Artificial neural network engineering of high carrier mobility single-layer graphene interfaces for ultrasensitive FET-aptasensors in clinical PSA detection.

In the current work artificial neural network (ANN) optimization of the Hummers' method (ANN-MHM) is introduced to enable deterministic synthesis of high-mobility, monolayer reduced graphene oxide (rGO) for ultrasensitive FET aptasensing. ANN-guided synthesis produced GO with large lateral dimensions (2-4 μm), monolayer thickness (∼1.1 nm), and low defect density (ID/IG = 0.99), yielding rGO FET channels with carrier mobility of 317 cm2/V·s, nearly 30-fold higher than conventional Hummers-derived films. Integrated into a prostate-specific antigen (PSA) aptasensor, the device delivered dual-parameter transduction (ΔVcnp and ΔIds/Ids0), a 1 pg/mL LOD, and a 1 pg/mL-10 ng/mL dynamic range at physiological pH. Clinical validation in human serum plasma showed excellent agreement with ELISA (R2 = 0.99), while selectivity tests (Trypsin, Human Kallikrein-2 (hK2), Bovine Serum Albumin (BSA) in female plasma) confirmed negligible off-target response. Reliability analysis across eight devices, four fabricated from the same GO batch and four from independently synthesized batches, evealed 6% intrinsic sensor variation and only 4% batch-to-batch variation, demonstrating the robustness of the ANN-engineered synthesis route for scalable production. The fabricated sensors retained 104- 86% functionality after 3 weeks at 5 °C, confirming practical storage stability. By coupling machine learning with nanoelectronic biosensing, this work establishes a reproducible and scalable pathway for next-generation point-of-care diagnostics.