Early breast cancer detection using logistic regression models

Abstract

MicroRNAs (miRNAs) play a central role in gene expression and have remarkable abundance in body fluids. They are candidate diagnostics for a variety of conditions and diseases, including breast cancer. Their main objective is to identify miRNAs for the discrimination of cancer and their intrinsic molecular subtypes in order to recognize potential biomarkers.More and more linear algebra and statistics methods are used to address issues in gene expression literature. RNAseq technology is one of the extended use tool for overall analysis of miRNAs expression allowing simultaneus investigation of hundreds or thousands of miRNAs in a sample and is characterized by a low sample size and a large number of characteristics (miRNAs) that impair measures of similarity and classification performance. To avoid the problem of "curse dimensionality" many authors have carried out the selection of characteristics or reduced the size of data matrix. We present new predictive models to classify breast cancer tumor samples in early stage. The methodologies allowed correct classification of early stage breast cancer data set GSE58606 from NCBI with sensibility and specificity greater than 0.95. Also, as a sub-product of the methodology we are able to identify a set of biomarkers already known in others types of cancer