NetworkAnalyst, another tool for pathway reconstruction

[ntalluri]

This is another piece of work we should explore that is related to frameworks similar to SPRAS: https://pmc.ncbi.nlm.nih.gov/articles/PMC4086107/

Abstract:

Biological network analysis is a powerful approach to gain systems-level understanding of patterns of gene expression in different cell types, disease states and other biological/experimental conditions. Three consecutive steps are required - identification of genes or proteins of interest, network construction and network analysis and visualization. To date, researchers have to learn to use a combination of several tools to accomplish this task. In addition, interactive visualization of large networks has been primarily restricted to locally installed programs. To address these challenges, we have developed NetworkAnalyst, taking advantage of state-of-the-art web technologies, to enable high performance network analysis with rich user experience. NetworkAnalyst integrates all three steps and presents the results via a powerful online network visualization framework. Users can upload gene or protein lists, single or multiple gene expression datasets to perform comprehensive gene annotation and differential expression analysis. Significant genes are mapped to our manually curated protein-protein interaction database to construct relevant networks. The results are presented through standard web browsers for network analysis and interactive exploration. NetworkAnalyst supports common functions for network topology and module analyses. Users can easily search, zoom and highlight nodes or modules, as well as perform functional enrichment analysis on these selections. The networks can be customized with different layouts, colors or node sizes, and exported as PNG, PDF or GraphML files. Comprehensive FAQs, tutorials and context-based tips and instructions are provided. NetworkAnalyst currently supports protein-protein interaction network analysis for human and mouse and is freely available at http://www.networkanalyst.ca.

[agitter]

There are relevant ideas here. Is this their "pathway reconstruction algorithm" though?

For each individual (seed) protein, a search algorithm is performed to identify proteins that directly interact with the seed proteins (first-order interactors). The results are used to build the default networks. This approach will typically return one large subnetwork (‘continent’) with several smaller ones (‘islands’). Most subsequent analyses are performed on the continent. We recommend users to control the number of nodes to be in the range of 200 - 2000 for practical (visual, biological and computational) reasons since larger networks will lead to a ‘hairball’ effect, which rarely produces any informative outcome, while smaller networks will not enable systems-level understanding.

Several functions have been implemented to allow users to further adjust the network size. For instance, if the default network constructed with first-order interactors is small (i.e. <100 nodes), users can search for higher-order interactions by treating all nodes in this network as new seed proteins. Users can also trim a large network to be composed only of nodes that connect the original seed proteins. When there are too many seed proteins (i.e. >1000) and the default network will be too large to be visualized, users can choose to focus only on networks within these seed proteins (zero-order interactors). In addition, NetworkAnalyst allows users to adjust the cutoffs for P values or fold changes to control the number of seed proteins, as well as to manually exclude certain proteins that are known to be involved in many non-specific interactions, to enable fine control of network size.

It sounds like they take the direct neighbors by default and have options to include second order neighbors if that is too small.

Version 3.0 here: https://doi.org/10.1093/nar/gkz240

Related R package here: https://github.com/xia-lab/NetworkAnalystR