Machine Learning-Assisted Screening of High-Entropy Sub‑1 nm Nanowires for Ultrasound-Augmented Pancatalytic Tumor Therapy

The
rational design of catalytic nanomaterials, particularly high-entropy
alloy (HEA) nanomaterials with their unique structures, is crucial
for advancing pancatalytic therapy and holds promise for catalytic
biomedical applications. Nevertheless, conventional trial-and-error
approaches to HEA development persist in being inefficient and resource-demanding,
underscoring the critical need for data-driven strategies to accelerate
the design of next-generation catalytic platforms. Herein, we report
a machine learning (ML)-assisted framework for the rational design
of subnanometer HEAs with enhanced multicatalytic activities for the
treatment of triple-negative breast cancer. Using an extensively curated
dataset of nanozyme compositions and activities, ML algorithms identified
key metal elements with the highest contribution to catalytic performance.
Guided by these insights, a distinct PtFeMoCoNiRu HEA subnanowire
(HESNW) was synthesized, in which the subnanostructure confers maximized
active-site exposure and superior atomic utilization, enabling exceptional
reactive oxygen species (ROS) generation. Ultrasound (US), characterized
by its noninvasive and deep penetration, is employed to activate catalytic
reactions. Excessive ROS production with HESNW under US irradiation
induces extensive DNA damage, activating the cGAS-STING signaling
pathway and leading to PANoptosis. This ML-guided design strategy
enables the precise tailoring of multicatalytic HEA nanomaterials
and provides a generalizable blueprint for accelerating the discovery
of multifunctional nanocatalysts through data-driven methodologies.