| Summary: | <p>Abstract</p> <p>Background</p> <p>Inference of gene regulatory networks is a key goal in the quest for understanding fundamental cellular processes and revealing underlying relations among genes. With the availability of gene expression data, computational methods aiming at regulatory networks reconstruction are facing challenges posed by the data's high dimensionality, temporal dynamics or measurement noise. We propose an approach based on a novel multi-layer evolutionary trained neuro-fuzzy recurrent network (ENFRN) that is able to select potential regulators of target genes and describe their regulation type.</p> <p>Results</p> <p>The recurrent, self-organizing structure and evolutionary training of our network yield an optimized pool of regulatory relations, while its fuzzy nature avoids noise-related problems. Furthermore, we are able to assign scores for each regulation, highlighting the confidence in the retrieved relations. The approach was tested by applying it to several benchmark datasets of yeast, managing to acquire biologically validated relations among genes.</p> <p>Conclusions</p> <p>The results demonstrate the effectiveness of the ENFRN in retrieving biologically valid regulatory relations and providing meaningful insights for better understanding the dynamics of gene regulatory networks.</p> <p>The algorithms and methods described in this paper have been implemented in a Matlab toolbox and are available from: <url>http://bioserver-1.bioacademy.gr/DataRepository/Project_ENFRN_GRN/</url>.</p>
|
|---|
