MIP-based electrochemical sensor with machine learning for accurate ZIKV detection in protein- and glucose-rich urine
Issued Date
2025-07-01
Resource Type
ISSN
00032697
eISSN
10960309
Scopus ID
2-s2.0-105001269084
Journal Title
Analytical Biochemistry
Volume
702
Rights Holder(s)
SCOPUS
Bibliographic Citation
Analytical Biochemistry Vol.702 (2025)
Suggested Citation
Sukjee W., Sirisangsawang P., Thepparit C., Auewarakul P., Puttasakul T., Sangma C. MIP-based electrochemical sensor with machine learning for accurate ZIKV detection in protein- and glucose-rich urine. Analytical Biochemistry Vol.702 (2025). doi:10.1016/j.ab.2025.115854 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/109421
Title
MIP-based electrochemical sensor with machine learning for accurate ZIKV detection in protein- and glucose-rich urine
Corresponding Author(s)
Other Contributor(s)
Abstract
Nowadays, a multitude of biosensors are being developed worldwide. However, a significant challenge arises when these biosensors are tested in real sample environments, as many of them fail to perform as expected. This can lead to ambiguous results and raise concerns about their reliability. In many cases, further data analysis is required to enhance the clarity and meaningfulness of the outputs. In this study, we investigated the acrylamide-methacrylic acid-methyl methacrylate-vinylpyrrolidone copolymer for fabrication of molecularly imprinted polymers, aimed at developing electrochemical sensors for the direct detection Zika virus in urine. Here, Zika virus detection by the biosensor in three types of urine possibly found in clinical samples including normal, high glucose (glucose >540 mg/dL) and high protein urines (protein >100 mg/dL). The results show that the signal obtained from normal urine increased with virus concentration, while it decreased in urine with high glucose or high protein level. Support vector machine was introduced to unify two opposite trends and resolve ambiguity of the data. It was able to sift through the noise and extract valuable information, thereby improving the reliability and achieved 91 % accuracy in detecting the analyte spiked into real patient samples.