Acoustic-wave biosensors work by detecting the frequency shifts resulting from the binding of target molecules to a functionalized resonator. We have developed a prototype membrane-based acoustic wave sensor fabricated as a MicroElectroMechanical Systems (MEMS) device [1]. With this device, we can monitor not only the fundamental frequency of vibration, but up to six higher modes as well. Coupled with the ability to chemically pattern the membrane surface to allow spatial grafting of different target molecules to different regions of the membrane, and the ability to accurately model frequency responses to distributed mass loadings [2], this gives us the ability to detect multiple species simultaneously, or to incorporate internal redundancy in measurements.

In our design, a single chip has eight independent membrane pairs, and each membrane pair functions as an individual sensor capable of sensing multiple targets through a differential measurement. We have previously developed a computationally efficient reduced order model which accurately captures the response surface under distributed mass loading [2]. Using this model, based on a matrix perturbation analysis, we will demonstrate how to use optimization schema for both sensitivity and robustness, at two discrete levels, as measured by the Receiver Operator Characteristic curve [3]. The first level of optimization is the single sensor with multiple functionalized regions; the second is the single chip with an array of sensors. In each case, we want to determine the optimal size, placement, and number of functionalized regions to maximize sensor performance. We will develop objective functions and constraints for both performance metrics at each level of optimization, such as maximizing discrimination in underdetermined systems [4], and maximizing sensitivity to multiple targets on a single membrane. At the single sensor level our work will build on our previous results on maximizing discrimination [4], giving us a complete set of optimization schemes for a single membrane.

To demonstrate the applicability of our results, we will perform validation studies using literature data on detection of airborne biological weapons, including Bacillus anthracis and Botulinum toxin [5], and compare our results to the naive approach wherein each individual sensor is used for the detection of a single target analyte.

[1] M.J. Bartkovsky, A. Liao, S. Hauan, T.M. Przybycien, and G.K. Fedder. "A MEMS based Acoustic-Wave Gravimetric Biosensor: Device Characterization and Performance." Presented at the AIChE annual meeting, Cincinnati, OH, USA, November 2005. Paper 423e.

[2] J.E. Valentine, T.M. Przybycien, and S. Hauan. "Design of Acoutic Wave Biochemical Sensors using MEMS." Submitted to Sensors and Actuators A, May 2006.

[3] M.H. Zweig and G. Campbell. "Receiver-operating characteristic plots: A fundmantal evaluation tool in clinical medicine." Clin. Chem. 1993, v39 i4.

[4] J.E. Valentine, T.M. Przybycien, and S. Hauan. "Quantitative Design Approach for a multi-analyte acoustic-wave sensor" Presented at the AIChE annual meeting, Cincinnati, OH, USA, November 2005. Paper 429b.

[5] M.G. Kortepeter and G.W. Parker. "Potential biological weapons threats." Emerging Infectious Diseases. 1999, v5 i4.