Bayesian Optimization Algorithm for the Non-unique Oligonucleotide Probe Selection Problem

DNA microarrays are used in order to recognize the presence or absence of different biological components (targets) in a sample. Therefore, the design of the microarrays which includes selecting short Oligonucleotide sequences (probes) to be affixed on the surface of the microarray becomes a major issue. This paper focuses on the problem of computing the minimal set of probes which is able to identify each target of a sample, referred to as Non-unique Oligonucleotide Probe Selection. We present the application of an Estimation of Distribution Algorithm (EDA) named Bayesian Optimization Algorithm (BOA) to this problem, for the first time. The presented approach considers integration of BOA and state-of-the-art heuristics introduced for the non-unique probe selection problem. This approach provides results that compare favorably with the state-of-the-art methods. It is also able to provide biologists with more information about the dependencies between the probe sequences of each dataset.