Forward Error Correcting Biosensors: Modeling,Algorithms and FabricationYang Liu∗, Evangelyn Alocilja †, Shantanu Chakrabartty∗[email protected],[email protected], [email protected]∗Department of Electrical and Computer Engineering†Department of Biosystems Engineering Michigan State UniversityEast Lansing, MI 48824 USAAbstract—Advances in micro-nano-biosensor fabrication areenabling the integration of a large number of biological recog-nition elements within a single package. As a result, hundredsto millions of tests can be performed simultaneously and canfacilitate rapid detection of multiple pathogens in a given sample.However, it is an open question as to how to exploit the high-dimensional nature of the multi-pathogen testing for improvingthe detection reliability of typical biosensor systems. Our researchover the past few years has addressed this question and inthis paper we briefly summarize our approach. Our underlyingprinciple is based on a forward error correcting (FEC) biosensorwhere redundant patterns are synthetically encoded on thebiosensor. A decoding algorithm then exploits this redundancy tocompensate for systematic errors due to experimental variationsand for random errors due to stochastic biomolecular interac-tions. The key milestones in this research are : (a) fabrication andmodeling of biomolecular circuit elements used for constructingthe FEC biosensor; (b) development of a simulation environmentfor rapid evaluation of encoding/decoding algorithms and (c)development of a “co-detection” protocol that exploits non-linearinteraction between different biomolecular circuit elements. Asa proof-of-concept our study and experimental results have beenbased on a conductimetric lateral flow immunosensor that usesantigen-antibody interaction in conjunction with a polyanilinetransducer to detect the presence or absence of pathogens in agiven sample.Index Terms—Forward error correcting biosensors, factorgraph, sum-product algorithm, co-detectionI. INTRODUCTIONBiosensors have emerged as important analytical tools forthe rapid detection of food-borne pathogens, which accordingto The United States Department of Agriculture (USDA)cause approximately 5,000 deaths every year [1]. A typicalarchitecture of a biosensor consists of a biological recognitionlayer as a reactive surface in proximity to a transducer whichconverts the binding between the analyte and the recognitionlayer into a measurable electrical or optical signal [2]. Withadvances in micro-nano fabrication, biosensor technology isnow available that can integrate a large number of recognitionand transducer sites on a portable device. As a result a largenumber of simultaneous detection experiments can be con-ducted in parallel, an example being nanoscale arrays reportedin [3]. At the same time detection technology has also seensignificant improvements where analyte concentration levelsranging from pico-molar (pM ) to femto-molar (fM) can nowbe detected [4]. Both these trends are summarized in Fig.1,Array density (probe/mm2)Reliability(error rate)Sensitivitymg/mLµg/mLng/mLpg/mL0.5%1%2% MicroarrayNanoarray102103104105106Array density (probe/mm2)Reliability(error rate)Sensitivitymg/mLµg/mLng/mLpg/mL0.5%1%2% MicroarrayMicroarrayNanoarrayNanoarray102103104105106Fig. 1. Summary of biosensor technology trends.which also depicts another important performance parameterthat is typically overlooked in biosensor design. Unfortunately,with miniaturization the reliability of biosensors typicallydetoriates as have been observed in large scale gene-chiparrays. Irrespective of the choice of the detection technology(DNA, aptamers or antibodies) the biosensors artifacts can becategorized as [5]: (a) systematic errors that were inducedby sample handling errors, variations in experimental condi-tions (pH or temperature), and errors introduced by deviceartifacts; and (b) random errors introduced due to stochasticnature of biomolecular interactions. From a modeling point ofview, these artifacts are similar to those that are observed innanoscale storage and computing systems for which forwarderror correction (FEC) techniques have been successfullyemployed [6].The objective of this research is to replicate the success ofFEC principles in designing reliable computing and storagesystems towards designing reliable biosensors. In this regard,the study addresses some of the key challenges in this long-term goal. The first step involves mathematical abstractionwhere simulation models are developed that capture the ex-978-1-4244-2879-3/08/$25.00 ©2008 IEEE 249Authorized licensed use limited to: Michigan State University. Downloaded on August 4, 2009 at 13:31 from IEEE Xplore. Restrictions apply.perimentally measured response of the biosensor circuit. Thesesimulation models are then used to: (a) understand the natureof the biosensor channel and in the process derive fundamentallimits of biosensor FEC; (b) rapid design and evaluationof different FEC encoding and decoding algorithms withoutresorting to painstaking experimental procedures. Our finalobjective is to close the design loop, where the reliability ofthe biosensor encoding and decoding algorithms are validatedusing in-lab experiments. In this paper, we summarize someof the highlights of this research and present some of the keyresults.II. FUNDAMENTAL BUILDING BLOCKS:FABRICATIONANDMODELINGEven though the encoding and decoding methods beingdeveloped can be applied to any multi-pathogen detectionprinciple based on either antibody, DNA or aptamers, wehave chosen a conductimetric lateral flow immunosensor asa model biosensor. The immunosensor utilizes a polyanilinenanowire based transducer to convert the binding between atarget antigen and its antibody into an electrical signal. Fig. 2shows the visualization of a multi-pathogen immunosensor andthe principle of a single immunosensor is illustrated in Fig. 3,which shows a cross-sectional view of the immunosensor.Before the sample is applied, the gap between the electrodesin the capture pad is open. Immediately after sample isapplied to the sample pad, the solution containing the antigenflows to the conjugate pad, dissolves with the polyaniline-labeled antibody and forms an antigen-antibody-polyanilinecomplex. The complex is transported using capillary actioninto the capture pad containing the immobilized antibodies. Asecond antibody-antigen reaction occurs and forms a sandwich(Fig. 3). Polyanilines in the sandwich extend