Smell cancer by machine learning-assisted peptide/MXene bioelectronic array
Jiawang Hu,Nanlin Hu,Donglei Pan,Yan Zhu,Xuan Jin,Shikai Wu,Yuan Lu
DOI: https://doi.org/10.1016/j.bios.2024.116562
2024-10-15
Abstract:Non-invasive detection of tumors is of utmost importance to save lives. Nonetheless, identifying tumors through gas analysis is a challenging task. In this work, biosensors with remarkable gas-sensing characteristics were developed using a self-assembly method consisting of peptides and MXene. Based on these biosensors, a mimetic biosensor array (MBA) was fabricated and integrated into a real-time testing platform (RTP). In addition, machine learning (ML) algorithms were introduced to improve the RTP's detection and identification capabilities of exhaled gas signals. The synthesized biosensor, with the ability to specifically bind to targeted gas molecules, demonstrated higher performance than the pristine MXene, with a response up to 150% greater. Besides, the MBA successfully detected 15 odor molecules affiliated with five categories of alcohols, ketones, aldehydes, esters, and acids by pattern recognition algorithms. Furthermore, with the ML assistance, the RTP detected the breath odor samples from volunteers of four categories, including healthy populations, patients of lung cancer, upper digestive tract cancer, and lower digestive tract cancer, with accuracies of 100%, 94.1%, 90%, and 95.2%, respectively. In summary, we have developed a cost-effective and precise model for non-invasive tumor diagnosis. Furthermore, this prototype also offers a versatile solution for diagnosing other diseases like nephropathy, diabetes, etc.