High-Throughput DNA Sequence Space Exploration for Carbon Nanotube Biosensor Design

<jats:p>Fluorescent single-walled carbon nanotubes (SWCNTs) can be non-covalently functionalized with biopolymers such as DNA to create biosensors. Noncovalent functionalization promises a vast design space, but synthesis is limited by low-throughput dispersion steps. Here, we present a high-throughput SWCNT-based platform that utilizes covalent DNAanchor chemistry to recruit ssDNA recognition motifs via hybridization. The approach circumvents low throughput processing steps such as sonication or centrifugation. By integrating automated liquid handling and spectroscopy, we bypass traditional fabrication bottlenecks and synthesize 180 unique sensors and test 540 analyte/sensor conditions in less than three days. This accelerated workflow generates large datasets that are combined with an iterative machine learning (ML) pipeline. We demonstrate the versatility of this approach by evolving sensors with enhanced sensitivity for the neurotransmitter dopamine over common interfering molecules such as ascorbic acid in complex media. Through iterative rounds of ML-driven prediction and synthesis, we establish a scalable framework for rapid synthesis of functional nanosensors and increase the sensitivity for dopamine for up to 89 % of the evolved sensors. This strategy bridges the gap between the vast DNA sequence space and the generation of sensor variants, offering a scalable path to next-generation sensors.</jats:p>