Smart electrochemical sensing of xylitol using a combined machine learning and simulation approach

• Published in: Talanta open Vol. 6; p. 100144
• Main Authors: Uwaya, Gloria Ebube, Sagrado, Salvador, Bisetty, Krishna
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2022; Elsevier
• Subjects: Density functional theory; Electrochemical sensor; gold; machine learning; molecular dynamics; Monte Carlo; MWCNT; xylitol; gold; Monte Carlo; Density functional theory; xylitol; MWCNT; Electrochemical sensor; molecular dynamics; machine learning
• ISSN: 2666-8319; 2666-8319
• DOI: 10.1016/j.talo.2022.100144

•Nanocomposites of multiwalled carbon nanotube (MWCNT)/Biosynthesised gold nanoparticles (AuNPs) have been employed in the design of a picomolar sensor for xylitol detection.•Chemically reactivity of the analyte confirmed by HOMO-LUMO plots obtained from density functional theory calculations.•Monte Carlo and molecular dynamics simulations of GCE/MWCNT/AuNPs-xylitol complex correlated with the experimental CV and EIS results.•ANN machine learning was used to prediction of the voltametric signal with good accuracy.•An excellent detection limit of 9.8 × 10−6 pM was obtained at the designed sensor and a good practicability with satisfactory recoveries of 97–100% with 2.83–3.33% RSDs in chewing gum. A novel sensor was proposed for the detection of xylitol in sugar free chewing gum using Au nanoparticles (NPs) derived from Callistemon viminalis leaf extract coupled with multiwalled carbon nanotubes (MWCNTs) doped onto glassy carbon electrode (GCE). In comparison to the bare GCE, the modified GCE/MWCNT/AuNPs sensor showed about 45-fold better electrochemical response to xylitol. Under the optimal conditions, the designed sensor achieved a detection limit of 9.8 × 10−6 pM for concentrations ranging from 9.9 × 10−6 to 2.9 × 10−5 pM. The practicability was tested on sugar-free sample yielding recoveries of 97–100% with RSDs of 2.83–3.33%. Machine learning (ML) was used to predict changes in voltammetric signal with changing potential over time demonstrating the fundamental knowledge of the electrochemical reaction. The performance of the Artificial Neural Network (ANN) provides good accuracy and precision in predicting the intensity (I) along with repeated ANN runs, with a mean square error (MSE) of 0.007 (± 0.002) and a determination coefficient (R2) of 0.9992 ± 0.0006. Additionally, the interaction of xylitol on the electrode surfaces were investigated using Monte Carlo adsorption studies and 1000 ps Molecular Dynamics simulations under NVT conditions. According to the frontier molecular orbitals obtained through Density Functional Theory calculations, the reactive sites of xylitol occur at the hydroxyl group on the second carbon. Using complementary measurement techniques, this new strategy exhibits a great potential for rapid detection of xylitol in food and dental products. [Display omitted]