Direct Identification and Quantification of Peptide Biomarkers in 20% Human Serum Using a Solid-State Nanopore Platform and Machine Learning

Abstract The early detection of disabling or life-threatening diseases remains a major challenge in healthcare and relies on the direct identification of disease-specific protein and peptide biomarkers in biological fluids. Here, we establish a polymer-engineered solid-state nanopore platform for the direct electrical identification and quantification of peptide biomarkers in 20% human serum. A semidilute polyethylene glycol network generates an osmotic driving force that promotes peptide capture while screening nonspecific interactions responsible for nanopore clogging. We combine this strategy based on polymer physics with machine-learning-assisted digital serum cleaning to overcome the complexity of biological samples. This method thus allows for the extraction of peptide-specific electrical fingerprints from the serum background. As a proof of concept, we investigate three closely related kinin biomarkers: bradykinin (BK), des-Arg9-bradykinin (DABK), and Lys-bradykinin (LBK), which differ by only one or two terminal amino acids. Despite their high structural similarity, each peptide generates a distinct electrical signature, allowing reliable single-molecule discrimination in both buffer and serum. The platform further identifies and quantifies these biomarkers in mixtures of varying composition in 20% serum. More broadly, this work establishes a general strategy for the direct electrical decoding of closely related peptide biomarkers in clinically relevant biofluids.