read | Technology
5 – Pyxis
A simple, scalable AI-powered platform for untargeted absolute quantitation of analytes
Produced by Matterworks, Inc.
Matterworks are taking powerful AI techniques developed for language and image processing and applying them to LC-MS data – enabling the direct transformation of uninterpreted, raw LC-MS data to a biologically actionable list of the identities and concentrations of detected metabolites. Pyxis standardizes the entire process of untargeted analyte quantitation – eliminating the need for per-analyte calibration, peak selection, integration, and manual data analysis.
What the judges say…
“Smart application of AI for the analysis of LC-MS data.”
Meet the Expert
Jack Geremia
Please introduce yourself
My name is Jack Geremia, a chemist by training and a serial biotech founder. I build companies with a focus on solving technical challenges that unlock a significant existing market and scientific need. I enjoy company ideation and technology strategy, and then building high-performing teams capable of solving difficult problems. Currently, I serve as CEO and co-founder of Matterworks Inc., where we are harnessing recent breakthroughs in artificial intelligence to unlock unstructured molecular data that predict and interpret complex biological phenomena.
What impact could your innovation have?
Despite massive human and financial investment in interpreting omic data, the life sciences are still not that great at predicting complex emergent biology from these data. A prime example is drug efficacy. Limited ability to predict whether a novel therapy will be safe and effective is evident in the immense cost of developing new drugs. We see the potential to more than double the success rate of life-science pipeline R&D.
Was there a key breakthrough or major “eureka” moment during development?
Our eureka moment was realizing that molecular data, the most powerful data for explaining and predicting complex emergent biology, are already acquired at sufficient scale to enable a fundamental shift in how we perform life-science R&D. As a discipline, we simply lacked the right tools to exploit these data because of their immense complexity. Recent breakthroughs in machine learning and AI in just the past few years offer new mathematics and machinery remarkably adapted to processing these otherwise underused but powerful scientific data. To put it simply, mass spectrometers see orders of magnitude more biochemistry than our traditional peak-based methodologies extract. ML changes that.
Did you collaborate with any external teams during development?
We believe that the best new technologies are developed via early engagement and critical feedback provided by smart users skilled at evaluating and embracing new technologies. At Matterworks, we have been fortunate to benefit from multiple impressively innovative groups across the pharma and biotech industries. These groups and scientific leaders recognize the potential to solve some of their biggest challenges and were willing to devote time and effort to evaluate early versions of our technology. We could not be more appreciative of that feedback and collaboration.
Do you have a “philosophy of innovation”?
My innovation philosophy is to identify large unmet needs where there is an almost obvious technical solution, just that doesn’t yet quite exist. Such solutions often lie at the interface of multiple traditional scientific and engineering disciplines. I enjoy building the multi-disciplinary team required to assemble a cross-functional expertise base and then work hard to get the different groups communicating and problem solving together.
Are you driven more by scientific curiosity or the desire to make an impact?
Definitely the desire to make an impact via scientific innovation
Is there anything missing from the analytical scientist’s toolbox today?
Generally, across the analytical sciences, the techniques and analyses are so complex, that we have had to separate data generation from data interpretation. Yet, experience teaches us that the more we consolidate data generation and interpretation, the better the use of those data become. Advances in machine intelligence are changing the rulebook for how we divide analytical sciences into data acquisition and insight generation. There is an immense opportunity in front of us to allow machine interpretation of raw data so that the scientific interpretation does not need to be oversimplified.
What big problems could analytical innovation help to solve over the next decade?
Predictive tools to leverage pre-clinical data to identify whether a drug will exhibit sufficient efficacy or unexpected toxicity, will be a game changer. A broader look at untargeted data will lead to new modalities for disease mechanism and intervention.
What’s next for your team?
Our first product demonstrated that high-quality annotation can be achieved with a fraction of the time and cost associated with the analysis of high-resolution mass spectrometry data. Now we are expanding into a greater scope of annotation. Most exciting: we have just seen the first examples of building pre-trained models that can replace certain high cost and effort laboratory assays. This is an immense opportunity that we are well positioned to tackle in the upcoming year.
