Collaboration-Based Function Prediction in Protein-Protein Interaction Networks

Abstract

Molecular biology and all the biomedical sciences are undergoing a true revolution as a result of the emergence and growing impact of a series of new disciplines and tools sharing the '-omics' suffix in their name. These include in particular genomics, transcriptomics, proteomics and metabolomics, devoted respectively to the examination of the entire systems of genes, transcripts, proteins and metabolites present in a given cell or tissue type. The availability of these new, highly effective tools for biological exploration is dramatically changing the way one performs research in at least two respects. First, the amount of available experimental data is not a limiting factor any more; on the contrary, there is a plethora of it. Given the research question, the challenge has shifted towards identifying the relevant pieces of information and making sense out of it (a 'data mining' issue). Second, rather than focus on components in isolation, we can now try to understand how biological systems behave as a result of the integration and interaction between the individual components that one can now monitor simultaneously, so called 'systems biology'. Machine learning naturally appears as one of the main drivers of progress in this context, where most of the targets of interest deal with complex structured objects: sequences, 2D and 3D structures or interaction networks. At the same time bioinformatics and systems biology have already induced significant new developments of general interest in machine learning, for example in the context of learning with structured data, graph inference, semi-supervised learning, system identification, and novel combinations of optimization and learning algorithms. This book contains the scientific contributions presented at the Fourth International Workshop on Machine Learning in Systems Biology (MLSB'2010), held in Edinburgh from October 15 to 16, 2010. The workshop was organized as a core event of the PASCAL2 Network of Excellence, under the IST programme of European Union. The aim of the workshop was to contribute to the cross-fertilization between the research in machine learning methods and their applications to systems biology (i.e., complex biological and medical questions) by bringing together method developers and experimentalists. The technical program of the workshop consisted of invited lectures, oral presentations and poster presentations. Invited lectures were given by Nir Friedman,Ursula Kummer, Hans Lehrach, Florence d'Alché-Buc, and Vebjorn Ljosa. Sixteen oral presentations were given, for which extended abstracts are included in this book: these were selected from 24 submissions, each reviewed by three members of the scientific program committee. Twenty-five poster presentations were given, for which abstracts of varying length are included here. We would like to thank all the people contributing to the technical programme, the scientific program committee, the local organizers and the sponsors for making the workshop possible.

Predicting Genes Involved in Human Cancer Using Network Contextual Information

“…One argument is that GO annotations are often incomplete, and by collecting GO information from the neighbors of a protein p, we may get more information about p itself. This argument is backed up by the fact that GO annotations of proteins can often be predicted well from the GO annotations of their neighbors; see, e.g., [4,27]. However, as we will show, this is not the only effect; there is also a direct relationship between a protein's involvement in cancer and the GO annotations of the proteins it interacts with.…”

“…Since it was already known that the functional annotation of a protein's neighbors can be used to predict the protein's own functions [4,27], and that the protein's own functions are relevant for its involvement in cancer [25,26], one might wonder to what extent our results are simply a consequence of these two facts. We can test this by enriching proteins in the PPI network with predicted GO annotations (predicted from the GO annotations of their neighbors), and next applying the FS method.…”

Using a Human Disease Network for augmenting prior knowledge about diseases

“…Selecting individual discriminative functions based on χ 2 (f i ) does not consider the network topology and the way different functions interact with each other in the network. Rahmani et al [38] showed that for the task of predicting cancer-related proteins, it is possible that a function f i does not correlate itself with cancer-involvement, but interaction of the same function with some function f j does correlate with the former protein being involved in a cancer. Rahmani et al [38] proposed a new way of calculating the χ 2 of the function pairs in the PPI network.…”

DeepFunc: A Deep Learning Framework for Accurate Prediction of Protein Functions from Protein Sequences and Interactions

“…This information may include secondary structures extracted from input protein sequences, or secondary structure, disordered regions, signal peptides, and motifs like in the case of the FFPred3 method . Finally, several approaches rely on the PPI‐derived information to accurately predict protein functions . The crucial idea behind these methods is that proteins which share similar topological features in the PPI networks may share similar functions .…”

“…Moreover, some protein function predictors utilize other types of data, such as genetic interactions, genomic context, protein structure, and gene expression . We focus on two classes of current predictors: sequence‐based methods that cover the use of domains, motifs and residue‐level information, and PPI‐based methods that rely on information extracted from these networks . These two classes of methods utilize somehow complementary information.…”