Optimal estimation in DNA microarrays via global optimization

DNA microarray technology relies on affinity between complementary nucleic acids to detect the presence and estimate the amounts of target molecules of interest. Molecular binding is stochastic in nature, having inherent uncertainty manifested as Poisson noise. This, along with interference due to non-specific binding, are among the main obstacles for achieving high accuracy of DNA microarrays. In this paper, optimal target estimation in DNA microarrays is shown to lead to a fractional program. A practically feasible approach which employs a branch-and-bound algorithm solving a convex optimization problem in each step is presented. Simulation results demonstrate that the proposed approach outperforms previously considered methods.