Machine Learning-Assisted Biomass-Derived Carbon Dots as Fluorescent Sensor Array for Discrimination of Warfarin and Its Metabolites

Abstract

Warfarin (WAR), an effective oral anticoagulant, is of utmost importance in treating many diseases. Despite its significance, rapid and precise discrimination of WAR remains a formidable challenge, especially facing its structural analogs of metabolites. Here, three kinds of herb-derived N-doped carbon dots (NCDs) were greenly synthesized via a fast and simple microwave-assisted method. Three NCDs showcased respectable blue fluorescent (FL) properties and sensing capabilities for the discrimination of WAR and its metabolites. To improve accuracy in identifying WAR and its metabolites, a sensor array composed of three unique herb-derived NCDs was meticulously designed. Combined with the machine learning model, the sensor array displayed a strong immunity to interference in the discrimination of the WAR, even in unknown samples. Meanwhile, the FL sensing mechanism is deeply expounded. The methodology proffers broad prospects for biomass-derived nanomaterials and provides an effective and feasible project for pharmaceutical analysis by capitalizing on machine learning.