Deep Learning-assisted design of fluorescence nanobiosensors

The ability to design and synthesize nanomaterials with specific photophysical properties is not only a great intellectual challenge, but also one with important practical consequences. To address this challenge, we are currently exploring a new class of biolabels termed few-atom noble metal nanoclusters. Noble metal nanoclusters are collections of small numbers of gold or silver atoms (2-30 atoms) with physical sizes close to the Fermi wavelength of an electron (~0.5 nm for gold and silver). Providing the missing link between atomic and nanoparticle behavior in noble metals, these nanoclusters exhibit optical, electronic, and chemical properties dramatically different from those of much larger nanoparticles or bulk materials. Among those water-soluble noble metal nanoclusters newly developed, DNA-templated silver nanoclusters (DNA/AgNCs) have attracted great interest in biosensing owing to a number of useful photophysical and photochemical properties. For instance, controlled conversion of DNA/AgNCs between bright and dark states by guanine proximity has led to the invention of a new molecular probe, termed a NanoCluster Beacon (NCB), that “lights up” upon binding with a DNA target. Not relying on Főrster energy transfer as the fluorescence on/off switching mechanism, NCBs have the potential to reach an ultrahigh signal-to-background (S/B) ratio in molecular sensing. Since the fluorescence enhancement is caused by intrinsic nucleobases, NCB detection is simple, inexpensive, and compatible with commercial DNA synthesizers. We hold 3 US patents on the silver nanocluster probes (10407715, 9499866, 8476014) and are currently collaborating with Dr. Jennifer Brodbelt in UT Chemisty and Dr. Minjun Kim at SMU in using mass spectrometry and nanopores to study DNA/AgNCs. More details can be foud in Yeh et al., Nano Letters, 2010; Yeh et al., JACS, 2012; Obliosca et al., ACS Nano, 2014; Chen et al., JACS, 2015; Blevins, ACS Nano, 2019.

We are currently using deep learning models to predictively design silver nanocluster sensors with desired colors, on/off ratios and photostabilities.