A Graph Algorithmic Approach to Separate Direct from Indirect Neural Interactions

Network graphs have become a popular tool to represent complex systems composed of many interacting subunits; especially in neuroscience, network graphs are increasingly used to represent and analyze functional interactions between multiple neural sources. Interactions are often reconstructed using pairwise bivariate analyses, overlooking the multivariate nature of interactions: it is neglected that investigating the effect of one source on a target necessitates to take all other sources as potential nuisance variables into account; also combinations of sources may act jointly on a given target. Bivariate analyses produce networks that may contain spurious interactions, which reduce the interpretability of the network and its graph metrics. A truly multivariate reconstruction, however, is computationally intractable because of the combinatorial explosion in the number of potential interactions. Thus, we have to resort to approximative methods to handle the intractability of multivariate interaction reconstruction, and thereby enable the use of networks in neuroscience. Here, we suggest such an approximative approach in the form of an algorithm that extends fast bivariate interaction reconstruction by identifying potentially spurious interactions post-hoc: the algorithm uses interaction delays reconstructed for directed bivariate interactions to tag potentially spurious edges on the basis of their timing signatures in the context of the surrounding network. Such tagged interactions may then be pruned, which produces a statistically conservative network approximation that is guaranteed to contain non-spurious interactions only. We describe the algorithm and present a reference implementation in MATLAB to test the algorithm’s performance on simulated networks as well as networks derived from magnetoencephalographic data. We discuss the algorithm in relation to other approximative multivariate methods and highlight suitable application scenarios. Our approach is a tractable and data-efficient way of reconstructing approximative networks of multivariate interactions. It is preferable if available data are limited or if fully multivariate approaches are computationally infeasible.     

Functional network organization of the human brain

Real-world complex systems may be mathematically modeled as graphs, revealing properties of the system. Here we study graphs of functional brain organization in healthy adults using resting state functional connectivity MRI. We propose two novel brain-wide graphs, one of 264 putative functional areas, the other a modification of voxelwise networks that eliminates potentially artificial short-distance relationships. These graphs contain many subgraphs in good agreement with known functional brain systems. Other subgraphs lack established functional identities; we suggest possible functional characteristics for these subgraphs. Further, graph measures of the areal network indicate that the default mode subgraph shares network properties with sensory and motor subgraphs: it is internally integrated but isolated from other subgraphs, much like a "processing" system. The modified voxelwise graph also reveals spatial motifs in the patterning of systems across the cortex.     

Density-dependence of functional development in spiking cortical networks grown in vitro

During development, the mammalian brain differentiates into specialized regions with distinct functional abilities. While many factors contribute to functional specialization, we explore the effect of neuronal density on the development of neuronal interactions in vitro. Two types of cortical networks, namely, dense and sparse with 50,000 and 12,500 total cells, respectively, are studied. Activation graphs that represent pairwise neuronal interactions are constructed using a competitive first response model. These graphs reveal that, during development in vitro, dense networks form activation connections earlier than sparse networks. Link entropy analysis of dense network activation graphs suggests that the majority of connections between electrodes are reciprocal in nature. Information theoretic measures reveal that early functional information interactions (among three electrodes) are synergetic in both dense and sparse networks. However, during later stages of development, previously synergetic relationships become primarily redundant in dense, but not in sparse networks. Large link entropy values in the activation graph are related to the domination of redundant ensembles in late stages of development in dense networks. Results demonstrate differences between dense and sparse networks in terms of informational groups, pairwise relationships, and activation graphs. These differences suggest that variations in cell density may result in different functional specializations of nervous system tissue in vivo.     

Laplacian Estrada and Normalized Laplacian Estrada Indices of Evolving Graphs

Large-scale time-evolving networks have been generated by many natural and technological applications, posing challenges for computation and modeling. Thus, it is of theoretical and practical significance to probe mathematical tools tailored for evolving networks. In this paper, on top of the dynamic Estrada index, we study the dynamic Laplacian Estrada index and the dynamic normalized Laplacian Estrada index of evolving graphs. Using linear algebra techniques, we established general upper and lower bounds for these graph-spectrum-based invariants through a couple of intuitive graph-theoretic measures, including the number of vertices or edges. Synthetic random evolving small-world networks are employed to show the relevance of the proposed dynamic Estrada indices. It is found that neither the static snapshot graphs nor the aggregated graph can approximate the evolving graph itself, indicating the fundamental difference between the static and dynamic Estrada indices.     

Algorithms for effective querying of compound graph-based pathway databases

Background  

Graph-based pathway ontologies and databases are widely used to represent data about cellular processes. This representation makes it possible to programmatically integrate cellular networks and to investigate them using the well-understood concepts of graph theory in order to predict their structural and dynamic properties. An extension of this graph representation, namely hierarchically structured or compound graphs, in which a member of a biological network may recursively contain a sub-network of a somehow logically similar group of biological objects, provides many additional benefits for analysis of biological pathways, including reduction of complexity by decomposition into distinct components or modules. In this regard, it is essential to effectively query such integrated large compound networks to extract the sub-networks of interest with the help of efficient algorithms and software tools.     

The large graph limit of a stochastic epidemic model on a dynamic multilayer network

We consider a Markovian SIR-type (Susceptible → Infected → Recovered) stochastic epidemic process with multiple modes of transmission on a contact network. The network is given by a random graph following a multilayer configuration model where edges in different layers correspond to potentially infectious contacts of different types. We assume that the graph structure evolves in response to the epidemic via activation or deactivation of edges of infectious nodes. We derive a large graph limit theorem that gives a system of ordinary differential equations (ODEs) describing the evolution of quantities of interest, such as the proportions of infected and susceptible vertices, as the number of nodes tends to infinity. Analysis of the limiting system elucidates how the coupling of edge activation and deactivation to infection status affects disease dynamics, as illustrated by a two-layer network example with edge types corresponding to community and healthcare contacts. Our theorem extends some earlier results describing the deterministic limit of stochastic SIR processes on static, single-layer configuration model graphs. We also describe precisely the conditions for equivalence between our limiting ODEs and the systems obtained via pair approximation, which are widely used in the epidemiological and ecological literature to approximate disease dynamics on networks. The flexible modeling framework and asymptotic results have potential application to many disease settings including Ebola dynamics in West Africa, which was the original motivation for this study.     

Graph sharpening plus graph integration: a synergy that improves protein functional classification

MOTIVATION: Predicting protein function is a central problem in bioinformatics, and many approaches use partially or fully automated methods based on various combination of sequence, structure and other information on proteins or genes. Such information establishes relationships between proteins that can be modelled most naturally as edges in graphs. A priori, however, it is often unclear which edges from which graph may contribute most to accurate predictions. For that reason, one established strategy is to integrate all available sources, or graphs as in graph integration, in the hope that the positive signals will add to each other. However, in the problem of functional prediction, noise, i.e. the presence of inaccurate or false edges, can still be large enough that integration alone has little effect on prediction accuracy. In order to reduce noise levels and to improve integration efficiency, we present here a recent method in graph-based learning, graph sharpening, which provides a theoretically firm yet intuitive and practical approach for disconnecting undesirable edges from protein similarity graphs. This approach has several attractive features: it is quick, scalable in the number of proteins, robust with respect to errors and tolerant of very diverse types of protein similarity measures. RESULTS: We tested the classification accuracy in a test set of 599 proteins with remote sequence homology spread over 20 Gene Ontology (GO) functional classes. When compared to integration alone, graph sharpening plus integration of four vastly different molecular similarity measures improved the overall classification by nearly 30% [0.17 average increase in the area under the ROC curve (AUC)]. Moreover, and partially through the increased sparsity of the graphs induced by sharpening, this gain in accuracy came at negligible computational cost: sharpening and integration took on average 4.66 (+/-4.44) CPU seconds. AVAILABILITY: Software and Supplementary data will be available on http://mammoth.bcm.tmc.edu/     

Driving developmental and evolutionary change: A systems biology view

Embryonic development is underpinned by ～50 core processes that drive morphogenesis, growth, patterning and differentiation, and each is the functional output of a complex molecular network. Processes are thus the natural and parsimonious link between genotype and phenotype and the obvious focus for any discussion of biological change. Here, the implications of this approach are explored. One is that many features of developmental change can be modeled as mathematical graphs, or sets of connected triplets of the general form <noun><verb><noun>. In these, the verbs (edges) are the outputs of the processes that drive change and the nouns (nodes) are the time-dependent states of biological entities (from molecules to tissues). Such graphs help unpick the multi-level complexity of developmental phenomena and may help suggest new experiments. Another comes from analyzing the effect of mutation that lead to tinkering with the dynamic properties of these processes and to congenital abnormalities; if these changes are both inherited and advantageous, they become evolutionary modifications. In this context, protein networks often represents what classical evolutionary genetics sees as genes, and the realization that traits reflect the output processes of complex networks, particularly for growth, patterning and pigmentation, rather than anything simpler clarifies some problems that the evolutionary synthesis of the 1950s has found hard to solve. In the wider context, most processes are used many times in development and cooperate to produce tissue modules (bones, branching duct systems, muscles etc.). Their underlying generative networks can thus be thought of as genomic modules or subroutines.     

An Adaptive Complex Network Model for Brain Functional Networks

Brain functional networks are graph representations of activity in the brain, where the vertices represent anatomical regions and the edges their functional connectivity. These networks present a robust small world topological structure, characterized by highly integrated modules connected sparsely by long range links. Recent studies showed that other topological properties such as the degree distribution and the presence (or absence) of a hierarchical structure are not robust, and show different intriguing behaviors. In order to understand the basic ingredients necessary for the emergence of these complex network structures we present an adaptive complex network model for human brain functional networks. The microscopic units of the model are dynamical nodes that represent active regions of the brain, whose interaction gives rise to complex network structures. The links between the nodes are chosen following an adaptive algorithm that establishes connections between dynamical elements with similar internal states. We show that the model is able to describe topological characteristics of human brain networks obtained from functional magnetic resonance imaging studies. In particular, when the dynamical rules of the model allow for integrated processing over the entire network scale-free non-hierarchical networks with well defined communities emerge. On the other hand, when the dynamical rules restrict the information to a local neighborhood, communities cluster together into larger ones, giving rise to a hierarchical structure, with a truncated power law degree distribution.     

Analyzing protein lists with large networks: edge-count probabilities in random graphs with given expected degrees.

We present an analytical framework to analyze lists of proteins with large undirected graphs representing their known functional relationships. We consider edge-count variables such as the number of interactions between a protein and a list, the size of a subgraph induced by a list, and the number of interactions bridging two lists. We derive approximate analytical expressions for the probability distributions of these variables in a model of a random graph with given expected degrees. Probabilities obtained with the analytical expressions are used to mine a protein interaction network for functional modules, characterize the connectedness of protein functional categories, and measure the strength of relations between modules.     

In search of the biological significance of modular structures in protein networks

Many complex networks such as computer and social networks exhibit modular structures, where links between nodes are much denser within modules than between modules. It is widely believed that cellular networks are also modular, reflecting the relative independence and coherence of different functional units in a cell. While many authors have claimed that observations from the yeast protein–protein interaction (PPI) network support the above hypothesis, the observed structural modularity may be an artifact because the current PPI data include interactions inferred from protein complexes through approaches that create modules (e.g., assigning pairwise interactions among all proteins in a complex). Here we analyze the yeast PPI network including protein complexes (PIC network) and excluding complexes (PEC network). We find that both PIC and PEC networks show a significantly greater structural modularity than that of randomly rewired networks. Nonetheless, there is little evidence that the structural modules correspond to functional units, particularly in the PEC network. More disturbingly, there is no evolutionary conservation among yeast, fly, and nematode modules at either the whole-module or protein-pair level. Neither is there a correlation between the evolutionary or phylogenetic conservation of a protein and the extent of its participation in various modules. Using computer simulation, we demonstrate that a higher-than-expected modularity can arise during network growth through a simple model of gene duplication, without natural selection for modularity. Taken together, our results suggest the intriguing possibility that the structural modules in the PPI network originated as an evolutionary byproduct without biological significance.     

SpectralNET – an application for spectral graph analysis and visualization

Background  

Graph theory provides a computational framework for modeling a variety of datasets including those emerging from genomics, proteomics, and chemical genetics. Networks of genes, proteins, small molecules, or other objects of study can be represented as graphs of nodes (vertices) and interactions (edges) that can carry different weights. SpectralNET is a flexible application for analyzing and visualizing these biological and chemical networks.     

Protein complex prediction based on maximum matching with domain-domain interaction

With the development of high-throughput methods for identifying protein-protein interactions, large scale interaction networks are available. Computational methods to analyze the networks to detect functional modules as protein complexes are becoming more important. However, most of the existing methods only make use of the protein-protein interaction networks without considering the structural limitations of proteins to bind together. In this paper, we design a new protein complex prediction method by extending the idea of using domain-domain interaction information. Here we formulate the problem into a maximum matching problem (which can be solved in polynomial time) instead of the binary integer linear programming approach (which can be NP-hard in the worst case). We also add a step to predict domain-domain interactions which first searches the database Pfam using the hidden Markov model and then predicts the domain-domain interactions based on the database DOMINE and InterDom which contain confirmed DDIs. By adding the domain-domain interaction prediction step, we have more edges in the DDI graph and the recall value is increased significantly (at least doubled) comparing with the method of Ozawa et al. (2010) [1] while the average precision value is slightly better. We also combine our method with three other existing methods, such as COACH, MCL and MCODE. Experiments show that the precision of the combined method is improved. This article is part of a Special Issue entitled: Computational Methods for Protein Interaction and Structural Prediction.     

Quantifying loopy network architectures

Biology presents many examples of planar distribution and structural networks having dense sets of closed loops. An archetype of this form of network organization is the vasculature of dicotyledonous leaves, which showcases a hierarchically-nested architecture containing closed loops at many different levels. Although a number of approaches have been proposed to measure aspects of the structure of such networks, a robust metric to quantify their hierarchical organization is still lacking. We present an algorithmic framework, the hierarchical loop decomposition, that allows mapping loopy networks to binary trees, preserving in the connectivity of the trees the architecture of the original graph. We apply this framework to investigate computer generated graphs, such as artificial models and optimal distribution networks, as well as natural graphs extracted from digitized images of dicotyledonous leaves and vasculature of rat cerebral neocortex. We calculate various metrics based on the asymmetry, the cumulative size distribution and the Strahler bifurcation ratios of the corresponding trees and discuss the relationship of these quantities to the architectural organization of the original graphs. This algorithmic framework decouples the geometric information (exact location of edges and nodes) from the metric topology (connectivity and edge weight) and it ultimately allows us to perform a quantitative statistical comparison between predictions of theoretical models and naturally occurring loopy graphs.     

Identification of functional modules in a PPI network by bounded diameter clustering

Dense subgraphs of Protein-Protein Interaction (PPI) graphs are assumed to be potential functional modules and play an important role in inferring the functional behavior of proteins. Increasing amount of available PPI data implies a fast, accurate approach of biological complex identification. Therefore, there are different models and algorithms in identifying functional modules. This paper describes a new graph theoretic clustering algorithm that detects densely connected regions in a large PPI graph. The method is based on finding bounded diameter subgraphs around a seed node. The algorithm has the advantage of being very simple and efficient when compared with other graph clustering methods. This algorithm is tested on the yeast PPI graph and the results are compared with MCL, Core-Attachment, and MCODE algorithms.     

Comparing graph representations of protein structure for mining family-specific residue-based packing motifs.

We find recurring amino-acid residue packing patterns, or spatial motifs, that are characteristic of protein structural families, by applying a novel frequent subgraph mining algorithm to graph representations of protein three-dimensional structure. Graph nodes represent amino acids, and edges are chosen in one of three ways: first, using a threshold for contact distance between residues; second, using Delaunay tessellation; and third, using the recently developed almost-Delaunay edges. For a set of graphs representing a protein family from the Structural Classification of Proteins (SCOP) database, subgraph mining typically identifies several hundred common subgraphs corresponding to spatial motifs that are frequently found in proteins in the family but rarely found outside of it. We find that some of the large motifs map onto known functional regions in two protein families explored in this study, i.e., serine proteases and kinases. We find that graphs based on almost-Delaunay edges significantly reduce the number of edges in the graph representation and hence present computational advantage, yet the patterns extracted from such graphs have a biological interpretation approximately equivalent to that of those extracted from distance based graphs.     

Protein complex prediction based on maximum matching with domain–domain interaction

With the development of high-throughput methods for identifying protein–protein interactions, large scale interaction networks are available. Computational methods to analyze the networks to detect functional modules as protein complexes are becoming more important. However, most of the existing methods only make use of the protein–protein interaction networks without considering the structural limitations of proteins to bind together. In this paper, we design a new protein complex prediction method by extending the idea of using domain–domain interaction information. Here we formulate the problem into a maximum matching problem (which can be solved in polynomial time) instead of the binary integer linear programming approach (which can be NP-hard in the worst case). We also add a step to predict domain–domain interactions which first searches the database Pfam using the hidden Markov model and then predicts the domain–domain interactions based on the database DOMINE and InterDom which contain confirmed DDIs. By adding the domain–domain interaction prediction step, we have more edges in the DDI graph and the recall value is increased significantly (at least doubled) comparing with the method of Ozawa et al. (2010) [1] while the average precision value is slightly better. We also combine our method with three other existing methods, such as COACH, MCL and MCODE. Experiments show that the precision of the combined method is improved. This article is part of a Special Issue entitled: Computational Methods for Protein Interaction and Structural Prediction.