A Bead-Based DNA Classifier for Multi-Input Cancer Diagnosis

Abstract DNA computation offers a promising route for multi-input cancer diagnosis, yet its translation into practical classifiers is often impeded by complex probe design and operational intricacies. Here, we introduce a bead-based DNA classifier (BDC) platform designed for multimiRNA-based diagnostic applications. This platform performs weighted molecular computation through a combination of magnetic bead capture and polymerase-mediated strand displacement. An in silico diagnostic model for breast cancer was built using an SVM algorithm trained on TCGA miRNA-seq data, which revealed a three-miRNA signature with assigned integer weights and yielded a computational AUC of 0.97. Molecular implementation involves processing the weighted miRNAs in separate parallel tubes and converting each input into a universal DNA barcode strand that is subsequently quantified via qPCR. A diagnostic readout is generated by directly comparing the CT values of the two reaction tubes. Clinical validation using 27 serum samples resulted in an accuracy of 81%. This study demonstrates, as a proof-of-concept, a streamlined DNA-computing workflow that simplifies probe design for multi-input classifiers, and shows preliminary utility for multi-input diagnostics using extracted miRNA from clinical samples.