Pancreatic ductal adenocarcinoma (PDAC) carries a dismal prognosis, and therapeutic drug monitoring (TDM) of gemcitabine (GEM) is critical for optimizing chemotherapy outcomes yet impeded by the complexity of conventional analytical techniques. Herein, we report a high-performance photoelectrochemical (PEC) sensor based on a novel BiVO 4 /Y 2 O 3 type-II n–n heterojunction for the ultrasensitive detection of GEM, integrated with a machine learning (ML) framework for PDAC treatment response prediction. BiVO 4 and Y 2 O 3 nanosheets were assembled via electrostatic self-assembly to establish an atomically sharp heterojunction interface. Density functional theory (DFT) calculations confirmed a staggered band alignment in which both the conduction band minimum and valence band maximum of Y 2 O 3 are thermodynamically positioned above those of BiVO 4 , driving efficient spatial separation of photogenerated charge carriers. This charge-separation mechanism reduces photoluminescence intensity 3.8-fold and lowers interfacial charge-transfer resistance to 181 Ω, yielding a photocurrent density of 7.02 µA cm −2 with a 1.8-fold enhancement over pristine BiVO 4 . The resulting PEC sensor exhibits a linear response toward GEM over 0.1–10 µM (R 2 = 0.9983), excellent selectivity against common blood-borne interferents, and satisfactory recovery (93.9–105.1%) in human serum. Sensor-derived GEM pharmacokinetic (PK) parameters were subsequently integrated as features into ML classifiers trained on a 400-patient PDAC cohort. The bestperforming logistic regression model achieved an AUC of 0.836 and significantly stratified progression-free survival (6.2 vs. 3.9 months; p = 7.91 × 10 −8 ). This work establishes a new paradigm bridging semiconductor heterojunction engineering with data-driven precision oncology for individualized chemotherapy management.
Zhang et al. (Fri,) studied this question.