New Publication: Classification of Colorectal Cancer cfDNA Using 3 nm Nanopores and Machine Learning
In the article we present a label- and amplification-free approach for classifying cell-free DNA (cfDNA) using ultra-small (~3 nm) solid-state nanopores paired with high-bandwidth electronics. By recording the ionic current signatures of individual cfDNA molecules, we capture biophysical information preserved in native, unmodified DNA.
To decode these signals, we developed a hybrid CNN-Transformer machine learning model that processes raw time series current, achieving ~95% accuracy in distinguishing healthy from colorectal cancer-derived cfDNA. This work establishes a proof-of-concept for rapid, low-volume cfDNA profiling as a potential point-of-care framework for cancer diagnostics.