Terahertz (THz) fingerprint sensing utilizes the absorption of fingerprints generated by the unique vibrational characteristics of molecules to achieve substance-specific identification. By taking full advance of the anapole mode induced-biosensor consisting of out-of-plane metal-insulator-metal (MIM) configuration, the D-glucose solutions down to physiological level are accurately detected by proposed metasurface biosensor through the electromagnetic induced absorption (EIA) effect induced by the interaction between the metasurface and molecular vibrational fingerprint. Besides, by utilizing the vibrational fingerprint sensing ability, the pure D-glutamic acid and D-lactose, as well as their mixture have been quantitatively characterized. In addition, with the aid of machine learning algorithms, the designed single resonance metasensor achieves 100% recognition of five molecules. This work brings a convincing strategy for trace label-free molecular recognition for various species, which might extend the promising potentials of THz sensing techniques toward biomedical testing and clinical diagnosis.