Machine learning-optimized charge transfer nanocrystals for synergistic NIR photothermal therapy, ferroptosis, and electrochemical detection of breast cancer.

Despite significant advances in near-infrared (NIR) photothermal agents, most current systems face challenges including large band gaps and poorly defined crystalline morphology, compromised responsiveness in the tumor microenvironment due to inefficient and time-consuming optimization, and dependence on separate, cancer type-specific diagnostic platforms for effective treatment. To overcome these challenges, we rationally synthesized a novel class of charge transfer complex-derived nanocrystals (CTNs) with well-defined morphology and a significantly reduced band gap in the NIR-I region, achieved for the first time through the unique combination of DBT as the electron donor with F4TCNQ, TCNB, and DTTCNQ as acceptors. Additionally, we employed machine learning (ML) to systematically optimize donor-to-acceptor ratios in CTNs, ultimately identifying DBT-F4TCNQ as the most effective formulation, with strong NIR absorption, excellent photothermal conversion, and superior electrocatalytic efficiency. Interestingly, ML-optimized DBT-F4TCNQ derived from CTNs demonstrated effective cysteine-triggered intracellular depletion of glutathione (GSH), thereby promoting reactive oxygen species (ROS) accumulation and inducing ferroptosis. Additionally, the in vivo antitumor efficacy was validated through NIR-triggered photothermal therapy combined with ROS-mediated ferroptosis, while the diagnostic potential of the designed CTNs was demonstrated via in-direct electrochemical monitoring of choline released from 4T1 cancer cells. Thus, the designed ML-optimized CTNs derived from the novel DBT-F4TCNQ combination hold great promise as a next-generation theranostic platform by integrating therapeutic efficacy with diagnostic precision.