Year: 2026
Type: preprint
Abstract: Water-soluble near-infrared (NIR) emitters are of interest for imaging in the tissue transparency window. DNA-stabilized silver nanoclusters (DNA-AgN) hold promise for this spectral window, but it remains challenging to select oligonucleotides that stabilize NIR-emitting DNA-AgN. We address this challenge using transfer learning (TL), constructing a model that learns from experimental data for DNA-AgN stabilized by short 10-base oligonucleotides to then design longer oligonucleotides for larger, NIR-emitting nanoclusters. Using this model, we experimentally uncover 321 new DNA-AgN with peak emission beyond 800 nm, increasing design success by 710%. DNA-AgN designed using TL exhibit favorable properties including substantial emission above 1,000 nm, high fluorescence quantum yield, upconversion fluorescence, and a 14-electron count - the highest reported for DNA-AgN. These results demonstrate the power of TL for extracting physiochemical information from a limited parameter space and extrapolating these principles for broader design challenges, which is especially important when training data is scarce.
Source: ChemRxiv
Institutions: University of California System, University of Copenhagen
Language: English
Subfield: Materials Chemistry
Field: Materials Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7212135451