Forward error correction biosensors: Principles, modeling, and fabrication

Reliability is a field of research that has been largely overlooked in the area of biosensing. As a result, rapid-response biosensors which have been shown to work remarkably well under controlled laboratory conditions, fail to reproduce similar results when deployed in the field. This degradation in reliability can be attributed to several factors which include device level artifacts, variations in experimental protocols, transducer and measurement noise, stochastic interaction between biomolecules and background interference. With advances in micro-nano-fabrication, the emerging biosensors can integrate an increasing number of detection elements on the same device. This has opened the possibility that exploiting spatial redundancies across multiple detection experiments could be used to alleviate the effects of biosensor noise and artifacts. This system level approach, also known as "forward error correction" (FEC) has been extensively used for designing ultra-reliable communication and storage systems However, its application in the area of biosensing has remained largely unexplored. This thesis represents a first-of-its-kind research that investigates novel FEC interfaces to biosensors which potentially could lead to detection systems that deliver near-perfect reliability in detecting pathogens.;This thesis demonstrates that the application of encoding-decoding principles in biosensors can improve the reliability and the accuracy of biosensors. Based on the nature of biomolecular interactions, we first develop a stochastic model for affinity-based biosensors that mathematically captures the nature and sources of "biological" noise. We then present a framework for designing and evaluating biosensor encoder and decoding algorithms based on FEC principles that can improve the reliability of pathogen detection. To demonstrate these concepts, two biosensor platforms (lateral flow immunosensor and silver-enhanced gold nanoparticle based biochip) have been adopted as model biosensor encoders. In each of these platforms different encoder structures have been generated by synthetically patterning redundant biological probes. One of the contributions of this thesis is to demonstrate that a biosensor asymmetric code delivers a more reliable performance than a repetition code which is commonly used in many of the existing biosensor platforms. The thesis also describes a corresponding factor-graph based decoding algorithm which is used to detect the presence or absence of target biomolecules in a given sample.;The equivalent biosensor circuit models and the encoding-decoding algorithms have been integrated into a computational framework which facilitates rapid evaluation of FEC strategies for biosensors without resorting to painstaking and time-consuming experimental procedures. In behavioral simulation study, we demonstrate the efficacy of employing biosensor encoding/decoding scheme. One of the salient outcome of this study is a novel "co-detection" principle that uses nonlinear coupling property of the asymmetric code to detect trace quantities of pathogen in a given sample. "Co-detection" is similar in spirit to many noise exploitation techniques like stochastic resonance which has been reported in physics and biology, where it has been shown that addition of random noise into a non-linear system in fact improves the system sensitivity. We report experimental results where the "co-detection" principle has been successfully used to detect of trace quantity of mouse IgG in the presence of large background biomolecules. One of the key applications where "co-detection" could be used in the future is in the early diagnosis of Human immunodeficiency virus (HIV).