(370e) Modeling and Optimization of Biosensors Exploiting Restriction Fragment Length Polymorphism

Restriction fragment length polymorphism is a widely used method for detecting the presence complex prokaryotes. Microfluidic technology holds promise for extending these methods to develop a new class of bacterial sensors for low cost environment and medical diagnostics. Although large databases of fragment lengths are now available from experimental data, bacterial organism identification is still limited by the number of distinct fragment lengths and the experimental resolution of fragment lengths. This paper develops and optimizes methodology for sensors based on restriction fragment length polymorphism. Maximum entropy and optimization randomized detector design are applied to estimate the microbial communities' present probability distribution and fraction distribution. Maximum information gain is also used to optimize the experimental protocols. Two typical applications are provided to illustrate the practical application of our analysis to biosensor design. The shortcomings of culture-based methods are well known and Polymerase Chain Reaction (PCR) based, culture-independent techniques provide new ways for microbial communities' analysis. Although terminal restriction fragment analysis is currently one of the most important ones, the fragment length analysis processed by Gel electrophoresis is time-consuming that limits its utility. Recent reports show that a rapid molecular mapping technology is possible for DNA fragment length analysis by using microfluidic stretching and single-molecule detection. With the application of microfluidic technology, a new rapid, low-cost biosensor is possible for environment and medical diagnostics. In this paper two different mathematic models are used to describe the new sensor based on restriction fragment length polymorphism that results in two optimization problems. Based on limited measurements of distinct fragment lengths polymorphism following digestion with restriction enzymes, one is estimating the present probability distribution for a microbial community and the other one is estimating the fraction distribution for a microbial community. The synthesis problem is then to detect maximum by optimizing experimental protocols and the selection of restriction enzymes for a given bacteria library. Here we propose maximum entropy, a randomized detector design, relative information (information gain) as the appropriate metric for the analysis. Several different information entropy and relative information are compared. Our computational study indicates that maximum entropy is an appropriate method for the computation of bacterial fraction distribution and optimization randomized detector design is a suitable method for the estimating bacterial present distribution. Both of them are solvable convex problems with well known solutions. Information gain establishes an upper bound for the sensor performance. Maximum information gain yields a convex geometric program. Two typical applications from references are also used to illustrate the practical application of our analysis to biosensor design.