Label-Free Biomolecule Detection in Physiological Solutions With Enhanced Sensitivity Using Graphene Nanogrids FET...

Recently, graphene nanogrid sensor has been reported to be capable of sub-femtomolar sensing of Hepatitis B (Hep-B) surface antigen in buffer. However, for such low concentration of Hep-B in serum, it has been observed during real-time operation that there is an overlap of around 50% in the drain-source current sensitivity values between different concentrations of the target biomolecule, in the range from 0.1 to 100 fM. This has been attributed to the fact that the concentration of non-specific antigen in serum being significantly higher than that of the target antigen, there is a considerable deviation in the number of captured target antigen for the same concentration. Further, this degree of overlap varies from one set to another set of sensor, depending on the statistical variations in the sensor fabrication process. This phenomenon challenges the quantification of target antigen for ultralow limit in physiological analyte. In this paper, we introduce probabilistic neural network (PNN) for quantification of Hep-B down to 0.1 fM in serum using graphene nanogrids field-effect transistor biosensor. The sensor has been operated in heterodyne mode in the frequency range of 100 kHz to 1 MHz applied between drain and source to overcome the problem of Debye screening effect. The application of PNN limits the quantification error within 10% in the range of 0.1 to 100 fM in contrast to 77% and 66% using polynomial fit and static neural network models, respectively. Further, the proposed methodology lowers the detection limit of Hep-B in serum by more than three orders of magnitude compared with the state-of-the-art, real-time, label-free sensors.