Uma extensão do algoritmo Fuzzy C-Means aplicado ao reconhecimento de spots em imagens oriundas de Eletroforese Bidimensional

Abstract

Proteomics is a methodology widely used, especially by molecular biology, which main goals consist of documenting the general distribution of proteins in a cell, identifying and characterizing individual proteins of interest and, mainly, elucidating associations and functions of proteins. Currently, one of the main techniques used in proteomics is Two-dimensional Electrophoresis (2D), which is based on separation and migration of molecules, placed in a solution under the influence of the application of an electric field. These proteins can be detected by a variety of revelation reagents, revealing a two-dimensional profile of spots. At the end of the process, the gel is scanned and the resulting image can be processed. Two-dimensional electrophoresis images often contain noise, such as dust particles, and even cracks in the gel, and this can interfere in the the analysis results. In this context, this work explores the application of fuzzy logic in image processing, and provides a model in order to find spots in Two-dimensional electrophoresis images. The proposed model uses fuzzy clustering and it only makes use of the pixels information to do the classification of the pixels and find the spots. The evaluation of the proposed model was made using two-dimensional electrophoresis images of SWISS-2DPAGE database. These images contain spots already checked in the laboratory. We used 29 images in the model tests, with the best result of recall reaching 93,30% of the spots found. Also, comparisons were made of the proposed model, of unsupervised learning, with supervised learning techniques (machine learning) in development in the research group.