Otimização e análise de algoritmos de ordenamento de redes proteicas

Abstract

Analysis by Transcriptogram was developed as a solution to noise reduction, usually present in the microarray measuring technique of the Transcriptome, and has demonstrated potential to be applied as a method of disease diagnostics. The noise reduction in the measure is achived by the protein interaction network ordering, allowing gene expression analysis in whole genome scale. The Transcriptogram's efficiency to noise reduction was analyzed, however, it still lacks an analisys of the ordering quality, so that the best parameter setting for the ordering algorithm is used by the Transcriptogram. So far, this analysis is hindered by the high runtime of the ordering algorithm. In this work, an analysis of the ordering algorithm stages allows some optimizations, and consequent reduction in execution time, also allowing further analysis on which parameters settings have the greatest influence on the ordering quality. Applying the Transcriptogram to a diagnostic problem, the diagnostic measure is used to characterize the influence of the parameters of the ordering algorithm to achive better diagnoses. The results show that the protein network used in previous works doesn't produce the best diagnostics. Moreover, the ordering minimization, achieved by executing the ordering algorithm for longer periods, does not necessarily increase the probability to find better diagnosis compared to random ordering. Eventhough the experimental diagnostic results could not statistically difFerentiate random ordering from optimized ordering, these results cannot be considered conclusive since a single disease has been evaluated.