Analysis of biomolecular networks through parametrized Hopfield Networks

Several node label ranking problems in Systems Biology ara characterized by a large unbalance between annotated and unannotated nodes: for instance in gene function prediction usually only a small subset of the genes are annotated for a specific GO term. In this context both supervised and semi-supervised algorithms usually fail to detect genes "positive" for a given class.

To deal with this problem we developed COSNet (COst Sensitive neural Network), a novel cost-sensitive family of parametrized Hopfield networks: by considering the values of the neuron states as parameters to be learned from the data, we designed a graph-based semi-supervised learning algorithm able to learn from imbalanced data, even when only a small subset of the nodes are labelled (Frasca and Valentini, 2016, Frasca et al. 2013, Bertoni et al. 2011).

A regularized version of COSNEt has also been proposed to deal with extremely unbalanced functional classes (Frasca et al. 2013). Variants of COSNet exploit gene multi-functionality to improve gene function predictions (Frasca 2015), and a new multi-parametric Hopfield network (Hopfield multi-category -- HoMCat), designed to take into account a priori-known "categories" of neurons within networks, has been successfully applied to the multi-species protein function prediction problem (Frasca et al. 2015).

We developed also UNIPred that combines unbalance-aware parametrized Hopfield networks with unbalance-aware network integration. This approach also introduces a valuable way to evaluate the informativeness of different sources of information for the prediction of specific GO terms (Frasca et al. 2015).

Recently we developed Multi-task Hopfield networks, able to predict multiple tasks simultaneously, by leveraging the synergy resulting from learning multiple related tasks, to improve the overall accuracy of the learning system (Frasca et al., 2019). This method can be applied to simultaneously learn multiple GO terms, as well as multiple associations between genes and multiple related diseases, with interesting perspectives for the prediction of protein functions or for the prediction of the molecular mechanisms underlying comorbidity.

We also developed methods and web services for the combination and visualization of complex biomolecular networks (Perlasca et a. 2016, 2019).

Publications

P. Perlasca, M. Frasca, C. Ba, M. Notaro, A. Petrini, E. Casiraghi, G. Grossi, J. Gliozzo, G. Valentini and M. Mesiti. UNIPred-Web: a web tool for the integration and visualization of biomolecular networks for protein function prediction. BMC Bioinformatics, BioMed Central 20(1), 2019.

M. Frasca, G. Grossi and G. Valentini. Multitask Hopfield Networks. European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, , Wurzburg (Germany), 2019.

M. Frasca and G. Valentini. COSNet: An R package for label prediction in unbalanced biological networks. Neurocomputing, Elsevier 237, 2017.

P. Perlasca, G. Valentini, M. Frasca and M. Mesiti. Multi-species protein function prediction: towards web-based visual analytics.. iiWAS, 2016.

M. Frasca, A. Bertoni and G. Valentini. UNIPred: Unbalance-aware Network Integration and Prediction of protein functions. Journal of Computational Biology, Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd Floor New Rochelle, NY 10801 USA 22(12), 2015.

M. Frasca. Automated gene function prediction through gene multifunctionality in biological networks. Neurocomputing, Elsevier 162, 2015.

M. Frasca, A. Bertoni, M. Re and G. Valentini. A neural network algorithm for semi-supervised node label learning from unbalanced data. Neural Networks, Elsevier 43, 2013.

M. Frasca, A. Bertoni and G. Valentini. An unbalance-aware network integration method for gene function prediction. MLSB 2013-Machine Learning for Systems Biology, 2013.

A. Bertoni, M. Frasca and G. Valentini. COSNet: a cost sensitive neural network for semi-supervised learning in graphs. Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 2011.

M. Frasca, S. Bassis and G. Valentini. Learning node labels with multi-category hopfield networks. Neural Computing and Applications, Springer 27(6), 2016.