MicroRNAs (miRNAs) serve as crucial biomarkers in disease diagnosis. Although silicon-based electronic machine learning models provide efficient means for analyzing the massive data associated with miRNAs and for disease diagnosis, they remain constrained by low parallelism and poor biocompatibility. DNA computing provides an ideal interface for the integration of information technology and biotechnology. Here, we employ strand displacement reaction (SDR) to realize the DNA-based support vector machine (SVM) models for disease diagnosis that sequentially integrates four functional modules: weighted summation module, subtraction module, signal restoration module, and reporter module. In contrast to conventional silicon-based computing architectures, the DNA based disease diagnosis model can directly and specifically identify the expression levels of target miRNAs from biological samples and drive model decisions in real time, enabling precise classification of disease states. We validate the disease diagnosis model using miRNAs expression levels, achieving diagnostic accuracies of 98.54% for cystic, mucinous and serous neoplasms (CMSN), 99.46% for Glioma, and 97.81% for clear cell carcinoma (CCC). The DNA-based disease diagnosis model simultaneously classifies the three disease states in parallel, and the results show a strong agreement with the actual disease states labeled in The Cancer Genome Atlas (TCGA). This work provides a new paradigm for engineering precise, intelligent and integrated disease diagnosis schemes at the molecular level.
Tang et al. (Thu,) studied this question.