Machine-Learning-Driven Dual-Mode Immunoassay Based on Nanobody and Natural Fluorescent Probe for Ultrasensitive Quantification of Pseudomonas aeruginosa

Abstract Pseudomonas aeruginosa is a Gram-negative opportunistic bacterial pathogen posing severe threat to environmental safety and human health. To achieve rapid and accurate detection, we developed a nanobody-based colorimetric-fluorescent dual-mode sandwich immunosensor using natural fluorescent probe and further integrated machine learning to enhance the signal readout and improve detection sensitivity. Here, specific nanobodies toward P. aeruginosa were successfully developed for the first time by biopanning against an immune phage-displayed nanobody library. Natural fluorescent quinine, which possesses intrinsic fluorescence and undergoes fluorescence quenching via an internal quenching effect, was employed as the fluorescent probe. With the baseline LODs obtained by direct fitting of the raw optical signals (9.005 × 105 CFU/mL for colorimetry and 3.02 × 103 CFU/mL for fluorescence), an XGBoost model was introduced and further reduces the LOD to 3.269 × 103 CFU/mL and 3.96 × 102 CFU/mL, achieving a 275-fold and 7.6-fold improvement in sensitivity, respectively. Furthermore, following a brief pre-enrichment, this method enables detection of P. aeruginosa at concentrations as low as 10 CFU/mL. Recoveries in water samples ranged from 81.47% to 119.47%, with a coefficient of variation below 8.44%, showing good agreement with machine learning predictions (R2 = 0.957). No cross-reactivity was observed with closely related or heterologous bacteria. This method provides a technical approach for the precise quantitative monitoring of P. aeruginosa in complex environments.