Automatic detection of bands in electrophoresis gel images by means of optimization of a target function

Biocontrols are one of the mandatory factors to be considered in the search for sustainable agricultural activities. The amount of samples to be analyzed while searching for such agents makes unavoidable the use of technological tools that accelerate the processes involved and improve the accuracy of the results. These results are obtained using different techniques, among them the gel electrophoresis, which is widely used to analyze the molecular separation of substances like proteins and nucleic acids. The outcomes given by this technique are based on the final location of bands formed by the molecule segments in the analyzed sample once the process is finished. This work has as main objective to automatically find the location of each of these bands by means using as input gel electrophoresis images. The proposed solution is based in the optimization of a target function using the Downhill Simplex method and genetic algorithms to guide the generation of the multidimensional points to be optimized. The target function is the mean square error between the intensity distribution of a lane in the image and the sum of gaussian functions that model the distribution of DNA/RNA or proteins along the lane. Hence, the intensity level and the central band locations are obtained by finding the set of parameters that best fit the sum of gaussian functions to the lane image.