Microarray data and image simulation

Microarray technologies have become very popular among biologists because a single experiment can produce a large amount of gene expression data. However, the experimental process, from slide manufacturing to data analysis, can be affected by random errors and/or systemic bias caused by several factors. Moreover, an evaluation of the data analysis tools can not be objectively performed because of the lack of specific benchmarks. The paper describes a method of data and microarray image simulation that provides researchers with reference tests for evaluating the performance of analysis tools. The relevant characteristics of a specific experiment are used for generating synthetic microarray images and the corresponding gene expression values. Hence, one can estimate expected errors comparing simulated data to the results of different analysis packages and then can choose the most suitable technique for the specific experiment. Three different segmentation algorithms have been tested on simulated images. The results on the classification correctness of gene activation are reported.