Machine Learning-Assisted Lung Cancer Diagnosis using Rapid Isothermal Amplification of Serum microRNAs

Abstract Early diagnosis of non-small cell lung cancer (NSCLC) is essential for improving patient prognosis. However, existing serum biomarkers remain limited in sensitivity, stability, and clinical translation. Here, we developed an NSCLC liquid biopsy strategy integrating tissue-serum consensus miRNA screening, machine-learning-based classification, and tunable cascade isothermal amplification detection. Candidate miRNAs were identified from tissue and serum datasets, and a four-miRNA-based support vector machine (SVM) model was constructed for NSCLC classification. Furthermore, a cascade isothermal amplification circuit composed of linear and exponential amplification modules was designed for rapid, programmable, and quantitative miRNA detection. This circuit achieved target recognition, signal transduction, and cascade amplification under constant-temperature conditions within 20 min. Clinical validation showed good agreement between the cascade circuit and RT-qPCR results. The four-miRNA SVM model achieved area under the curve (AUC) values of 0.935 and 0.905 using RT-qPCR and cascade circuit data, respectively. In addition, a two-miRNA comparator model further demonstrated the compatibility of this platform with alternative classification strategies. Overall, this study provides a rapid, tunable, and clinically translatable molecular diagnostic strategy for serum miRNA-based auxiliary diagnosis of NSCLC.