IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 7, NO. 4, AUGUST 2013 451 FAST: A Framework for Simulation and Analysis of Large-Scale Protein-Silicon Biosensor Circuits Ming Gu, Member, IEEE, and Shantanu Chakrabartty, Senior Member, IEEE Abstract—This paper presents a computer aided design (CAD) framework for verification and reliability analysis of protein-sil- icon hybrid circuits used in biosensors. It is envisioned that similar to integrated circuit (IC) CAD design tools, the proposed frame- work will be useful for system level optimization of biosensors and for discovery of new sensing modalities without resorting to la- borious fabrication and experimental procedures. The framework referred to as FAST analyzes protein-based circuits by solving in- verse problems involving stochastic functional elements that admit non-linear relationships between different circuit variables. In this regard, FAST uses a factor-graph netlist as a user interface and solving the inverse problem entails passing messages/signals be- tween the internal nodes of the netlist. Stochastic analysis tech- niques like density evolution are used to understand the dynamics of the circuit and estimate the reliability of the solution. As an example, we present a complete design flow using FAST for syn- thesis, analysis and verification of our previously reported conduc- tometric immunoassay that uses antibody-based circuits to imple- ment forward error-correction (FEC). Index Terms—Biomolecular circuit, biosensors, computer-aided design, factor-graphs, inverse problems, message passing, simulation. I. INTRODUCTION D ESIGN of reliable biosensors requires understanding, modeling and characterization of fundamental noise, stochastic interactions between proteins and device artifacts. In a theoretical study reported in [1], [2], it was shown that the signals acquired from biosensors could potentially exhibit large variability due to random interactions between biomolecules or due to the noise at the interface of the transducer. While the ef- fects of variability could potentially be alleviated by improving experimental protocols and device fabrication process, in [3] we had proposed using a forward error-correcting (FEC) approach to improve the reliability of biosensors. The FEC biosensor, whose architecture is shown in Fig. 1, uses protein-based reactive circuits (using antibodies, aptamers, or DNA) in con- junction with a transducer which converts the binding of an analyte with the protein into a measurable optical or electrical signal. A biomolecular encoder synthetically introduces re- dundancy into the protein-protein interaction before the signal Manuscript received May 17, 2012; revised August 17, 2012; accepted September 28, 2012. Date of publication December 25, 2012; date of current version July 24, 2013. This work was supported by a research grant from the National Science Foundation (CCF:1117186). This paper was recommended by Associate Editor K. C. Cheung. The authors are with the Department of Electrical and Computer Engi- neering, Michigan State University, East Lansing, MI 48824 USA (e-mail: shantanu@egr.msu.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBCAS.2012.2222403 Fig. 1. System level architecture of an FEC biosensor interface which can be analyzed using FAST. generated by the transducer is read out [3]. Acquisition and de- coding of the sensor signal is performed using silicon (CMOS) circuits and the final result is a score which is proportional to the probability of the target analyte being present in the sample. In [3] we reported examples of protein (antibody) based com- binatorial circuits and in [4] we experimentally demonstrated the feasibility of a small-scale biomolecular encoder using lateral-flow immunoassays. However, these preliminary exper- iments indicated that the full potential of an FEC biosensor can be only be realized using large-scale biomolecular encoders which integrates millions of protein-based circuits. This would require a simulation framework that could be used to model, analyze and predict the reliability of large-scale biomolecular encoders without resorting to laborious and expensive experi- mental procedures. In this paper, we present a computer aided design (CAD) framework called FAST (Factor-graph based Analysis of Stochastic circuiTs), which can be used for design, synthesis and verification of large-scale hybrid protein-silicon circuits. A FAST based design-flow of FEC biosensors is summarized by the chart in Fig. 2. First, simple protein circuits are proto- typed and their responses are experimentally measured. The ex- perimental data is then used for generating a library of behav- ioral models, equivalent circuit models and channel (or noise) models. These models are instantiated and imported when the user specifies a topology comprising of the basic protein-cir- cuits. During analysis, FAST allows additional constraints to be incorporated into the design, which includes limitations on the size of the biosensor substrate, cross-talk due to the sub- strate, analyte propagation and transducer artifacts. FAST then generates a factor-graph netlist that captures the probabilistic interdependencies between different circuit elements. Inference or estimation of probability distributions of target variables are obtained using Monte-Carlo simulations and the dynamics of the factor-graph circuit is understood using a density-evolution analysis. The outcome of the simulations are detection error-rate (DER) curves which can then be used to determine the reliability of the circuit. As a final step in the design-flow, the encoder cir- cuit is fabricated to validate the simulated reliability metrics or some important detection property as predicted by FAST. 1932-4545/$31.00 © 2012 IEEE