Integrated High-Throughput Microfluidic-Mass Spectrometry for Enhanced Biomedical Data Generation and AI Model Training

About This Grant

ABSTRACT Drug development is significantly hindered by traditional reaction optimization methods that are low-throughput and costly. Although AI models could accelerate the discovery of optimal reaction conditions, they require massive, high-quality datasets that current methodologies cannot generate. Conventional automation systems using liquid handlers and microwell plates can perform only thousands of reactions per day, while robust AI-driven prediction demands millions of data points. Our peer-reviewed studies demonstrate that a micro-robotic nanoliter dispenser combined with custom-engineered nanowell arrays featuring innovative microlens geometries and MALDI-MS imaging can reduce reaction volumes from 40 μL to 0.5 nL, cut processing times from 60 hours to 4 hours, and scale throughput from thousands to millions of reactions. This proposal addresses the critical data-generation bottleneck by developing a scalable, fully integrated micro-robotic screening platform capable of performing millions of nanoliter-scale chemical reactions in parallel. Our approach combines precise, on-demand droplet dispensing with integrated dielectrophoretic steering, depositing droplets onto densely packed nanowell arrays maintained under controlled incubation conditions. Rapid matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) imaging then generates high-dimensional spectral fingerprints that quantitatively reflect reaction outcomes, providing rich datasets for an AI-driven analysis pipeline employing state-of-the-art machine learning for dimensionality reduction, clustering, and regression. To demonstrate our technology's capabilities, we will specifically optimize the Suzuki–Miyaura cross-coupling reaction—a widely utilized synthetic process in medicinal chemistry—and, in doing so, establish a scalable framework for AI-driven optimization of complex pharmaceutical syntheses using MIRACS. In Phase I (1 year), we will integrate and validate the core micro-robotic components and MALDI-MS workflow using our proven methodologies, generating preliminary quantitative benchmarks for system performance. In Phase II (2–3 years), we will scale the platform, further refine droplet handling and nanowell technologies for industrial-scale throughput, and fully integrate the AI predictive pipeline. Led by Fluid Discovery Inc. and supported by expert academic collaborators, our multidisciplinary team is uniquely qualified to commercialize this technology, significantly enhancing capabilities to synthesize complex drug molecules.