Gene Expression Analysis using Markov Chains extracted from Recurrent Neural Networks

Igor Lorenzato AlmeidaDenise Regina PechmannMaicon de Brito do AmaranteAdelmo Luis Cechin

This paper presents a new approach for the analysis of microarray data by the use of Recurrent Neural Networks (RNNs) as a time model of the gene regulatory network. Our method extracts a Markov Chain (MC) from a trained RNN and the relations among genes in each MC state. We propose to use the learning ability of RNNs for the automatic construction of the model with the gene interactions represented by the weights and afterwards to use an algorithm to extract these relations in the form of MCs and linear matrices easily visualized in the form of graphs of states and genes. The graph of states show the evolution of the gene expression levels in time while the gene graph shows the dependencies among genes in each Markov state.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: