Piercing the Shadows: Exploring the Influence of Signal Preprocessing on Interpreting Ultrasensitive Bioelectronic Sensor Data

Abstract

The development of ultrasensitive electronic sensors for in vitro diagnostics is essential for the reliable monitoring of asymptomatic individuals before illness proliferation or progression. These platforms are increasingly valued for their potential to enable timely diagnosis and swift prognosis of infectious or progressive diseases. Typically, the responses from these analytical tools are recorded as digital signals, with electronic data offering simpler processing compared to spectral and optical data. However, preprocessing electronic data from potentiometric biosensor arrays is still in its infancy compared to more established optical technologies. This study utilized the Single-Molecule with a Large Transistor (SiMoT) array, which has achieved a Technology Readiness Level of 5, to explore the impact of data preprocessing on electronic biosensor outcomes. A dataset consisting of plasma and cyst fluid samples from 37 patients with pancreatic precursor cyst lesions was analyzed. The findings revealed that standard signal preprocessing can produce misleading conclusions due to artifacts introduced by mathematical transformations. The study offers strategies to mitigate these effects, ensuring that data interpretation remains accurate and reflective of the underlying biochemical information in the samples.

Keywords: Chemometric data processing; Data preprocessing; Electrolyte gated field effect transistors; Pancreatic cancer early detection; Single molecule with a large transistor-SiMoT.

Figure 1

(a) Schematic representation of the SiMoT array, featuring the ELISA‐like disposable cartridge, the ELISA plate lid, displaying the sensing gates, and the case housing the read‐out electronics connected to a smart device. (b) Close‐up view of a single EG‐OFET device located at the bottom of each well in the 96‐well ELISA plate, along with a top biofunctionalized sensing gate. Visual depictions of biofunctionalized gates tested in triplicate, to the assay of (c) MUC1, (d) CD55 and (e) KRAS. (f) Schematic representation of the gate used for the negative control experiment, which is coated with Bovine Serum Albumin.

Figure 2

Plasmon peak reflectance intensity vs SPR angle measured for the deposition of (a) anti‐MUC1, (b) anti‐CD55 and (c) b‐KRAS. The black curve refers to the gold surface plus chemical SAM, while red curve refers to the modified surface with the three different bio‐recognition elements. In the inset, the angular shift measured after the deposition is shown. Current shift compared to the baseline current level measured when the gate is exposed to a concentration of 10 nM of (d) 7 negative control experiments, and to a concentration of 10 zM of (e) MUC1, (f) CD55 and (g) KRAS.

Figure 3

Representative example of the I_D current data points measured at V_D=−0.4 V and V_G=−0.5 V vs. cycles in a well of a SiMoT multiplexing array. (a) Average current data points along with the standard deviation (grey shadow) measured for 7 negative control experiments (see main text for details) with BSA biofunctionalized gates. These experiments provided the Gaussian distribution of the noise's statistical parameters and the LOI level. (b) Data‐points for a MUC1 assay in a cyst fluid are given as the average of three replicates on three distinct gates while the shaded area are the standard deviations. Operatively, the gate functionalized with anti‐MUC1is at first incubated in PBS and the baseline (I_{0_1–20,} blue labels) is measured in deionized water; subsequently the same gate is incubated in plasma sample from a high‐grade mucinous cyst patient and the sensing current (I_1–20, red labels) is measured in water. The dashed green line is the LOI level (see panel a) while f₁, f₂, and f₃ are the features used for the Principal Component Analysis (see text for details).

Figure 4

(a) Score plot evaluated on the SiMoT array data matrix after column autoscaling. Red, green, and black labels represent the high‐grade, potentially low‐grade, and potentially non‐mucinous cysts, respectively. PC1 explains the 52.7 % of the variance, while PC2 the 13.1 %. (b) Loading scatter plot, depicting how each feature contributes to PC1 and PC2; the features f1, f2, and f3 are represented by grey, blue, and violet arrows, respectively. (c) Score‐Plot and (d) Loading scatter plot evaluated on the SiMoT array data matrix after SNV row scaling. PC1 explains the 45.1 % of the variance, while PC2 the 18.9 %.