NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration

of Network Medicine Data

Marco Angelini, Graziano Blasilli, Lorenzo Farina, Simone Lenti and Giuseppe Santucci

DIAG, Department of Computer, Control, and Management Engineering Antonio Ruberti,

Sapienza University of Rome, Italy

Keywords:

Network Medicine, Visual Analytics, Interactome.

Abstract:

The emerging Network Medicine domain is causing a shift between diagnosis based on the conventional

reductionist approach, arguing that biological factors work in a simple linear way, and the analysis of per-

turbations within the comprehensive network map of molecular components and their interactions, i.e., the

”Interactome”. As a consequence, clinicians are investigating more than 140,000 interactions between more

than 13,000 genes and their connections with drugs and diseases, along a sequence of ”networks”. Making

sense of this complex structure is a challenging activity and the visual analytics application NEMESIS tries to

attack such a problem allowing for interactively exploring this large body of knowledge, focusing on subsets of

data and investigating their relationships with other relevant dimensions, pursuing the main goal of facilitating

hypothesis formulation and validation.

1 INTRODUCTION

Until recently, the investigation of disease etiology,

diagnosis and treatment, has been based on a con-

ventional reductionist approach. This tenet argues

that critical biological factors work in a simple linear

mechanism to control disease pathobiology. Rather,

they are nearly always the result of multiple pathobi-

ological pathways that interact through an intercon-

nected network: a disease is rarely a direct conse-

quence of an abnormality in a single gene or molec-

ular component (see, e.g., (Chan and Loscalzo, 2012;

Gustafsson et al., 2014)). For example, complex dis-

eases like cancers of different types, have extraor-

dinary complex biological phenomena that underlie

them. Today, big data, genomics, and quantitative in

silico methodologies integration have the potential to

push forward the frontiers of medicine in an unprece-

dented way.

Clinicians, diagnosticians and therapists have long

strived to determine single molecular traits that lead

to diseases. What they had in mind was the idea

that a single golden bullet drug might provide a cure.

However, this reductionist approach largely ignored

the essential complexity of human diseases. Indeed,

a large body of evidence that is now emerging from

new genomic technologies, points out directly to the

cause of disease as perturbations within the interac-

tome, i.e., the comprehensive network map of molec-

ular components and their interactions. Precisely,

the human interactome is composed of direct physi-

cal, regulatory (transcription factors binding), binary,

metabolic enzyme-coupled, protein complexes and

kinase/substrate interactions. Such network is largely

incomplete as well as the connections between genes

and disease. Currently, more than 140,000 interac-

tions between more than 13,000 proteins/genes are

known (see, e.g.,(Korcsmaros et al., 2017; Gustafs-

son et al., 2014)). Consequently, a paradigm shift

is needed towards the development of temporal and

spatial multi-level models, from molecular machiner-

ies to single cells, whole organism and individuals,

including the environment, to reveal the underlying

links among components. This new type of medical

paradigm is called Network Medicine.

The gap between the biological and the informa-

tional mindset can be daunting and might impair from

the beginning the development of shared concepts.

However, the network medicine visualization setting

will certainly also facilitate communications across

disciplines given the immediate and intuitive under-

standing of the network concept and representation, a

visual metaphor that can be used by molecular biolo-

gists to visualize their knowledge in a structured way

ready to be translated into an algorithm on the avail-

able data, as medical or biological goals are deﬁned.

322

Angelini, M., Blasilli, G., Farina, L., Lenti, S. and Santucci, G.

NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration of Network Medicine Data.

DOI: 10.5220/0007577003220329

In Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2019), pages 322-329

ISBN: 978-989-758-354-4

Most of the available proposals have the goal of

visualizing the structure of the relationships existing

among genes, diseases, drugs, and biological process,

showing them as networks (see, e.g., (Sharma et al.,

2015; Gladilin, 2017)) offering, in most cases, only

navigational means to traverse their complex struc-

ture. This paper tries to proceed further in this direc-

tion designing a visual analytics solution that allow

for interacting with such networks in order to steer

analysis patterns. Moreover, taking into account that

exploring a single layer is not enough, the proposed

system, NEMESIS, aims at providing an integrated

vision of both information coming from a single ref-

erence network and new derived data coming from the

analysis of one ore more additional networks.

Summarizing, the contribution of the paper is a

novel visual analytics solution encompassing the fol-

lowing main characteristics:

• the integrated visualization of a single network

with multidimensional data derived from the anal-

ysis of the other connected networks;

• the possibility of focusing the analysis on rele-

vant subsets of data (e.g., subsets of diseases) to

steer the explorations and formulate and validate

hypothesis;

• the availability of complex and long lasting pre-

computed analytics that can be used either for

identifying speciﬁc subsets of data (e.g., similar

diseases) or presenting multidimensional aggre-

gation of one or more networks to provide sum-

mary information.

The paper is structured as follows: Section 2 de-

scribes the Network Medicine application domain,

Section 3 deals with related proposals, Section 4 de-

scribes the system implementation, and Section 5 de-

scribes the results of an informal user study, con-

cludes the paper and outlines future activities.

2 APPLICATION DOMAIN

The human genome sequencing using high-

throughput next-generation devices is being deeply

affecting current visions of biomedical and clinical

research. More recently, entering the era of per-

sonal whole-genome sequencing, 38 million genetic

variants have been discovered, some of which are

rare mutations and thus may be associated with

large size effect. How to use this large amount of

data to generate better understanding of disease and

ﬁnd appropriate drug targets? Looking at networks

without a speciﬁc biomedical bias in mind and let

the data speak by themselves, so to formulate new

hypothesis to be further validated by experimentalist

and so on, moving within a virtuous circle of shared

knowledge. The unifying framework of visual

explorative data analysis perfectly ﬁts the need

for integrated network-based algorithms in order

to reconcile biological network representation and

large-scale data integration. Networks can, in fact,

be obtained from any sort of information: known

genes-genes interactions, gene expression proﬁles,

functional annotation, etc. Recently, it has been

shown that the genes associated with a disease are

localized in speciﬁc neighborhoods, or ’disease mod-

ules’, within such an interaction network called the

interactome (Vidal et al., 2011) (Khler et al., 2008).

The overall ambition of researchers working in this

ﬁeld is to both developing a global understanding

of how interactome perturbations result in disease

traits, and to translate computational insights into

concrete clinical applications, such as new drugs and

therapies, diagnostic tools or prognostic/predictive

markers.

According to this, the paper investigates novel

mechanisms for exploring the different network lay-

ers that constitute the complex information network

medicine has to deal with, i.e., the gene-gene network,

drugs-genes network, diseases-genes network, etc.,

with the main goal of supporting explorative analy-

sis, in order to generate and validate hypothesis. Ac-

cording to this goal, the NEMESIS system allows for

focusing on a subset of data, showing its structure and

relationships on both a reference network and on inte-

grated analysis of the other networks, see Fig. 1. Suit-

able coordinated visualizations and analytics, detailed

on Section 4, provide additional support to the inter-

active analysis.

3 RELATED WORK

Information Visualization and Visual Analytics are

long recognized ﬁelds that provided beneﬁts for med-

ical data analysis, like stated in (Chittaro, 2001)

(Shneiderman et al., 2013). In particular, the ﬁeld of

network medicine is becoming prominent in the last

decade, with several research contributions focused

on applying principles of network analysis to the med-

ical ﬁeld.

Several contributions focused on the visualiza-

tion of interactome ((Chaurasia et al., 2009) (Lu

et al., 2004)), disease module and gene pathways

((Cerami et al., 2010) (Mlecnik et al., 2005)), Elec-

tronic Health Records ((Wang et al., 2011)), phe-

notypes ((Bottomly et al., 2016)) and most of them

use the well-known node-link diagram representation

NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration of Network Medicine Data

323

Figure 1: NEMESIS exploration of the different network

layers. The user selects one reference network (e.g.,the

gene-gene links network) that is represented in the visual-

ization using topological pieces of information of the refer-

ence network (e.g.,distance among nodes) together with de-

rived data computed through a generic aggregation function

that uses the other networks (e.g., the number of diseases

that a pair of genes have in common).

for explaining results and analysis outcomes((Sharma

et al., 2015),(Gladilin, 2017)). However, typically

they do not provide any control or only very basic vi-

sual interactions for dealing with the visual environ-

ment. Moreover, general purpose framework exist,

in the form of environments or libraries, that allow

to visualize large biological networks like Cytoscape

((Smoot et al., 2011)), NetBioV (Tripathi et al., 2014)

or HitWalker 2 (Bottomly et al., 2016). The work in

(Merico et al., 2010) proposes a technique for gene set

enrichment visualization; this technique ﬁnds func-

tionally coherent gene-sets, such as pathways, that are

statistically over-represented in a given gene list. Ide-

ally, the number of resulting sets is smaller than the

number of genes in the list, thus simplifying interpre-

tation. Differently from our approach no interaction

is provided to analyze the results of the application of

the technique, limiting the analysis capabilities. For

what regards visual analytics solutions, Gerasch et al.

(Gerasch et al., 2014) propose a system for visually

analyzing high-throughput omics data in the context

of networks, in particular for the differential analy-

sis and the analysis of time series data. Perer and

Sun (Perer and Sun, 2012) and Basole et al. (Ba-

sole et al., 2015) propose visual analytics solutions

that analyze clinical patients data: the former takes

clinical patients data as event sequences, constructs

time-evolving networks and visualizes them as a tem-

poral ﬂow of matrices; the latter allows for exploring

data about pediatric asthma care processes. Differ-

ently from our approach they only consider patients

data, effectively focusing only on one plane of anal-

ysis and do not include any multidimensional anal-

ysis and or interactome data. The work in (Huan

et al., 2008) presents PRoteoLEns, a JAVA-based vi-

sual analytics tool for creating, annotating, and ex-

ploring multi-scale biological networks. Nonetheless,

the tool seems very proﬁcient in exploring subparts

of a biological network while does not seems good in

communicating an overview. Finally, inspired by the

visual encoding proposed in (Dietzsch et al., 2009),

NEMESIS relies on easy-relatable visual paradigms

for medical and bioinformatics people, that are not

computer scientist and could not necessarily relate to

more abstract visual representations.

4 THE NEMESIS SYSTEM

This section presents NEMESIS (NEtwork MEdicine

analySIS), a visual analytics solution that allows for

exploring groups of similar diseases, studying the

associated genes, their interactions through the in-

teractome, and the relations that could exist among

them, focusing on single dimensions of analysis or

considering multidimensional properties. NEME-

SIS has been developed in collaboration with med-

ical and bioinformatics personnel through an iter-

ative development cycle that produced 3 different

versions, of which the last is the one presented in

this paper. A working prototype is available at:

http://awareserver.dis.uniroma1.it/nemesis/. Actually

the prototype uses the OMIM dataset [2] of diseases

genes characteristics, composed by:

• 13.401 genes;

• 70 diseases, 20-60 genes per disease (avg: 40);

• 138.405 direct interactions between genes (path of

length 1);

• 5.230.666 computed indirect interactions (derived

data, computing path of length 2).

The NEMESIS environment is visible in Figure 2;

it is composed of 3 main panes: the Interactions Ma-

trix pane that shows data about the relations and path-

ways of interest, the Genes scatter-plot pane, that vi-

sualizes characteristics speciﬁc to genes, and the Dis-

eases Analysis pane, that shows data about the con-

sidered diseases. Each of them is described in the fol-

lowing sections.

4.1 Diseases Analysis Pane

This pane allows to select the list of diseases the anal-

ysis will concentrate on. It reports the set of all dis-

eases under analysis, i.e., cancer diseases, whose se-

lection affects all the other linked visualizations. Each

disease is selectable/deselectable through the associ-

ated check-box, with the environment reconﬁguring

IVAPP 2019 - 10th International Conference on Information Visualization Theory and Applications

324

Figure 2: The NEMESIS user interface. The left side contains the Interactions matrix pane, representing interactions among

genes using an enriched matrix representation; the top-right contains the Genes scatter-plot, allowing for inspecting charac-

teristics of speciﬁc genes or subsets of them; ﬁnally, the bottom-right part contains the interactive diseases list that allows the

analyst to steer the analysis and activate analytics.

accordingly. For each disease, the number of associ-

ated genes and the number of genes involved by the

actual analysis workﬂow are reported. For obtaining

rapid steering of the analysis, accelerators are present

in the form of “Select All/Deselect All” buttons. In

this way the analyst can choose to inspect genes com-

ing from a starting set of diseases, reﬁning the anal-

ysis on smaller subsets, or even inspecting the genes

belonging to a single disease. An example of these

three cases is visible in Figure 3.

Selecting a subset of diseases will inﬂuence the

data presented in the Genes scatter-plot and Interac-

tions Matrix panes, and so it is possible by mouse-

hovering on a disease to see the corresponding genes

highlighted in the Genes scatter-plot (in red), with the

capability to incrementally select more diseases and

explore their intersection by clicking on the name of

the disease. The ﬁxed disease are represented with a

red background, as visible in Figure 4.

Additionally, while computing interactions among

disease genes it is possible to compute paths of length

2 either using only diseases genes or considering all

the interactome genes (by deselecting the “interac-

tome disease only” ﬂag).

An important feature of the NEMESIS system is

the capability to drive the analysis using complex an-

alytics, obtained computing from million to hundreds

of millions of combinations. These functionalities are

labeled as “Biomarkers”, and allow to compute the

Top 5 intersecting diseases considering only the direct

interactions (maximizing intersection at distance=1),

the Top 5 intersecting diseases considering both direct

and indirect interactions (maximizing intersection up

to distance=2), and the Top 5 diseases that maximize

the genes sensitivity. In Figure 8 is visible an ex-

ample in which is computed the Top 5 of diseases

that maximizes intersection considering direct inter-

actions. The highlighting of regular patterns in the

matrix and the strong reduction in cardinality of data

to process make the use of biomarkers very helpful in

conducting analysis.

4.2 Interactions Matrix Pane

This pane contains an enriched interactive visualiza-

tion of the interactome data of interest. Differently

from many contributions that focus on representing

the genes interactions as a node-link diagram, we used

the well-known matrix visual paradigm for represent-

ing the interctome in order to exploit its better read-

ability properties (Ghoniem et al., 2004) and to repre-

sent paths of any length. Indeed, it represents three

types of interactions between genes; genes that do

not have any interactions are represented with a black

square, genes that are directly interacting (connected

through a single link in the interactome) are repre-

sented as a blue square and genes that are indirectly

interacting (a pathway of length 2 exists in the in-

teractome connecting them) are represented as an or-

ange square. In this way the analyst can inspect the

overview of the existing relations among genes that

belongs to the inspected set of diseases, spotting ar-

eas with high number of interactions (likely deserv-

ing further investigations) or low number of interac-

NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration of Network Medicine Data

325

Figure 3: The ﬁgure shows three different conﬁgurations of the NEMESIS system, obtained by different selections of diseases.

On the left is visible the environment conﬁgured to analyze all the available diseases; in the center the analyst has selected a

subset of diseases, deselecting the ones she is not interested in, with both the Interactions matrix and the Genes scatter-plot

reconﬁguring accordingly; on the right the analyst has chosen to analyze a single disease, effectively lowering the cardinality

of genes and interactions to analyze.

Figure 4: The analyst selects the lynphoma and leukemia b-cell diseases, blocking the ﬁrst and hovering the second for

comparing modules. The analysis, superimposed on the previous steps (selection of high sensitivity genes), highlights that in

the modules of these diseases is present only one gene with an high sensitivity, ﬁlled in green with a red border.

tions (likely to be discarded). In order to help the

analyst in her work a set of metadata is represented

atop the matrix, showing the number of considered

genes, number of interacting genes, total number of

interactions, number of genes common to the cho-

sen set of diseases expanding the set of genes asso-

ciated to a disease considering only direct interaction

(Intersecting D1) or even indirect interactions (Inter-

secting D2). The analyst can choose to show only

a speciﬁc type of interactions (length 1, length 2, or

both) by acting on the interactive legend present at the

right side of the matrix; this mechanism helps in re-

ducing the cardinality of elements to explore starting

by no particular preconditions; the analyst can at any

time explore single interactions obtaining an informa-

tive tooltip containing the IDs of the two interacting

genes, their distance in the interactome and the path

that connects them. An example is visible in Figure 5.

By mouse-hovering on one element of the matrix

the interacting genes are represented using blinking

animation in the Genes scatter-plot pane in order to

identify their derived characteristics. Finally, it is pos-

sible, by clicking on the “Multidimensional similar-

ity” ﬂag in the Diseases analysis pane, to switch be-

tween this mono-dimensional view, considering only

the interactions, and a multi-dimensional one consid-

ering all the dimensions that are tied to a pair of genes

and that are considered to deﬁne a similarity measure

between interacting genes. The resulting encoding

shows all the previous data regarding interactions in

the inferior triangular matrix, while the superior tri-

angular matrix is used to map the multidimensional

similarity measure. For this encoding we get inspired

by the work done in (Berger et al., 2008). The simi-

larity measure is normalized in the range [0,1] and as-

sociated with a linear color-scale ranging from black

to purple. An example of its use is visible in Figure 6.

4.3 Genes Scatter-plot Pane

The matrix representation is very effective in present-

ing an overview of the interactions and similarities

and, in order to complement it with information about

other genes characteristics, NEMESIS includes a vi-

IVAPP 2019 - 10th International Conference on Information Visualization Theory and Applications

326

Figure 5: The ﬁgure shows a detail of the Interactions ma-

trix pane. The analyst has selected to show only direct inter-

actions, producing a visualization that shows different den-

sities of direct interactions, higher in the top-left part. An

informative tooltip shows, on demand, detailed information.

Figure 6: Having selected the multidimensional similarity

check-box, the analyst is presented with the view in ﬁg-

ure. The inferior triangular matrix represents the interac-

tions among genes, while the superior triangular matrix rep-

resents their similarity (w.r.t., the number of diseases that

are associated with both of them). The analyst can spot

similarity between interactions, and check whether genes

interact through the interactome and/or are similar.

sual environment called Genes scatter-plot pane that

relies on the classic scatter-plot visual paradigm to

represent other genes properties. The visual encod-

ing is the following: the x-axis is associated with the

ID of the genes (allowing to manage and identify the

genes available for analysis), the y-axis encodes the

genes sensitivity, the size encodes the genes degree,

and a color-coding is used to communicate the pres-

ence or not of the gene in the current selected set of

Figure 7: The ﬁgure shows an example of selection of in-

teresting part of the interactome. The analyst selects all the

gens with an high sensitivity (represented in green) by click-

ing them, and obtain the subselection of the Interactions

matrix with all the genes interacting with the selected ones.

Additionally, the Diseases Analysis pane reports in green

the part of the selected genes that belong to each disease.

diseases. The sensitivity of a gene is a derived charac-

teristic and represents the weight that a gene has in a

disease, in terms of all the genes associated with that

disease. As an example, let assume that we are focus-

ing on genes G1 and G2, and considering a set of only

2 diseases (d1 and d2), with their modules expressed

in the form:

• d

= [c

(1,1)

G1,c

(2,1)

G2,c

(3,1)

G3,c

(4,1)

G4,c

(5,1)

G5]

• d

= [c

(1,2)

G1,c

(9,2)

G9]

Assuming c

(i, j)

= 1/|d

• sensitivity

(G1) = 1/5 = 20%

• sensitivity

(G1) = 1/2 = 50%

Mouse-hovering on a gene allows the analyst to

inspect a tooltip presenting additional characteristics

of the gene, like its ID and degree, the percentage of

diseases in which it appears (out of all the 70 classi-

ﬁed diseases), and the diseases to which it belongs to.

Additionally, the mouse-hover will highlight in the In-

teractions matrix pane the row of all the genes that in-

teract with the selected one and in the Diseases Anal-

ysis pane all the diseases it belongs to. The analyst

can even lock this gene by clicking on it, effectively

allowing to select a subset of genes and resulting ma-

trix rows to be inspected. This interaction allows for

selecting additional sub-areas of interest in the matrix,

by leading the analyst choosing either genes with high

sensitivity (and/or percentage,degree) or by selecting

genes belonging to speciﬁc diseases (highlighted in

the Diseases analysis pane). Figure 7 presents an ex-

ample of this behavior.

4.4 Workﬂow

The NEMESIS system provides the analyst with

several ways of conducting her analyses: choosing

the starting point considering the genes interactions

NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration of Network Medicine Data

327

Figure 8: The ﬁgure shows the selection of the ﬁrst

biomarker (top 5 intersecting D1); the cardinality of inter-

actions to consider is strongly reduced to 384, showing evi-

dent regular patterns of interaction among genes.

overview, or starting from the set of diseases to fo-

cus the analysis on, or starting from speciﬁc genes

characteristics and identifying the set of diseases that

have them in common. Independently from the start-

ing point of analysis, the available interaction means

and the interconnection among the 3 visual panes al-

low the analyst to progressively reﬁne the analysis

results, identifying speciﬁc areas of interest and pro-

jecting them in the other panes to focus on different

dimensions. The result is a fast way to reduce the

cardinality of elements to consider facilitating the de-

cision on where to lead the analysis next. Finally, the

capability to relate the state of the analysis with the

multidimensional similarity, expand the analysis to-

wards the parts of the interactome not directly con-

nected to the considered diseases, allowing the ana-

lyst to conduct the analysis beyond initial hypothesis

and to better understand the characteristics of the un-

derlying models and to formulate/validate hypothesis.

4.5 Analytics

The actual version of NEMESIS encompasses several

analytics that support the exploration activities:

Off line analytics. We have computed useful de-

rived data that is available during the explorative anal-

ysis. As an example, in order to create relevant

analysis entry points on diseases, we have precom-

puted all the top 5 diseases (more than 12 millions

of combinations, two months of running time on a

server with 32 cores and 96 GB of ram) in order to

select those that maximize relevant objective func-

tions, like those that maximize the number of shared

genes, or those that maximize the average sensitiv-

ity of the associated genes. Moreover we have pre-

computed interactome paths of increasing length to

quickly switch between direct and indirect interac-

tions. We are currently computing summary infor-

mation associated with such paths, information that

comes from other “networks”, e.g., we have precom-

puted for each pair of genes connected by a path on

the interactome the number of shared diseases (used

to drive the actual NEMESIS multidimensional sim-

ilarity analysis). Summary information coming from

multiple networks can be combined to produce mul-

tidimensional information that feeds the aggregation

functions depicted on Figure 6.

On line analytics. While the user is exploring the

data, it is possible to derive information starting from

the current selection, like gene sensitivity, or expand-

ing the set of genes associated with a selected disease

with all genes reachable through the interactome, us-

ing paths of length one and two.

5 CONCLUSION & FUTURE

WORK

NEMESIS has been informally evaluated both during

its development and during two meetings with medi-

cal personnel coming from the oncology department

of Sapienza Medical School and specialists in the net-

work medicine ﬁelds. The prototype has been well re-

ceived and the support to exploratory analysis in par-

ticular has been found very useful for identifying ar-

eas of uncertainty in the interactome and or hypothe-

sizing possible interactions between genes and/or be-

tween disease modules. Suggestions for integrating

clinical images related to particular diseases and capa-

bility to annotate and share analysis results have been

made and we are considering their implementation in

the next version.

In conclusion this paper presented NEMESIS, a

novel visual analytics solution aiming at fostering the

interactive visual exploration of the complex network

medicine data. The proposed solution provides means

for interactively exploring different facets of the com-

plex body of data, inspecting both the data associ-

ated to topological properties of a single network and

summary multidimensional information coming from

other relevant networks. The summary information

relies on several off line analytics. The prototype has

been informally evaluated with oncology doctors, get-

ting positive feedback on the used visualizations and

high interest for the explorative visual analytics ap-

proach that has been perceived as a novelty in the

ﬁeld. Concerning short term future work we are cur-

rently working on the usability of the system and in-

corporating the suggestions raised during the informal

evaluation. As a more ambitious objective we are de-

signing and experimenting more comprehensive deﬁ-

nition of similarity with the goal of producing a more

informative and useful summary overview, combining

the information coming from a larger set of relevant

networks. Moreover, being the actual implementation

IVAPP 2019 - 10th International Conference on Information Visualization Theory and Applications

328

based on several off line analytics we are studying

how to modify the analytical workﬂow adopting the

solutions coming from the emerging ﬁeld of Progres-

sive Visual Analytics (see, e.g., (Schulz et al., 2016)).

REFERENCES

Basole, R. C., Braunstein, M. L., Kumar, V., Park, H.,

Kahng, M., Chau, D. H. P., Tamersoy, A., Hirsh,

D. A., Serban, N., Bost, J., Lesnick, B., Schissel,

B. L., and Thompson, M. (2015). Understanding

variations in pediatric asthma care processes in the

emergency department using visual analytics. Jour-

nal of the American Medical Informatics Association,

22(2):318–323.

Berger, P., Chegini, M., Schumann, H., and Tominski, C.

(2008). Integrated visualization of structure and at-

tribute similarity of multivariate graphs.

Bottomly, D., McWeeney, S. K., and Wilmot, B. (2016).

Hitwalker2: visual analytics for precision medicine

and beyond. Bioinformatics, 32(8):1253–1255.

Cerami, E., Demir, E., Schultz, N., Taylor, B. S., and

Sander, C. (2010). Automated network analysis

identiﬁes core pathways in glioblastoma. PloS one,

5(2):e8918.

Chan, S. Y. and Loscalzo, J. (2012). The emerging

paradigm of network medicine in the study of human

disease. Circulation research, 111(3):359–374.

Chaurasia, G., Malhotra, S., Russ, J., Schnoegl, S., Hnig,

C., Wanker, E. E., and Futschik, M. E. (2009). Unihi

4: new tools for query, analysis and visualization of

the human proteinprotein interactome. Nucleic Acids

Research, 37(suppl1):D657–D660.

Chittaro, L. (2001). Information visualization and its appli-

cation to medicine. Artiﬁcial intelligence in medicine,

22(2):81–88.

Dietzsch, J., Heinrich, J., Nieselt, K., and Bartz, D. (2009).

Spray: A visual analytics approach for gene expres-

sion data. In 2009 IEEE Symposium on Visual Analyt-

ics Science and Technology, pages 179–186.

Gerasch, A., Faber, D., K

untzer, J., Niermann, P.,

Kohlbacher, O., Lenhof, H.-P., and Kaufmann, M.

(2014). Bina: a visual analytics tool for biological

network data. PloS one, 9(2):e87397.

Ghoniem, M., Fekete, J.-D., and Castagliola, P. (2004). A

comparison of the readability of graphs using node-

link and matrix-based representations. In INFOVIS

2004. IEEE Symposium on, pages 17–24. Ieee.

Gladilin, E. (2017). Graph-theoretical model of global

human interactome reveals enhanced long-range

communicability in cancer networks. PloS one,

12(1):e0170953.

Gustafsson, M., Nestor, C. E., Zhang, H., Barab

asi, A.-

L., Baranzini, S., Brunak, S., Chung, K. F., Federoff,

H. J., Gavin, A.-C., Meehan, R. R., et al. (2014). Mod-

ules, networks and systems medicine for understand-

ing disease and aiding diagnosis. Genome medicine,

6(10):82.

Huan, T., Sivachenko, A. Y., Harrison, S. H., and Chen, J. Y.

(2008). Proteolens: a visual analytic tool for multi-

scale database-driven biological network data mining.

BMC Bioinformatics, 9(9):S5.

Khler, S., Bauer, S., Horn, D., and Robinson, P. N. (2008).

Walking the interactome for prioritization of candi-

date disease genes. The American Journal of Human

Genetics, 82(4):949 – 958.

Korcsmaros, T., Schneider, M. V., and Superti-Furga, G.

(2017). Next generation of network medicine: inter-

disciplinary signaling approaches. Integrative Biol-

ogy, 9(2):97–108.

Lu, H., Zhu, X., Liu, H., Skogerb, G., Zhang, J., Zhang, Y.,

Cai, L., Zhao, Y., Sun, S., Xu, J., Bu, D., and Chen, R.

(2004). The interactome as a treean attempt to visu-

alize the proteinprotein interaction network in yeast.

Nucleic Acids Research, 32(16):4804–4811.

Merico, D., Isserlin, R., Stueker, O., Emili, A., and Bader,

G. D. (2010). Enrichment map: a network-based

method for gene-set enrichment visualization and in-

terpretation. PloS one, 5(11):e13984.

Mlecnik, B., Scheideler, M., Hackl, H., Hartler, J., Sanchez-

Cabo, F., and Trajanoski, Z. (2005). Pathwayexplorer:

web service for visualizing high-throughput expres-

sion data on biological pathways. Nucleic Acids Re-

search, 33(suppl2):W633–W637.

Perer, A. and Sun, J. (2012). Matrixﬂow: temporal net-

work visual analytics to track symptom evolution dur-

ing disease progression. In AMIA annual symposium

proceedings, volume 2012, page 716. American Med-

ical Informatics Association.

Schulz, H.-J., Angelini, M., Santucci, G., and Schumann,

H. (2016). An enhanced visualization process model

for incremental visualization. IEEE transactions

on visualization and computer graphics, 22(7):1830–

1842.

Sharma, A., Menche, J., Huang, C. C., Ort, T., Zhou, X.,

Kitsak, M., Sahni, N., Thibault, D., Voung, L., Guo,

F., et al. (2015). A disease module in the interactome

explains disease heterogeneity, drug response and cap-

tures novel pathways and genes in asthma. Human

molecular genetics, 24(11):3005–3020.

Shneiderman, B., Plaisant, C., and Hesse, B. W. (2013).

Improving healthcare with interactive visualization.

Computer, 46(5):58–66.

Smoot, M. E., Ono, K., Ruscheinski, J., Wang, P.-L., and

Ideker, T. (2011). Cytoscape 2.8: new features for data

integration and network visualization. Bioinformatics,

27(3):431–432.

Tripathi, S., Dehmer, M., and Emmert-Streib, F. (2014).

Netbiov: an r package for visualizing large net-

work data in biology and medicine. Bioinformatics,

30(19):2834–2836.

Vidal, M., Cusick, M., and Barabsi, A.-L. (2011). Interac-

tome networks and human disease. Cell, 144(6):986 –

998.

Wang, T. D., Wongsuphasawat, K., Plaisant, C., and Shnei-

derman, B. (2011). Extracting insights from electronic

health records: Case studies, a visual analytics process

model, and design recommendations. Journal of Med-

ical Systems, 35(5):1135–1152.

NEMESIS (NEtwork MEdicine analySIS): Towards Visual Exploration of Network Medicine Data

329