Machine learning-enhanced molecular modeling of near-infrared fluorescent materials for bioimaging

The National Institute of Biomedical Imaging and Bioengineering awarded the Research Foundation of The City University of New York $151,168 on September 5, 2026, under the Discovery and Applied Research for Technological Innovations to Improve Human Health program (CFDA 93.286) to develop machine learning-enhanced molecular modeling of near-infrared fluorescent materials for bioimaging applications.

The project addresses limitations in current near-infrared fluorescent probes used for surgical guidance and deep-tissue diagnostics. Existing probes operating in the NIR-I window (650–950 nanometers) are constrained by scattering and autofluorescence, while NIR-II probes (1000–1700 nanometers), particularly polymer dots, suffer from low quantum yields caused by aggregation-induced quenching and narrow energy band gaps that impede clinical translation. The recipient will establish predictive design rules for donor-acceptor copolymers through three complementary research aims: developing supervised machine learning models to predict electronic coupling across molecular dynamics-sampled configurations; applying unsupervised machine learning methods to identify backbone conformations governing band gaps and orbital distributions; and curating high-level ab initio datasets of optoelectronic descriptors integrated with experimental measurements and data-driven optimization to discover promising copolymer frameworks. By uniting machine learning with physics-based modeling, the project will replace trial-and-error synthesis with predictive design methodologies.

Work is performed in Bronx, New York. The period of performance runs through July 31, 2030.

Abstract

Near-infrared (NIR) fluorescence imaging has become indispensable in modern medicine, enabling real-time visualization for surgical guidance and deep-tissue diagnostics. Current FDA-approved probes are limited to the NIR-I window (650-950 NM), where scattering and autofluorescence compromise performance. The second NIR window (NIR-II, 1000-1700 NM) offers deeper penetration, reduced background, and improved signal-to-noise, yet existing NIR-II probes, including polymer dots (PDOTS), remain constrained by low quantum yields due to aggregation-caused quenching (ACQ) and narrow energy band gaps. This GAP limits clinical translation of otherwise promising PDOTS, which uniquely combine exceptional brightness, photostability, and biocompatibility. This project addresses these barriers through machine learning (ML)-enhanced molecular modeling integrated with multiscale simulations. The overarching objective is to establish predictive design rules for donor-acceptor copolymers that yield bright, stable, and biocompatible NIR-II PDOTS. Three complementary aims will be pursued: (1) develop supervised ML models to efficiently predict electronic coupling, a key quantum descriptor of ACQ, across millions of molecular dynamics-sampled configurations; (2) apply unsupervised ML methods, including dimensionality reduction and feature-extraction networks, to identify backbone conformations that govern band gaps and orbital distributions critical for NIR-II emission; and (3) curate high-level AB initio datasets of optoelectronic descriptors for donor and acceptor building blocks, integrating them with experimental measurements and data-driven optimization to discover promising copolymer frameworks. By uniting ML with physics-based modeling, the project will overcome computational bottlenecks, replace trial-and-error synthesis with predictive design, and establish transferable principles for NIR-II materials. Outcomes will include open datasets, interpretable ML models, and systematic guidelines for suppressing ACQ and enhancing emission efficiency. These advances will accelerate the development of next-generation fluorescent probes for image- guided surgery, oncology, and cardiovascular interventions. Beyond scientific impact, the project will provide interdisciplinary training at the interface of chemistry, materials science, and artificial intelligence, preparing students for careers in AI-enabled biomedical research.