Feature selection methods for the generation of signatures to study the non alcoholic fatty liver disease

Resumen

This thesis focuses in the analysis of microarray expression data, proteomics and machine learning centering on the development of feature subset selection algorithms for biomarker identification and genomic signature generation of genes and pathways to study the Non Alcoholic Fatty Liver Disease. Molecular signatures from microarrays have been used to determine the Non Alcoholic Fatty Liver Disease outcomes, prognostic, mechanisms, targets, and treatment strategies by understanding the processes responsible for the differences between states. To generate the signatures in this study we have adapted and implemented a representative set of 31 feature selection methods making use of adequate evaluation criteria and stable and robust feature selection models. Many of the feature selection methods are adapted for microarray analysis so they generate signatures classifying disease based not only on the activity of individual genes but also on the activity of signaling pathways. One of the goals of these modeling studies is to build classifiers to predict the outcome of future observations. We work towards designing some diagnostic and prognostic models for the Non Alcoholic Fatty Liver Disease. These methods are also used to predict posttranslational phosphorylation modifications in MS/MS spectra.