Tackling the SNR problem in miniaturized arrayed biosensors for water

The SNR is the main factor determining the lowest limit of detection (LOD) of a sensor [7]. The most straightforward way to increase the SNR is by reducing the noise. All substrates exhibit some level of autofluorescence and reflect the exciting light that leaks through emission filters to some extent [8]. These factors can be controlled with the chemical properties of the surface and its coating. Moreover, nonspecific binding might add high background noise into the system, making it difficult to detect a weak signal, thus increasing the LOD. This problem is even more pronounced with protein arrays; unlike the negatively charged nucleic acids, the amphiphatic protein may interact with the surface due to hydrophobic interactions, electrostatic interactions, and van der Waals forces. The large variation between interactions requires optimization of blocking protocols for each sample solution, when one sample may contain as many as thousands of different molecules to be analyzed [9]. An ideal blocking protocol will reduce the nonspecific binding, while not disturbing the specific binding, thus reducing the noise without harming the signal. Increasing the signal is obviously an additional way of increasing SNR. The signal intensity is closely related to the substrate and the surface chemistry that immobilizes the capture layer to the solid support. The majority of microarray substrates are made of glass coated with an immobilization layer that interacts with the biomolecules by either physical adsorption or covalent binding. Physical adsorption can be achieved through positively charged residues (like aminosilane) that attract the negatively charged DNA strands, or electrostatic and hydrophobic interactions of the protein with the surface. It is the simplest process; however, it is unstable and susceptible to stringent washing in comparison to other methods. For covalent immobilization, the glass slide is coated with functional groups that form a covalent link to the amine residues found on both nucleotides and proteins (e.g., epoxy or aldehyde coatings). According to Ekins’s ambient analyte theory [10], microspots are expected to achieve better sensitivity than ligand-binding assays, even though the latter use much larger active sites. As the spot size is reduced and its area decreases, fewer target molecules are needed in order to achieve maximal signal density (full coverage of the spot); thus a lower LOD is expected (see Fig. 5.1) [11]. Also, for small enough spots, the binding process does not significantly reduce the concentration of the target molecules in the sample; thus quantification of the analyte is more accurate. Therefore, an “ambient analyte assay” is not affected by the volume of the sample in use, and it allows high sensitivity with relatively low sample volume [12]. For this reason it is postulated that optimal SNRs can be achieved only in spots smaller than a certain size, depending on the binding site density. However, ambient analyte theory assumes an ideal surface, that is, homogenous and infinite binding site density of the surface, so the signal density (signal intensity per unit area) remains constant as the spot area is reduced. Recently we have shown [13] that this assumption is not valid for spots around ~1 µm. Instead, the density and homogeneity of binding sites play a critical role in sensitivity and reproducibility of miniaturized arrays with spots of these dimensions.