Investigating the Sensing Mechanisms of Carbon Nanotube Aptasensors for Point of Care Hormone Tests

<p><strong>Point of care medical tests have the potential to greatly improve the quality, affordability, and accessibility of health care services. This work uses carbon nanotube network field effect transistors as a biosensing platform for point of care tests. The primary aim of this work is the detection of oestradiol, a steroidal sex hormone that impacts many areas of health. To do so, transistors are functionalised with DNA aptamers, which act as selective receptors. Despite the proven performance of carbon nanotube aptasensors, further work is required to fully characterise their behaviour and understand their sensing mechanisms to optimise the sensing signal. This work investigates the behaviour of these sensors by directly comparing two different oestradiol aptamers on the same carbon nanotube field effect transistor platform, showing qualitative differences in sensor performance that relate to the receptor molecule. The performance of the underlying transistor is compared against the strength of the sensing signal, for oestradiol sensors alongside carbon nanotube aptasensors for phenylalanine. Lastly, because the method for quantifying the real time sensing signal of a field effect transistor aptasensor is not thoroughly investigated in the literature, this work develops several methods to extract a numerical reading of the sensing response and compares them using real sensing data.</strong></p><p>Oestradiol sensors were developed using two different aptamers as receptors: a widely published aptamer from Alsager et al, AL-35, and a new truncation of a more recent aptamer from Jauset-Rubio et al, JR-31. We find similar overall performance, with both aptamers consistently detecting oestradiol at a concentration of 1.0 µM. We found qualitative differences between the behaviour of the two sensors. While AL-35 aptasensors give stronger signals, at concentrations of oestradiol well below the limit of detection the behaviour of the AL-35 aptasensors was far more erratic, resulting in an overall worse signal to noise ratio than the JR-31 aptasensors. This shows that there are aspects of aptasensor behaviour that cannot be easily determined by characterising the aptamer separately from the platform.</p><p>The relationship between transistor figures of merit and the magnitude of the sensing signal was investigated in both oestradiol sensors and phenylalanine sensors. While the expected result would be for a strong correlation between sensitivity and one or both of transconductance and device current, no such correlation is found. Instead, there is a moderate correlation found between sensitivity and threshold voltage, showing that the impact of transistor performance on sensor performance is small relative to a confounding variable. The correlation with threshold voltage suggests that the confounding variable is the functionalision of the CNT network with aptamers, which is an aspect of these devices that requires further work.</p><p>The techniques for quantifying the response of real time FET aptasensors are rarely discussed in the literature. CNT aptasensors can exhibit significant noise and drift in current, which introduces ambiguity in the magnitude of the signal. To systematically quantify the response of these sensors, we develop three techniques with increasing capacity for drift rejection, and compare these techniques on the sensing data collected for this thesis. We show that the lack of drift rejection in simple averaging results in almost unintelligible numerical results for most sensor signals, with large responses sometimes registering as smaller than non responses due to drift in the current. The simple differencing method introduces drift rejection and gives much more consistent numerical values that correlate strongly with a qualitative analysis of the sensing data. Attempting to improve the drift rejection further by using linear fits of local regions of the sensing data resulted in a dramatic loss of robustness to noise, resulting in the quantified responses being less consistent and more difficult to interpret.</p>