Visualizing Molecular Fingerprints in Living Systems at Unprecedented Physical Limits through Advanced Instrumentation and Artificial Intelligence

Visualizing the biochemical processes that govern subcellular biological functions is crucial for decoding various facets of human physiology and identifying biomarkers for diseases. I will introduce our efforts in developing computation-driven, bond-selective micro-spectroscopy technologies for vibrational imaging of intrinsic biomolecules. Through synergistic integration between advanced instrumentation and artificial intelligence, we overcome the conventional physical limits of vibrational imaging in speed, molecular specificity, and spatial resolution. First, I will introduce methods to surpass the imaging speed barrier through compressive sampling and deep learning-augmented ultrafast scanning, achieving dynamic metabolic imaging in live cells and high-throughput mapping of metabolites in large tissue sections. Next, I will share innovations in hyperspectral chemical imaging that enhance specificity and multiplexity for small biomolecules, with applications in cancer metabolic reprogramming and bacterial biochemical production profiling. Finally, I will present recent advances in ultrasensitive label-free chemical nanoscopy of cell metabolism via highly chirped visible stimulated Raman scattering and noise-robust spatial frequency amplification, revealing novel metabolic nanostructures related to viral replication within host cells and bacterial production of biofuels.