Machine learning-assisted multi-aptamer encoding enables pattern-level discrimination of HbA1c in complex biological matrices

A multi-aptamer, machine-learning–assisted sensor array is presented that integrates polyadenine (poly-A)–oriented aptamer functionalization with salt-stable, gold nanoparticles (AuNPs) colorimetry. Poly-A anchoring enables controlled aptamer orientation and improved colloidal stability under high ionic strength, allowing sensitive aggregation-based detection. Despite the inability of individual aptamers to discriminate glycated hemoglobin (HbA1c) from total hemoglobin (tHb), their combined multichannel responses provide distinct optical patterns that machine learning resolves into non-overlapping clusters. The platform successfully differentiated between clinical samples from diabetic and healthy individuals. This work establishes a generalizable strategy, interface engineering coupled with supervised pattern recognition, for resolving highly similar proteins and developing practical, low-cost point-of-care (POC) diagnostic tools.