arXiv subjects

Tijana Milenkovic

Publications and source records attributed to Tijana Milenkovic.

At least 19 recordsLinked to original sources

arXiv2021

Modeling multi-scale data via a network of networks

Prediction of node and graph labels are prominent network science tasks. Data analyzed in these tasks are sometimes related: entities represented by nodes in a higher-level (higher-scale) network can themselves be modeled as networks at a lower level. We argue that systems involving such entities should be integrated with a "network of networks" (NoN) representation. Then, we ask whether entity label prediction using multi-level NoN data via our proposed approaches is more accurate than using each of single-level node and graph data alone, i.e., than traditional node label prediction on the higher-level network and graph label prediction on the lower-level networks. To obtain data, we develop the first synthetic NoN generator and construct a real biological NoN. We evaluate accuracy of considered approaches when predicting artificial labels from the synthetic NoNs and proteins' functions from the biological NoN. For the synthetic NoNs, our NoN approaches outperform or are as good as node- and network-level ones depending on the NoN properties. For the biological NoN, our NoN approaches outperform the single-level approaches for just under half of the protein functions, and for 30% of the functions, only our NoN approaches make meaningful predictions, while node- and network-level ones achieve random accuracy. So, NoN-based data integration is important.

q-bio.MN↗

arXiv2021

Dynamic network analysis improves protein 3D structural classification

Protein structural classification (PSC) is a supervised problem of assigning proteins into pre-defined structural (e.g., CATH or SCOPe) classes based on the proteins' sequence or 3D structural features. We recently proposed PSC approaches that model protein 3D structures as protein structure networks (PSNs) and analyze PSN-based protein features, which performed better than or comparable to state-of-the-art sequence or other 3D structure-based approaches in the task of PSC. However, existing PSN-based PSC approaches model the whole 3D structure of a protein as a static PSN. Because folding of a protein is a dynamic process, where some parts of a protein fold before others, modeling the 3D structure of a protein as a dynamic PSN might further help improve the existing PSC performance. Here, we propose for the first time a way to model 3D structures of proteins as dynamic PSNs, with the hypothesis that this will improve upon the current state-of-the-art PSC approaches that are based on static PSNs (and thus upon the existing state-of-the-art sequence and other 3D structural approaches). Indeed, we confirm this on 71 datasets spanning ~44,000 protein domains from CATH and SCOPe

q-bio.MN↗

arXiv2020

Data-driven biological network alignment that uses topological, sequence, and functional information

Many proteins remain functionally unannotated. Sequence alignment (SA) uncovers missing annotations by transferring functional knowledge between species' sequence-conserved regions. Because SA is imperfect, network alignment (NA) complements SA by transferring functional knowledge between conserved biological network, rather than just sequence, regions of different species. Existing NA assumes that it is topological similarity (isomorphic-like matching) between network regions that corresponds to the regions' functional relatedness. However, we recently found that functionally unrelated proteins are almost as topologically similar as functionally related proteins. So, we redefined NA as a data-driven framework, TARA, which learns from network and protein functional data what kind of topological relatedness (rather than similarity) between proteins corresponds to the proteins' functional relatedness. TARA used topological information (within each network) but not sequence information (between proteins across networks). Yet, its alignments yielded higher protein functional prediction accuracy than alignments of existing NA methods, even those that used both topological and sequence information. Here, we propose TARA++ that is also data-driven, like TARA and unlike other existing methods, but that uses across-network sequence information on top of within-network topological information, unlike TARA. To deal with the within-and-across-network analysis, we adapt social network embedding to the problem of biological NA. TARA++ outperforms protein functional prediction accuracy of existing methods.

q-bio.MN↗

arXiv2020

Data-driven network alignment

Biological network alignment (NA) aims to find a node mapping between species' molecular networks that uncovers similar network regions, thus allowing for transfer of functional knowledge between the aligned nodes. However, current NA methods do not end up aligning functionally related nodes. A likely reason is that they assume it is topologically similar nodes that are functionally related. However, we show that this assumption does not hold well. So, a paradigm shift is needed with how the NA problem is approached. We redefine NA as a data-driven framework, TARA (daTA-dRiven network Alignment), which attempts to learn the relationship between topological relatedness and functional relatedness without assuming that topological relatedness corresponds to topological similarity, like traditional NA methods do. TARA trains a classifier to predict whether two nodes from different networks are functionally related based on their network topological patterns. We find that TARA is able to make accurate predictions. TARA then takes each pair of nodes that are predicted as related to be part of an alignment. Like traditional NA methods, TARA uses this alignment for the across-species transfer of functional knowledge. Clearly, TARA as currently implemented uses topological but not protein sequence information for this task. We find that TARA outperforms existing state-of-the-art NA methods that also use topological information, WAVE and SANA, and even outperforms or complements a state-of-the-art NA method that uses both topological and sequence information, PrimAlign. Hence, adding sequence information to TARA, which is our future work, is likely to further improve its performance.

q-bio.MN↗

arXiv2020

Pairwise versus multiple network alignment

Biological network alignment (NA) aims to identify similar regions between molecular networks of different species. NA can be local or global. Just as the recent trend in the NA field, we also focus on global NA, which can be pairwise (PNA) and multiple (MNA). PNA produces aligned node pairs between two networks. MNA produces aligned node clusters between more than two networks. Recently, the focus has shifted from PNA to MNA, because MNA captures conserved regions between more networks than PNA (and MNA is thus considered to be more insightful), though at higher computational complexity. The issue is that, due to the different outputs of PNA and MNA, a PNA method is only compared to other PNA methods, and an MNA method is only compared to other MNA methods. Comparison of PNA against MNA must be done to evaluate whether MNA's higher complexity is justified by its higher accuracy. We introduce a framework that allows for this. We evaluate eight prominent PNA and MNA methods, on synthetic and real-world biological networks, using topological and functional alignment quality measures. We compare PNA against MNA in both a pairwise (native to PNA) and multiple (native to MNA) manner. PNA is expected to perform better under the pairwise evaluation framework. Indeed this is what we find. MNA is expected to perform better under the multiple evaluation framework. Shockingly, we find this not to always hold; PNA is often better than MNA in this framework, depending on the choice of evaluation test.

q-bio.MN↗

arXiv2020

Inference of the Dynamic Aging-related Biological Subnetwork via Network Propagation

Gene expression (GE) data capture valuable condition-specific information ("condition" can mean a biological process, disease stage, age, patient, etc.) However, GE analyses ignore physical interactions between gene products, i.e., proteins. Since proteins function by interacting with each other, and since biological networks (BNs) capture these interactions, BN analyses are promising. However, current BN data fail to capture condition-specific information. Recently, GE and BN data have been integrated using network propagation (NP) to infer condition-specific BNs. However, existing NP-based studies result in a static condition-specific subnetwork, even though cellular processes are dynamic. A dynamic process of our interest is human aging. We use prominent existing NP methods in a new task of inferring a dynamic rather than static condition-specific (aging-related) subnetwork. Then, we study evolution of network structure with age - we identify proteins whose network positions significantly change with age and predict them as new aging-related candidates. We validate the predictions via e.g., functional enrichment analyses and literature search. Dynamic network inference via NP yields higher prediction quality than the only existing method for inferring a dynamic aging-related BN, which does not use NP.

q-bio.MN↗

arXiv2020

Network-based protein structural classification

Experimental determination of protein function is resource-consuming. As an alternative, computational prediction of protein function has received attention. In this context, protein structural classification (PSC) can help, by allowing for determining structural classes of currently unclassified proteins based on their features, and then relying on the fact that proteins with similar structures have similar functions. Existing PSC approaches rely on sequence-based or direct 3-dimensional (3D) structure-based protein features. In contrast, we first model 3D structures of proteins as protein structure networks (PSNs). Then, we use network-based features for PSC. We propose the use of graphlets, state-of-the-art features in many research areas of network science, in the task of PSC. Moreover, because graphlets can deal only with unweighted PSNs, and because accounting for edge weights when constructing PSNs could improve PSC accuracy, we also propose a deep learning framework that automatically learns network features from weighted PSNs. When evaluated on a large set of ~9,400 CATH and ~12,800 SCOP protein domains (spanning 36 PSN sets), our proposed approaches are superior to existing PSC approaches in terms of accuracy, with comparable running time.

q-bio.MN↗

arXiv2020

Heterogeneous network approach to predict individuals' mental health

Depression and anxiety are critical public health issues affecting millions of people around the world. To identify individuals who are vulnerable to depression and anxiety, predictive models have been built that typically utilize data from one source. Unlike these traditional models, in this study, we leverage a rich heterogeneous data set from the University of Notre Dame's NetHealth study that collected individuals' (student participants') social interaction data via smartphones, health-related behavioral data via wearables (Fitbit), and trait data from surveys. To integrate the different types of information, we model the NetHealth data as a heterogeneous information network (HIN). Then, we redefine the problem of predicting individuals' mental health conditions (depression or anxiety) in a novel manner, as applying to our HIN a popular paradigm of a recommender system (RS), which is typically used to predict the preference that a person would give to an item (e.g., a movie or book). In our case, the items are the individuals' different mental health states. We evaluate four state-of-the-art RS approaches. Also, we model the prediction of individuals' mental health as another problem type - that of node classification (NC) in our HIN, evaluating in the process four node features under logistic regression as a proof-of-concept classifier. We find that our RS and NC network methods produce more accurate predictions than a logistic regression model using the same NetHealth data in the traditional non-network fashion as well as a random-approach. Also, we find that the best of the considered RS approaches outperforms all considered NC approaches. This is the first study to integrate smartphone, wearable sensor, and survey data in an HIN manner and use RS or NC on the HIN to predict individuals' mental health conditions.

cs.SI↗

arXiv2019

Weighted graphlets and deep neural networks for protein structure classification

As proteins with similar structures often have similar functions, analysis of protein structures can help predict protein functions and is thus important. We consider the problem of protein structure classification, which computationally classifies the structures of proteins into pre-defined groups. We develop a weighted network that depicts the protein structures, and more importantly, we propose the first graphlet-based measure that applies to weighted networks. Further, we develop a deep neural network (DNN) composed of both convolutional and recurrent layers to use this measure for classification. Put together, our approach shows dramatic improvements in performance over existing graphlet-based approaches on 36 real datasets. Even comparing with the state-of-the-art approach, it almost halves the classification error. In addition to protein structure networks, our weighted-graphlet measure and DNN classifier can potentially be applied to classification of other weighted networks in computational biology as well as in other domains.

stat.ML↗

arXiv2019

The power of dynamic social networks to predict individuals' mental health

Precision medicine has received attention both in and outside the clinic. We focus on the latter, by exploiting the relationship between individuals' social interactions and their mental health to develop a predictive model of one's likelihood to be depressed or anxious from rich dynamic social network data. To our knowledge, we are the first to do this. Existing studies differ from our work in at least one aspect: they do not model social interaction data as a network; they do so but analyze static network data; they examine "correlation" between social networks and health but without developing a predictive model; or they study other individual traits but not mental health. In a systematic and comprehensive evaluation, we show that our predictive model that uses dynamic social network data is superior to its static network as well as non-network equivalents when run on the same data.

cs.SI↗

arXiv2019

Network analysis of synonymous codon usage

Most amino acids are encoded by multiple synonymous codons. For an amino acid, some of its synonymous codons are used much more rarely than others. Analyses of positions of such rare codons in protein sequences revealed that rare codons can impact co-translational protein folding and that positions of some rare codons are evolutionary conserved. Analyses of positions of rare codons in proteins' 3-dimensional structures, which are richer in biochemical information than sequences alone, might further explain the role of rare codons in protein folding. We analyze a protein set recently annotated with codon usage information, considering non-redundant proteins with sufficient structural information. We model the proteins' structures as networks and study potential differences between network positions of amino acids encoded by evolutionary conserved rare, evolutionary non-conserved rare, and commonly used codons. In 84% of the proteins, at least one of the three codon categories occupies significantly more or less network-central positions than the other codon categories. Different protein groups showing different codon centrality trends (i.e., different types of relationships between network positions of the three codon categories) are enriched in different biological functions, implying the existence of a link between codon usage, protein folding, and protein function.

q-bio.MN↗

arXiv2018

Graphlets versus node2vec and struc2vec in the task of network alignment

Network embedding aims to represent each node in a network as a low-dimensional feature vector that summarizes the given node's (extended) network neighborhood. The nodes' feature vectors can then be used in various downstream machine learning tasks. Recently, many embedding methods that automatically learn the features of nodes have emerged, such as node2vec and struc2vec, which have been used in tasks such as node classification, link prediction, and node clustering, mainly in the social network domain. There are also other embedding methods that explicitly look at the connections between nodes, i.e., the nodes' network neighborhoods, such as graphlets. Graphlets have been used in many tasks such as network comparison, link prediction, and network clustering, mainly in the computational biology domain. Even though the two types of embedding methods (node2vec/struct2vec versus graphlets) have a similar goal -- to represent nodes as features vectors, no comparisons have been made between them, possibly because they have originated in the different domains. Therefore, in this study, we compare graphlets to node2vec and struc2vec, and we do so in the task of network alignment. In evaluations on synthetic and real-world biological networks, we find that graphlets are both more accurate and faster than node2vec and struc2vec.

cs.SI↗

arXiv2018

From homogeneous to heterogeneous network alignment via colored graphlets

Network alignment (NA) compares networks with the goal of finding a node mapping that uncovers highly similar (conserved) network regions. Existing NA methods are homogeneous, i.e., they can deal only with networks containing nodes and edges of one type. Due to increasing amounts of heterogeneous network data with nodes or edges of different types, we extend three recent state-of-the-art homogeneous NA methods, WAVE, MAGNA++, and SANA, to allow for heterogeneous NA for the first time. We introduce several algorithmic novelties. Namely, these existing methods compute homogeneous graphlet-based node similarities and then find high-scoring alignments with respect to these similarities, while simultaneously maximizing the amount of conserved edges. Instead, we extend homogeneous graphlets to their heterogeneous counterparts, which we then use to develop a new measure of heterogeneous node similarity. Also, we extend $S^3$, a state-of-the-art measure of edge conservation for homogeneous NA, to its heterogeneous counterpart. Then, we find high-scoring alignments with respect to our heterogeneous node similarity and edge conservation measures. In evaluations on synthetic and real-world biological networks, our proposed heterogeneous NA methods lead to higher-quality alignments and better robustness to noise in the data than their homogeneous counterparts. The software and data from this work is available upon request.

q-bio.MN↗

arXiv2017

Network approach integrates 3D structural and sequence data to improve protein structural comparison

Initial protein structural comparisons were sequence-based. Since amino acids that are distant in the sequence can be close in the 3-dimensional (3D) structure, 3D contact approaches can complement sequence approaches. Traditional 3D contact approaches study 3D structures directly. Instead, 3D structures can be modeled as protein structure networks (PSNs). Then, network approaches can compare proteins by comparing their PSNs. Network approaches may improve upon traditional 3D contact approaches. We cannot use existing PSN approaches to test this, because: 1) They rely on naive measures of network topology. 2) They are not robust to PSN size. They cannot integrate 3) multiple PSN measures or 4) PSN data with sequence data, although this could help because the different data types capture complementary biological knowledge. We address these limitations by: 1) exploiting well-established graphlet measures via a new network approach, 2) introducing normalized graphlet measures to remove the bias of PSN size, 3) allowing for integrating multiple PSN measures, and 4) using ordered graphlets to combine the complementary PSN data and sequence data. We compare both synthetic networks and real-world PSNs more accurately and faster than existing network, 3D contact, or sequence approaches. Our approach finds PSN patterns that may be biochemically interesting.

q-bio.MN↗

arXiv2017

Alignment of dynamic networks

Networks can model real-world systems in a variety of domains. Network alignment (NA) aims to find a node mapping that conserves similar regions between compared networks. NA is applicable to many fields, including computational biology, where NA can guide the transfer of biological knowledge from well- to poorly-studied species across aligned network regions. Existing NA methods can only align static networks. However, most complex real-world systems evolve over time and should thus be modeled as dynamic networks. We hypothesize that aligning dynamic network representations of evolving systems will produce superior alignments compared to aligning the systems' static network representations, as is currently done. For this purpose, we introduce the first ever dynamic NA method, DynaMAGNA++. This proof-of-concept dynamic NA method is an extension of a state-of-the-art static NA method, MAGNA++. Even though both MAGNA++ and DynaMAGNA++ optimize edge as well as node conservation across the aligned networks, MAGNA++ conserves static edges and similarity between static node neighborhoods, while DynaMAGNA++ conserves dynamic edges (events) and similarity between evolving node neighborhoods. For this purpose, we introduce the first ever measure of dynamic edge conservation and rely on our recent measure of dynamic node conservation. Importantly, the two dynamic conservation measures can be optimized using any state-of-the-art NA method and not just MAGNA++. We confirm our hypothesis that dynamic NA is superior to static NA, under fair comparison conditions, on synthetic and real-world networks, in computational biology and social network domains. DynaMAGNA++ is parallelized and it includes a user-friendly graphical interface.

cs.SI↗

arXiv2016

IGLOO: Integrating global and local biological network alignment

Analogous to genomic sequence alignment, biological network alignment (NA) aims to find regions of similarities between molecular networks (rather than sequences) of different species. NA can be either local (LNA) or global (GNA). LNA aims to identify highly conserved common subnetworks, which are typically small, while GNA aims to identify large common subnetworks, which are typically suboptimally conserved. We recently showed that LNA and GNA yield complementary results: LNA has high functional but low topological alignment quality, while GNA has high topological but low functional alignment quality. Thus, we propose IGLOO, a new approach that integrates GNA and LNA in hope to reconcile the two. We evaluate IGLOO against state-of-the-art LNA (NetworkBLAST, NetAligner, AlignNemo, and AlignMCL) and GNA (GHOST, NETAL, GEDEVO, MAGNA++, WAVE, and L-GRAAL) methods. We show that IGLOO allows for a trade-off between topological and functional alignment quality better than the existing LNA and GNA methods considered in our study.

q-bio.MN↗

arXiv2016

SCOUT: simultaneous time segmentation and community detection in dynamic networks

Many evolving complex systems can be modeled via dynamic networks. An important problem in dynamic network research is community detection, which identifies groups of topologically related nodes. Typically, this problem is approached by assuming either that each time point has a distinct community organization or that all time points share one community organization. In reality, the truth likely lies between these two extremes, since some time periods can have community organization that evolves while others can have community organization that stays the same. To find the compromise, we consider community detection in the context of the problem of segment detection, which identifies contiguous time periods with consistent network structure. Consequently, we formulate a combined problem of segment community detection (SCD), which simultaneously partitions the network into contiguous time segments with consistent community organization and finds this community organization for each segment. To solve SCD, we introduce SCOUT, an optimization framework that explicitly considers both segmentation quality and partition quality. SCOUT addresses limitations of existing methods that can be adapted to solve SCD, which typically consider only one of segmentation quality or partition quality. In a thorough evaluation, SCOUT outperforms the existing methods in terms of both accuracy and computational complexity.

cs.SI↗

arXiv2016

Multiple network alignment via multiMAGNA++

Network alignment (NA) aims to find a node mapping between molecular networks of different species that identifies topologically or functionally similar network regions. Analogous to genomic sequence alignment, NA can be used to transfer biological knowledge from well- to poorly-studied species between aligned network regions. Pairwise NA (PNA) finds similar regions between two networks while multiple NA (MNA) can align more than two networks. We focus on MNA. Existing MNA methods aim to maximize total similarity over all aligned nodes (node conservation). Then, they evaluate alignment quality by measuring the amount of conserved edges, but only after the alignment is constructed. Directly optimizing edge conservation during alignment construction in addition to node conservation may result in superior alignments. Thus, we present a novel MNA approach called multiMAGNA++ that can achieve this. Indeed, multiMAGNA++ generally outperforms or is on par with the existing MNA methods, while often completing faster than the existing methods. That is, multiMAGNA++ scales well to larger network data and can be parallelized effectively. During method evaluation, we also introduce new MNA quality measures to allow for more complete alignment characterization as well as more fair MNA method comparison compared to using only the existing alignment quality measures.

q-bio.MN↗