CascadeMAP: Autonomous Closed-loop Optimization of Enzyme Cascades via Microfluidics, Machine Learning and Agentic AI
- 1Masaryk University
Abstract
Enzyme cascades enable complex biochemical transformations, but optimizing them requires searching high-dimensional spaces of enzyme stoichiometry, reaction conditions, and buffer composition. Here we introduce CascadeMAP, an autonomous microfluidic platform for closed-loop optimization of enzyme cascades, integrating high-throughput microfluidics with Bayesian optimization and assisted by agentic analyses. We demonstrate the platform across two three-enzyme cascades: (i) a glycerol detection pathway monitored by fluorescence and (ii) a 1,2,3-trichloropropane degradation pathway monitored by label-free Raman spectroscopy, providing orthogonal detection modalities. Bayesian optimization reached comparable optima with threefold fewer experimental conditions than Design of Experiments. A multi-agent AI system automated processing 2.1 GB of experimental data, pattern recognition, and insight synthesis. Operating without human intervention for 7 days, CascadeMAP processed ~295,000 reactions across ~7,400 different conditions. This capability establishes a generalizable, resource-efficient framework for autonomously optimizing enzyme cascades and metabolic pathways and accelerates the development of biocatalytic and synthetic biological systems.
Related
- Biotechnological approaches for food frauds and rapid detection methods as preventive measures: a global perspective”, with a special emphasis on microfluidics
- 3D Printed X-ray Compatible Microfluidics for Online Characterization of Hexosomes: A Synchrotron SAXS-on-Chip Study with Molecular Dynamics Insights
- Artificial Avidity Optimization of T-cell Receptors by Integrating Deep Learning Model and Mammalian Cell Surface Display
- Active-learning-guided optimization of cell-free systems for genome-wide transcriptomic profiling reveals progressive layers of regulation
- Machine learning-guided dual optimization of the substrate specificity and thermostability of yeast alcohol dehydrogenase.
BibTeX
@misc{vasina2026cascademap,
title = {{CascadeMAP: Autonomous Closed-loop Optimization of Enzyme Cascades via Microfluidics, Machine Learning and Agentic AI}},
author = {M. Vasina and D. Kovar and M. Kizovsky and D. Lacko and P. Vanacek and M. Herich and E. Volf and L. Drdla and S. Cabalova and P. Sikorova and M. Jirasek and P. Solansky and J. Jezek and O. Samek and F. Dousek and H. Walner and P. Zemanek and A. deMello and Z. Pilat and J. Damborsky and S. Stavrakis and S. Mazurenko and Z. Prokop},
howpublished = {bioRxiv},
year = {2026},
month = jun,
doi = {10.64898/2026.06.04.730034},
url = {https://www.biorxiv.org/content/10.64898/2026.06.04.730034v2},
note = {CC BY},
}
RIS
TY - UNPB
TI - CascadeMAP: Autonomous Closed-loop Optimization of Enzyme Cascades via Microfluidics, Machine Learning and Agentic AI
AU - M. Vasina
AU - D. Kovar
AU - M. Kizovsky
AU - D. Lacko
AU - P. Vanacek
AU - M. Herich
AU - E. Volf
AU - L. Drdla
AU - S. Cabalova
AU - P. Sikorova
AU - M. Jirasek
AU - P. Solansky
AU - J. Jezek
AU - O. Samek
AU - F. Dousek
AU - H. Walner
AU - P. Zemanek
AU - A. deMello
AU - Z. Pilat
AU - J. Damborsky
AU - S. Stavrakis
AU - S. Mazurenko
AU - Z. Prokop
PY - 2026
DA - 2026/06/04
PB - bioRxiv
DO - 10.64898/2026.06.04.730034
UR - https://www.biorxiv.org/content/10.64898/2026.06.04.730034v2
AB - Enzyme cascades enable complex biochemical transformations, but optimizing them requires searching high-dimensional spaces of enzyme stoichiometry, reaction conditions, and buffer composition. Here we introduce CascadeMAP, an autonomous microfluidic platform for closed-loop optimization of enzyme cascades, integrating high-throughput microfluidics with Bayesian optimization and assisted by agentic analyses. We demonstrate the platform across two three-enzyme cascades: (i) a glycerol detection pathway monitored by fluorescence and (ii) a 1,2,3-trichloropropane degradation pathway monitored by label-free Raman spectroscopy, providing orthogonal detection modalities. Bayesian optimization reached comparable optima with threefold fewer experimental conditions than Design of Experiments. A multi-agent AI system automated processing 2.1 GB of experimental data, pattern recognition, and insight synthesis. Operating without human intervention for 7 days, CascadeMAP processed ~295,000 reactions across ~7,400 different conditions. This capability establishes a generalizable, resource-efficient framework for autonomously optimizing enzyme cascades and metabolic pathways and accelerates the development of biocatalytic and synthetic biological systems.
ER -