Machine Learning methods

Most popular Machine Learning techniques, such as neural networks or Support Vector Machine (SVM), made it possible to successfully address issues of overriding interest in Bioinformatics field, such as automatic disease diagnosis and the recognition of relevant signals in genomics sequences, the classification of gene expression profiles. The models generated by these methods achieve high accuracy, but do not understand the link between inputs and outputs, thus preventing a deeper analysis of the problem under consideration.

To overcome this difficulty are emerging new Machine Learning techniques, known as methods for extracting rules, which produce models consisting in a set of intelligible rules that describe a physical interest. In this way the researchers in the biomedical field can make use of information extracted from the data collected to infer new knowledge about the problem that is intended to be analyzed.

The generality of Machine Learning techniques allows them to be widely used in the processes of data mining, in order to produce a structured representation of the data available, both in the treatment and classification of experiments conducted at experimental home. For example in the gene expression data produced by experiments with DNA microarray, the analysis of images is a fundamental aspect and the amount of data to be treated requires the use of automatic procedures for the location of the spot, and the measurement of assessing reliability. Both these phases and the next classification of gene expression values obtained, and in determining the most relevant genes for a physio-pathological state interest, methods of Machine Learning constitute an irreplaceable technical investigation. In this area, researchers are present in several national projects (LITBIO) and international BIOPATTERN, BIOIFOGRID.