Fast DNA-Based Molecular Classifier for Cancer Diagnosis Using Freeze-Thaw Cycling†

Abstract

Monitoring and analyzing expression levels of multiple biomarkers in biological samples can improve disease risk prediction and guide precision medicine but suffers from high cost and being time-consuming. Here, we construct a fast molecular classifier based on freeze-thaw cycling that implements an in silico support vector machine (SVM) classifier model at the molecular level with a panel of disease-related biomarkers expression patterns for rapid disease diagnosis. The molecular classifier employs DNA reaction networks as the computing module and repeated dehydration and concentration process as the driving force to implement a set of simplified mathematical operations (such as multiplication, summation and subtraction) for efficient classification of complex input patterns. We demonstrate that the fast DNA-based molecular classifier enables precise cancer diagnosis within a short turnaround time in synthetic samples compared to those of free diffusion classifiers. We envision that this all-in-one molecular classifier will create more opportunities for inexpensive, accurate, and rapid disease diagnosis, prognosis and therapy, particularly in emergency departments or the point of care.