Available online at www.sciencedirect.com
ScienceDirect
Current Opinion in
Structural BiologyConnecting the dots: Computational network analysis
for disease insight and drug repurposing
Nicoleta Siminea1,2, Eugen Czeizler3,2,
Victor-Bogdan Popescu4, Ion Petre5,2 and Andrei P
aun1,2Network biology is a powerful framework for studying the
structure, function, and dynamics of biological systems, offer-
ing insights into the balance between health and disease
states. The field is seeing rapid progress in all of its aspects:
data availability, network synthesis, network analytics, and
impactful applications in medicine and drug development. We
review the most recent and significant results in network
biomedicine, with a focus on the latest data, analytics, software
resources, and applications in medicine. We also discuss what
in our view are the likely directions of impactful development
over the next few years.
Addresses
1 Faculty of Mathematics and Computer Science, University of
Bucharest, Romania
2 National Institute of Research and Development for Biological Sci-
ences, Romania
3 Faculty of Medicine, University of Helsinki, Finland
4 Revvity, Inc., Finland
5 Department of Mathematics and Statistics, University of Turku,
Finland
Corresponding authors: Petre, Ion (ion.petre@utu.fi); P
aun, Andrei
(apaun@fmi.unibuc.ro)Current Opinion in Structural Biology 2024, 88:102881
This review comes from a themed issue on Biophyiscal Methods
(2024): Protein Networks in Health and Disease
Edited by Gabriela Chiosis and Elizabeth A. Komives
For a complete overview see the Issue and the Editorial
Available online xxx
https://doi.org/10.1016/j.sbi.2024.102881
0959-440X/© 2024 The Author(s). Published by Elsevier Ltd. This is an
open access article under the CC BY license (http://creativecommons.
org/licenses/by/4.0/).Introduction
Networks have been successfully applied in biology and
medicine for several decades now [1]. Placed at the
intersection of computational and life sciences, they
offer concepts and techniques from computational
systems biology, bioinformatics, network science, graph
theory, and medicine to analyse complex interactions
within biological systems and their perturbations in
disease states. The advent of advanced technologieswww.sciencedirect.comlike genomics, proteomics, big data analytics, network
analysis, and machine learning has enabled this growth,
allowing for a deeper understanding of diseases at a
molecular level. Network methods play a key role in
personalized and precision medicine, exploring the
possibility of tailoring medical decisions to individuals
based on their unique genetic makeup [2] (Figure 1).
Network biomedicine is evolving fast, both on its
computational side, as well as on its experimental and
applicative side. This is made possible by the wider
availability of more diverse and ever bigger datasets
(Table 1) and by the fast development of computational
tools (Table 2), especially related to artificial intelli-
gence. We review some of the most significant recent
developments in data integration, network synthesis,
network analysis, software, and applications in medi-
cine. We selected the studies in this review to reflect on
the current state of the art on network analytics and its
medical and pharmacological applications. Even with
our focus being just on the literature of the last couple of
years, the amount of new reported research is staggering,
showing the high interest in this research field, and
making this review unavoidably selective. To keep the
review focused, we had to leave out many significant
developments that are related but not right in the
intersection of network science and bio/pharma appli-
cations, including AI-based protein folding and complex
structure identification, drug target and drug synergy
prediction, disease and patient stratification, and many
others. Our goal is to offer practical indications to the
types of datasets that can be used in network studies,
methods of generating a network, types of analyzes that
can be done, and some recent applications.Data resources
The first step in a network project is to clarify the
availability of data and identify the relevant datasets.
There is a variety of datasets that can be used (Table 1).

 Protein data (DPro), which may include identifiers,
synonyms, gene data, variants, organism details, and
links to corresponding entries in external databases.

 Interaction data (DInt), which may include specific
identifiers, type, pathways, identification method,
and degree of confidence.Current Opinion in Structural Biology 2024, 88:102881
Figure 1
Current Opinion in Structural Biology
The flow of a typical network study. Multiple data resources get integrated into a network, either in a top-down fashion by filtering a comprehensive,
standard network, or in a bottom-up fashion, by binding together interactions, pathways, and modules. The network can also be augmented through
machine learning techniques. A multitude of network analysis methods can be applied (topological analysis, systems controllability, network dynamics,
stochastic analysis) to investigate questions around drug repurposing, drug target identification, biomarker identification, drug synergies, mechanisms of
action.
2 Biophyiscal Methods (2024): Protein Networks in Health and Disease
 Disease data (DDis), which may include essential
genes, disease drivers, and common disease-specific
expression abnormalities or mutations.

 Drug data (DDru), which may include drug-targets,
pharmacokinetics, and diseases where a drug might
be used as either main, secondary, or experimental
line of treatment.

 Patient data (DPat), which may include patient-
specific expression abnormalities or mutations, and
previous responses to lines of treatment.Table 1 lists some of the widely used, freely available
resources that offer open APIs and/or web-based data
browsers, or that allow to download their datasets in
standard formats (e.g., JSON, XML, Microsoft Excel, or
text-based comma- or tab-separated values).
The typical challenges that the modeler must handle
concern the integration of data from multiple sources,
the incompleteness of data, and the integration into the
networks of drug targets and of possible disease targets
(such as disease-specific essential proteins).
Data integration, incompleteness
The integration between different data resources is an
important challenge [3], requiring the cross-referencing
of different resource-specific identifiers, with an addi-
tional challenge caused by the lack of standardisation in
identifier types. For example, the oncogene “YES1” is
known, in different databases, as ENSG00000176105
(Ensembl), 7525 (NCBI), 12841 (HGNC), P07946
(UniProt), and IDBG-363.6 (InnateDB). Furthermore,
the data is often incomplete [4], predicted throughCurrent Opinion in Structural Biology 2024, 88:102881homology [5], or obtained through computational, pre-
dictive methods of suboptimal accuracy [6], which rep-
resents an additional challenge for validation [7].Essential proteins
The integration of disease-specific, survivability-essen-
tial proteins [8] into the disease networks is of signifi-
cant interest. Most often obtained through CRISPR-
based knockout, synthetic lethality experiments,
essential genes offer the possibility of targeted therapies
[9]. Their integration into larger interaction networks
and connecting them to available drugs offers the
chance to investigate repurposing available drugs to
target them indirectly through network-induced
cascaded influences [2]. Recent results from the study
by Hasibi et al. [10] expand this line of research by
showing how metabolic fluxes can be used to predict
their essentiality based on graph neural networks.
Drug targets
The integration of drug targets [11,12] into the net-
works allows tracing the effects of drugs through the
network, making it possible to investigate questions
about drug repurposing, drug synergies, optimal drug
combination design, and identifying the action mecha-
nism of drugs in a customised, stratified setup [13].
Several comprehensive drug target datasets with recent
updates are listed in Table 1.Network synthesis
Different types of networks can reveal different insights
into disease [35]: metabolic networks, regulatory net-
works, proteineprotein interaction networks, co-www.sciencedirect.com
Table 1
An annotated collection of data resources: protein data (DPro), interaction data (DInt), disease data (DDis), drug data (DDru), and patient
data (DPat).
Data Resource DPro DInt DDis DDru DPat
BioGRID [14]: protein and genetic interactions. ✓
Advantages: various species, interactions manually annotated, posttranslational
modifications sites.
ChEMBL: bioactive molecules with drug-like properties [15]. ✓ ✓
Advantages: curated data, compounds nomenclature, bioactivity, chemical information.
CORUM [16]: manually annotated data on protein complexes. ✓
Advantages: manually annotated protein complexes.
Limitations: last release one year and a half ago; data can only be downloaded directly,
no specific API available.
CTD [17]: interactions between genes or proteins and chemicals or diseases. ✓ ✓ ✓
Advantages: contains analysis tools, toxicogenomics data.
Limitations: free noncommercial use, commercially licensed.
DisGeNET [18]: drugs, drug-targets, and related pharmaceuticals. ✓
Advantages: it integrates data from various sources, has a separate section for curated
data.
Limitations: only for noncommercial use; for commercial use, DISGENET Plus is
offered; last update two and a half years ago.
DrugBank [11]: drugs, drug-targets, and related pharmaceuticals. ✓ ✓ ✓
Advantages: contains spectral data.
Limitations: free for academic purpose, fee-based for commercial use.
Ensembl [19]: genome browser, gene and protein data. ✓
Advantages: automatically annotated genome.
Limitations: only small amounts of data can be directly downloaded, while large amounts
can be accessed via API.
Genomic Data Commons [20]: cancer progression and therapy response. ✓ ✓
Advantages: very large dataset, genetic data obtained by various methods, tumor
images, patient clinical data.
Limitations: not all data are public because there can be sensitive information.
HGNC [21]: human genes and related nomenclature. ✓
Advantages: comprehensive resource showing multiple names for a gene, links to
orthologs, clinical resources.
Human Interactome Atlas [22]: binary protein interactions. ✓
Advantages: human protein interactions identified by high throughput yeast two-hybrid
screens.
Limitations: provided by only direct download; last data update 4 years ago.
InnateDB [23]: interactions involved in the innate immune response. ✓ ✓
Advantages: has a separate section for interactions validated experimentally; bovine
data in addition to human and mouse data.
Limitations: not clear when last updated.
IntAct [24]: curated and user-submitted data on interactions. ✓
Advantages: experimental data from articles published or submitted for publication in a
peer reviewed journal; interactions are grouped by diseases.
KEGG [25]: high-level functions and utilities of biological systems. ✓ ✓ ✓ ✓
Advantages: visual diagrams with pathways.
Limitations: user licence required for commercial purposes.
MINT [26]: experimentally verified data on interactions. ✓ ✓
Advantages: open source; interactions verified experimentally; identified by expert
curators; it provides links to other resources; data for different species: Homo sapiens,
Mus musculus, Drosophila Melanogaster, Saccharomyces Cerevisiae.
Limitations: last update 5 years ago.
NCBI Gene [27]: gene nomenclature and related information from multiple sources. ✓
Advantages: genes from different species can be searched; links to orthologs, to
BLAST, and to Genome Data Viewer; information about lineage, official full name, a
comprehensive description among others.
NetControl4BioMed [28]: aggregated data on proteins and interactions. ✓ ✓ ✓
Advantages: data from different sources can be combined; the possibility of creating and
analyzing the control over a network.
OGEE [8]: gene essentiality and cancer cell-lines. ✓ ✓
(continued on next page)
Connecting the dots Siminea et al. 3
www.sciencedirect.com Current Opinion in Structural Biology 2024, 88:102881
Table 1 (continued )
Data Resource DPro DInt DDis DDru DPat
Advantages: genes are categorized according to tissue, cell line, detection method.
Limitations: last updated 3 years ago.
Open Targets [29]: drugs and drug-targets. ✓ ✓
Advantages: integrates data from various sources; can view the targets associated with
each disease, association score or a prioritization score.
PharmGKB [30]: genetic drug responses. ✓
Advantages: data manually curated, but also extracted with an automated natural
language processing tool.
Reactome [31]: curated data on interactions and pathways. ✓ ✓ ✓ ✓
Advantages: visual interface; can be zoomed in; own genes can be uploaded for
analysis.
SIGNOR [32]: annotated data on interactions and pathways. ✓
Advantages: causal interactions data on humans, mouses and rats; relevance score for
each interaction; disease browser; pathway browser; advanced browser.
STRING [5]: aggregated data on physical interactions and functional associations. ✓ ✓
Advantages: search by protein id, by amino acid sequence, or even by disease; you can
analyze data within the platform and you can export the results.
UniProt [33]: annotated data on proteins and protein sequences. ✓
Advantages: detailed information and links to relevant databases.
WikiPathways [34]: biological pathways. ✓ ✓ ✓
Advantages: different applications to import the data, visualize and analyze it.
4 Biophyiscal Methods (2024): Protein Networks in Health and Diseaseexpression networks, signaling networks, or combina-
tions thereof. There are two approaches to constructing
a network [35].

 bottom-up, de-novo building the network around
some key nodes of interest, or

 top-down, trimming a comprehensive interaction
network (e.g., the Human Interactome [22]) to focus
it on an area of interest, such as a certain pathway or
tissue.Either way, the modeler has some key decisions to make
on how to acquire or to filter the data, and how to
augment the data through predictions for missing links.
Data acquisition/filtering
The interaction data can be acquired from general
repositories (Table 1), from disease-specific resources,
from cell-line-specific resources, or even from the
sample itself (if applicable). A major challenge here is
that the sources of the interaction data are heteroge-
neous and occasionally even unspecified [5], bringing
different levels of validation and trust in their validity:
experimentally validated data (from the specific
sample being studied or, less suitable, from related
assays), data from in-vitro experiments (including
those done on other organisms or on cell lines), or data
inferred from in-silico studies. Another challenge is
the selection and the integration of the relevant types
of interactions [31]: co-localization, physical associa-
tion, co-expression, activation, inhibition, post-
translational modification, and many others.Current Opinion in Structural Biology 2024, 88:102881Additionally, some of these functions are compart-
ment- and tissue-specific [22]. Some of these are
relevant in dynamical, quantitative studies on systems
behaviour, while others are useful in binding associa-
tions and regulatory studies. Such details are typically
more difficult to clarify in the top-down approach,
where they are already integrated in a ready-made
comprehensive network.
Link prediction
There is a flurry of recent results on link predictions,
many of them based on machine learning. Graph neural
networks are a popular method to predict missing in-
teractions between different node types, such as a drug
and its targets [36], or to learn the structural patterns of
the proteineprotein interactions [37]. Graph convolu-
tional networks and graph attention networks were used
in the study by Sun et al. [38] to identify
metaboliteedisease associations, and in the study by
Wang et al. [39] to identify interactions between long
noncoding RNA (lncRNA) and microRNA (miRNA). An
attention-based link prediction model that uses gene
ontology (GO) terms together with network topology to
predict protein function was developed in the study by
Zhang et al. [40]. A proteineprotein interaction network
was extended in the study by Gysi et al. [41] with
noncoding interactions, increasing the number of in-
teractions by a staggering 107% and its ability to detect
diseaseedisease relationships that could not be seen
before. An example of a unusual network was in the study
by Guthrie et al.. [42], where the authors integrated a set
of innate immune errors into the network with the aim of
identifying the peculiarities of rare diseases.www.sciencedirect.com
Table 2
A collection of recent software tools: network generation software (NGen), network visualization software (NVis), network analysis
software (NAna).
Software Tool NGen NVis NAna
CytoscapeJS [78] is an open-source JavaScript library for the analysis and visualization of complex
networks, supporting modern web browsers and Node.js.
✓ ✓
Advantages: widespread usage; can run both in-browser and on-server; allows extensions;
includes multiple graph theory algorithms;
Disadvantages: JavaScript is out-performed by many other languages; performance depends on
device when used in-browser; does not scale well with large networks; uses only JSON file format;
Graphia [79] is an open-source desktop application for the visualisation, analysis, and interpretation
of large and complex datasets.
✓ ✓
Advantages: does not require coding; accepts multiple standard file formats; fast and fluid
rendering for large networks after the initial layout; can build tables and charts for analysing
network data;
Disadvantages: larger initial layout time; performance depends on device; no graph theory analysis
algorithms;
iGraph [80] is an open-source C/C++/Mathematica/Python/R library for the analysis of networks,
designed for efficiency and ease of use.
✓ ✓
Advantages: is highly configurable; allows for random graph generation; includes multiple graph
theory algorithms; is very suitable for analysing alread-built networks; integrates with external
plotting libraries;
Disadvantages: different release schedules for the different languages that it supports; can layout a
graph, but requires an external library to plot it; can handle multiple data formats, but some
standard ones are missing (e.g., JSON or Cytoscape CX); has performance issues when dealing
with dynamic networks;
NDEx IQuery [81] is a web application for the generation of networks, integrating with multiple
external bioinformatics databases.
✓
Advantages: is focused on bioinformatics; contains extensive data to be used in network
generation; integrates closely with Cytoscape; server-based, does not depend on device
performance;
Disadvantages: provides only basic subnetwork building algorithms; does not allow for direct
network analysis; no support for additional file formats; offers accounts with limited, albeit
generous, storage space;
NetControl4BioMed [28] is an open-source web application for the generation and controllability
analysis of networks, integrating with multiple external bioinformatics databases.
✓ ✓
Advantages: contains extensive data to be used in network generation; provides several algorithms
for network generation and analysis; server-based, does not depend on device performance;
accepts multiple standard file formats;
Disadvantages: limited in-app visualisation and interactivity; data is automatically deleted after
three months of inactivity; only allows for controllability network analysis;
Retina [82] is an open-source web application for the online sharing of graph or network
visualisations, without a server requirement.
✓
Advantages: no download, no installation, no coding required; server-based, performance does not
depend on device; allows direct importing of online-available networks;
Disadvantages: accepts only GEXF and GraphML file formats; allows only for network
visualisation, no network interactivity; limited configuration options, without any layout
configuration;
Connecting the dots Siminea et al. 5Network analysis
The diversity of network analysis methods is stag-
gering and the field has developed tremendously in
the last couple of years. There are three main cate-
gories of methods: topological analysis, with the main
topic being to identify key nodes, modules, commu-
nities and structural patterns in the network; systems
analysis and control, concerned with questions aroundwww.sciencedirect.comthe network dynamics, emerging behaviour and
influencing/controlling the functioning of the
network; machine learning and deep learning,
exploiting the paradigm of data acquisition/genera-
tion/augmentation and automatic learning of patterns
to answer a variety of questions such as target and
link prediction. We discuss recent progress in each of
these categories.Current Opinion in Structural Biology 2024, 88:102881
6 Biophyiscal Methods (2024): Protein Networks in Health and DiseaseTopological analysis
Much of the recent network analysis research is focused
on identifying important/driver/essential nodes or
modules within the studied networks. One category of
methods is based on probabilistic analysis; we discuss
here two such studies. The first, proposed in the study
byMeng et al. [43], uses a two-stage random walk was to
discover driver genes: identify subnetworks with a high
correlation with the seed nodes and then rank possible
driver nodes. The second [44], identifies influential
nodes based on a local propagation probability model,
which ranks nodes based on total comprehensive scores
of neighbor nodes within their three level neighborhood.
Another category of methods is based on scoring and
ranking proteins to predict their essentiality; a new such
study is in Ref. [45], based on a modified Jaccard coef-
ficient. Yet another category of methods is on system-
level approaches that associate essentiality to the
concept of graph domination; a new such study is in
Ref. [46], based on a new heuristic for the group cen-
trality problem, a well-known variant of the graph
dominating problem.
Systems control
This topic, well studied over several decades, has also
seen significant progress in the last two years. The most
common approaches are based on the minimum domi-
nating set (MDS) problem. A recent new study [47]
proposes a new approach to identify the unavoidable
nodes in the many solutions that the MDS problem may
have. Another interesting proposal is in the study by
Wong et al. [48], offering a solution to theMDS problem
based on a biomolecular and quantum algorithm that
was run on both IBM Quantum’s QASM simulator and
the Brooklyn superconducting quantum device. Another
method that answers the same question in Boolean
networks was presented in the study by Parmer et al.
[49]. One other concept to address this problem is the
intermittent nodes, with a new criticality metric for
them proposed in the study by Someya et al. [50]. The
related topic of network controllability has also received
new results based on the probabilistic adaptive attack
model [51], or on Boolean network models [52].
Machine learning for network analysis
This is one of the fastest growing research direction for
network analysis. Graph neural networks and graph
convolutional networks are the most common architec-
ture in use, combining the power of deep learning with
the idea of customising the network connections based
on interaction data. One way to combine them is
demonstrated in the study by Wang et al. [53] that in-
troduces a graph neural network to infer gene regulatory
networks from RNA-seq data, and then applies network
control theory to identify key nodes. The study in the
study by Peng et al. [54] does the combination in the
other order: it first constructs a geneesampleCurrent Opinion in Structural Biology 2024, 88:102881association matrix and then it applies graph convolu-
tional networks to eventually predict personalised driver
genes for individual samples. Similarly, in the study by
Duan et al. [55] the authors start from a set of
proteineprotein interaction (PPI) networks in order to
identify low-dimensional node representations, and use
them to generate de-novo driver gene prioritizations for
new gene embedding. Another approach is introduced in
the study by Wu et al. [56], that uses graph neural
networks to learn drugeprotein association networks.
An attention-based deep learning model was developed
in the study by Theodoris et al. [57] for predicting key
gene regulators or therapeutic targets. Generative AI has
also been tested for network analysis, a recent example
being that of [58], that uses generative adversarial
networks (GAN) to identify protein complexes from
within PPI networks. Machine learning has been suc-
cessfully applied to identify biomarkers, for example in
papillary thyroid carcinoma [59], in DNA methylation in
Graves’ orbitopathy/thyroid associated orbitopathy [60],
in colon cancer [61], in stomach adenocarcinoma [62],
or in osteoarthritis starting from mRNA osteoarthritis
datasets [63].Applications in medicine
Precision medicine is one of the fields in which network
science has been influential and holds promises for
further impactful results, in disease network analysis,
drug target identification and drug repurposing, precision
oncology, biomarker discovery, bacterial resistance, clin-
ical decision support systems, patient stratification and
risk prediction. Integrating disease-specific data with
patient-specific information into comprehensive inter-
action networks makes it possible to ask questions about
personalised optimal therapies and drug repurposing.
Drug repurposing
On drug repurposing there are several recent network
studies. In the study by Bai et al. [64] bortezomib was
identified as a potential drug in nonsmall cell lung
cancer, using a novel network analysis method, based on
difference analysis, Spearman correlation, biological
function analysis and Bayesian causality analysis. In the
study by Chirom et al. [65] the authors found that 4
drugs (uprosertib, progesterone, solitomab and regorafenib)
have strong interactions with key regulators in ovarian
cancer. The study was based on the human interactome
network and the roles of hubs in it in the disease states.
While uprosetib is still under investigation for different
types of cancer, regorafenib was already approved for
metastatic colorectal cancer, advanced gastrointestinal
tumors and hepatocellular carcinoma. Selecting drugs
and drug combinations with optimal effect was investi-
gated in the study by Ahmed et al. [66] by integrating an
HPV-induced cervical cancer network with drug target
interaction data, followed by a drug target network
proximity analysis.www.sciencedirect.com
Connecting the dots Siminea et al. 7Drug target identification
On drug target identification recent progress includes
[67] where a network was created to find out how far the
targets of various traditional Chinese herbs are from the
module of symptoms they relieve. In the study by
Dwivedi et al. [68] the authors use the DisGeNET
database to map breast cancer targets to those of sinapic
acid using a diverse combination of methods: KEGG to
view disease pathways, and Cytoscape to visualize them,
then apply molecular docking, molecular dynamics sim-
ulations, molecular mechanics Poisson-Boltzmann sur-
face area analysis, together with cluster analysis and
GO analysis.
Network hub identification
Identifying hubs is one of the central topics of research in
network science in biomedicine, as hubs tend to be asso-
ciated with key regulators and disease drivers. This line of
research continues to be very active. In the study by Zhai
et al. [69] a study into network hubs was used to identify
dequalinium, a drug that may decrease the impact of
Escherichia coli on mutagenesis after fluoroquinolone admin-
istration. E3 ligase UBR5 was pinpointed in the study by
Mark et al. [70] as a signaling hub that helps degrade
unpaired subunits. Metabolic drivers and signature clus-
ters in Alzheimer’s disease were identified in the study by
Chang et al. [71] and RE1-silencing transcription factor
was found in the study byCheng et al. [72] as an upstream
suppressor involved in axon regeneration.
Module identification
Modularity and community structure was successfully
applied in the study by Shah et al. [73] to identify 5
modules associated with heart failure risk based on
protein co-regulation networks, in the study by
Humphrey et al. [74] where gene co-expression net-
works were exploited to identify activated microglia
modules in amyotrophic lateral sclerosis, and in the
study by van Oostrum et al. [75] where shared synaptic
protein modules were identified to pinpoint the prote-
omic areas involved in synapse specialization. One step
further was taken in the study by Berry et al. [76] where
the subtypes of Sjogren’s syndrome were mapped onto
the identified modules, and different therapeutic tar-
gets for each subtype were identified. In the study by
Kong et al. [77] the authors predicted the response of
patients with melanoma, gastric cancer, or bladder
cancer at immune checkpoint inhibitors through a
network-based machine learning model.Software
There are many software tools available for aspects
related to the synthesis, analysis, and visualization of
proteineprotein interaction network. An overview of
software developed or updated in the last two of years,
and of particularly significant older software can be
found in Table 2. We focused on freely available andwww.sciencedirect.comopen-source resources that accept, parse, and output
data in standard formats (e.g., Cytoscape, JSON, XML,
Microsoft Excel, or text-based comma- or tab-separated
values for network generation, or PDF, PNG, or SVG for
network visualization). The availability of import/export
in standard formats makes it versatile to combine these
tools for more complex analyses. We categorise the
software based on the following three classes.

 Network generation software (NGen), which allows
for the generation of PPI networks, starting from
custom lists of proteins, using different generation
algorithms, and integrating multiple data resources.

 Network visualization software (NVis), which allows
for the visualization of PPI networks, using different
layout algorithms, and offering static or dynamic on-
screen graphic rendering.

 Network analysis software (NAna), which allows for
the analysis of PPI networks, using different analysis
measures and algorithms (e.g., topographical or
controllability analysis).We discuss now in more details the most significant
recent tools for network inference and visualisation.
LEVELNET [83] is a recent, versatile tool for inte-
grating and exploring proteineprotein interaction net-
works synthesized from multiple data sources.
LEVELNET represents networks as multilayered
graphs, offering the possibility to zoom in and compare
their subnetworks. Starting from a set of proteins with
known structures, this tool builds a grid of networks for
each data source, offering different views on the in-
teractions. Interactions supported by multiple evidence
are represented as multiedges in the aggregate graph.
Also, each edge is tagged with a weight to reflect the
reliability of the evidence supporting that edge.
LEVELNET facilitates the identification of potential
inconsistencies between different data sources.
The Cytoscape stringApp has been improved with anal-
ysis and visualisation for heterogeneous networks [84].
A visualisation program is also GraphBio [85], which
provides visualisation options for a large palette of ana-
lyses, such as text group analysis, principal component
analysis, correlation analysis among others. An R package
built for cellecell signaling network analysis and visu-
alisation is presented in the study by Raredon et al. [86].
Obviously, neural network models have also materialized
for the visualisation of networks, one of them being
NeuroDAVIS [87], which is a neural network model that
can be used for the visualisation of large data sets as
well. Another novelty is the use of the Unity engine for
the visualisation of the front-end network simulation
exemplified by NetVision [88]. OpenPIP is an open
source platform for the analysis and visualisation ofCurrent Opinion in Structural Biology 2024, 88:102881
8 Biophyiscal Methods (2024): Protein Networks in Health and Diseaseproteineprotein interactions that is based on Cyto-
scape. js for network visualisation [89]. Another built
platform for analysing and visualising omni data is
presented in the study by Koh et al. [90], where the
authors also provide the code so that users can change it
if they need something else similar.
PathwayNexus [91] helps to observe relationships be-
tween metabolites and compounds as matrices, but also
to group, sort, visualise pathways. DeepMAPS [92] is a
tool for biological network inference from single-cell
multiomics data. The networks are represented as het-
erogeneous graphs and analysed with a multihead graph
transformer. The tool can be used through the Deep-
MAPS webserver.Perspectives
Network biology, although a relatively young scientific
field, has already offered many impactful applications in
medicine and drug development [1,93,94]. The field is
actively evolving and we reviewed some of the most
significant results published in the last two years. Based
on these recent results, we foresee several research di-
rections that are likely to receive attention in the next
few years and we anticipate that they will offer signifi-
cant new developments in this field. We discuss them
shortly here.
Machine learning has found deep applications in a myriad
of fields in science and outside of it. This includes many
exciting applications in biology and medicine, including
network biomedicine [36,57]. Still, we believe that even
more breakthroughs will be made when machine learning
becomes integrated in a deeper way in network-based
studies. The current applications in network studies are
mostly focused on graph representation learning [55],
that facilitates an easier identification of nodes, com-
munities, and patterns. To a large extent, this does not
yet take fully into account the global structure of the
network as a model for how signals and information are
distributed and influence each other. The combination of
the explainability brought forward by the structural in-
sights and mechanisms encoded into networks, and of
the de-novo, data-driven ability of machine learning to
discover hidden patterns with complex sources remains
yet to be convincingly demonstrated. For example, the
incomplete nature of the data [4], the noise in it, the
uncertainty in its validity [7], the search for system
controls [50] and other global patterns are all topics that
can be addressed with machine learning. To some extent,
difficult computational problems that have hindered
network science (for example, graphmatching), may start
to be addressed with machine learning. Generative graph
models have started to be investigated, but they are still
in their infancy, and have not yet seen convincing proofs
of concepts in network synthesis. Deep neural networksCurrent Opinion in Structural Biology 2024, 88:102881already brought forward several exciting new approaches:
node classification using few-shot learning [95];
proteineprotein interaction prediction using a multi-
graph neural network [96], a learning hierarchical graph
[97], or a graph neural network [98]; drug-target pre-
diction using a heterogeneous biological network [99], or
a metapath-aggregated heterogeneous graph neural
network [100]; drugedisease association prediction using
graph neural networks on multiple biological relation-
ships [101]; proteineligand affinity identification using
graph neural networks based on 3D structures and
physical interactions [102]; drug repurposing with graph
neural networks [103]; drug discovery with graph atten-
tion mechanisms [104].
Breakthroughs in protein structure identification using
AlphaFold [105] have also brought significant new re-
sults in network studies. Significant recent results
include predicting proteineligand binding affinity
[106]; designing protein sequences that can fit into a
specific desired structure [107]; protein dynamics based
on residue flexibility [108]; predicting multimeric pro-
tein complex structures; predicting proteineprotein
interactions [109], visualising the quality of predicted
protein complexes [110]; drug repurposing based on the
structural similarity of protein interfaces [111].
A research direction that we believe to be on the brink of
breakthroughs is that of personalised medicine [30,54].
It has been clear for some time now that networks allow
for the personalised integration of data and networks to
include aging, medical background, genetic data, life-
style, and environmental conditions. Adding these to
disease-, biological- and drug-data offers an insightful,
patient- and disease-specific insight into the personal-
ised mechanisms of disease, that may be analysed to
extract suggestions for optimal drug combinations and
treatment planning.Funding
This work was partially supported by the Core Program
within the Romanian National Research, Development
and Innovation Plan 2022e2027, carried out with the
support of MRID, project no. 23020101(SIA-PRO),
contract no 7N/2022, and project PNRR-I8 no.
842027778., contract no 760096.CRediT authorship contribution statement
Nicoleta Siminea: Writing e original draft, Writing e
review & editing. Eugen Czeizler: Writing e original
draft, Funding acquisition. Victor-Bogdan Popescu:
Writinge original draft. Ion Petre: Study design, Writing
e review & editing, Funding acquisition. Andrei P
aun:
Study Design, Writing e review & editing, Fund-
ing acquisition.www.sciencedirect.com
Connecting the dots Siminea et al. 9Declaration of competing interest
The authors declare that they have no known competing
financial interests or personal relationships that could
have appeared to influence the work reported in
this paper.Data availability
No data were used for the research described in
the article.References
Papers of particular interest, published within the period of review,
have been highlighted as:
* of special interest
* * of outstanding interest
1. Pandey AK, Loscalzo J: Network medicine: an approach to
complex kidney disease phenotypes. Nat Rev Nephrol 2023,
19:463–475, https://doi.org/10.1038/s41581-023-00705-0.
2. Nushi E, Popescu V-B, Sanchez Martin J-A, Ivanov S, Czeizler E,
Petre I: Network modelling methods for precision medicine. John
Wiley & Sons, Ltd; 2022:363–423, https://doi.org/10.1002/
9781119716600.ch10.
3. Wang X, Jing X, Dou F, Cao H: An approach for proteins and
their encoding genes synonyms integration based on pro-
tein ontology. BMC Bioinf 2023, 24:339, https://doi.org/10.1186/
s12859-023-05464-0.
4. Li Y, Sun C, Wei J-M, Liu J: Drug-Protein interaction prediction
by correcting the effect of incomplete information in het-
erogeneous information. Bioinformatics 2022, 38:5073–5080,
https://doi.org/10.1093/bioinformatics/btac629.
5. Szklarczyk D, Kirsch R, Koutrouli M, Nastou K, Mehryary F,
Hachilif R, Gable AL, Fang T, Doncheva NT, Pyysalo S, et al.:
The STRING database in 2023: protein-protein association
networks and functional enrichment analyses for any
sequenced genome of interest. Nucleic Acids Res 2023, 51:
D638–D646, https://doi.org/10.1093/nar/gkac1000.
6. Tsuchiya Y, Yamamori Y, Tomii K: Protein-protein interaction
prediction methods: from docking-based to AI-based ap-
proaches. Biophysical Reviews 2022, 14:1341–1348, https://
doi.org/10.1007/s12551-022-01032-7.
7. Kewalramani N, Emili A, Crovella M: State-of-the-art computa-
tional methods to predict protein-protein interactions with
high accuracy and coverage. Proteomics 2023, 23, 2200292,
https://doi.org/10.1002/pmic.202200292.
8. Gurumayum S, Jiang P, Hao X, Campos TL, Young ND,
Korhonen PK, Gasser RB, Bork P, Zhao X-M, He L-j, Chen W-H:
OGEE v3: online GEne Essentiality database with increased
coverage of organisms and human cell lines. Nucleic Acids
Res 2020, 49:D998–D1003, https://doi.org/10.1093/nar/
gkaa884.
9. Funk L, Su K-C, Ly J, Feldman D, Singh A, Moodie B,
Blainey PC, Cheeseman IM: The phenotypic landscape of
essential human genes. Cell 2022, 185:4634–4653.e22,
https://doi.org/10.1016/j.cell.2022.10.017.
10
* *
. Hasibi R, Michoel T, Oyarzún DA: Integration of graph neural
networks and genome-scale metabolic models for predicting
gene essentiality. npj Systems Biology and Applications 2024,
10:24, https://doi.org/10.1038/s41540-024-00348-2.
It uses a graph where nodes are reactions, edges are shared reactants
or products, node features are generated based on the graph structure
and adjacent edge weights, and the learning mechanism is based on a
message passing scheme, all to predict essential genes.
11
* *
. Knox C, Wilson M, Klinger CM, Franklin M, Oler E, Wilson A,
Pon A, Cox J, Chin NEL, Strawbridge SA, et al.: DrugBank 6.0:
the DrugBank knowledgebase for 2024. Nucleic Acids Res
2023, 52:D1265–D1275, https://doi.org/10.1093/nar/gkad976.www.sciencedirect.comIts latest version has not only increased the number of drugs and for
each drug, the number of drug interactions, new indications, if any, etc.,
but also includes new sections such as spectral data.
12. Zhou Y, Zhang Y, Lian X, Li F, Wang C, Zhu F, Qiu Y, Chen Y:
Therapeutic target database update 2022: facilitating drug
discovery with enriched comparative data of targeted agents.
Nucleic Acids Res 2021, 50:D1398–D1407, https://doi.org/
10.1093/nar/gkab953.
13. Trapotsi Maria-Anna BA: Hosseini-Gerami Layla, Computa-
tional analyses of mechanism of action (MoA): data, methods
and integration. RSC Chem Biol 2021, 3:170–200, https://
doi.org/10.1039/d1cb00069a.
14. Oughtred R, Rust J, Chang C, Breitkreutz B-J, Stark C,
Willems A, Boucher L, Leung G, Kolas N, Zhang F, et al.: The
BioGRID database: a comprehensive biomedical resource of
curated protein, genetic, and chemical interactions. Protein
Sci 2021, 30:187–200, https://doi.org/10.1002/pro.3978.
15. Zdrazil B, Felix E, Hunter F, Manners EJ, Blackshaw J, Corbett S,
de Veij M, Ioannidis H, Lopez DM, Mosquera JF, et al.: The
ChEMBL Database in 2023: a drug discovery platform span-
ning multiple bioactivity data types and time periods. Nucleic
Acids Res 2023, 52:D1180–D1192, https://doi.org/10.1093/nar/
gkad1004.
16. Tsitsiridis G, Steinkamp R, Giurgiu M, Brauner B, Fobo G,
Frishman G, Montrone C, Ruepp A: CORUM: the comprehen-
sive resource of mammalian protein complexes-2022. Nucleic
Acids Res 2022, 51:D539–D545, https://doi.org/10.1093/nar/
gkac1015.
17
*
. Davis AP, Wiegers TC, Johnson RJ, Sciaky D, Wiegers J,
Mattingly CJ: Comparative toxicogenomics database (CTD):
update 2023. Nucleic Acids Res 2022, 51:D1257–D1262, https://
doi.org/10.1093/nar/gkac833.
The database includes now a new feature that can connect chemicals,
genes, phenotypes and diseases.
18. Piñero J, Ramírez-Anguita JM, Saüch-Pitarch J, Ronzano F,
Centeno E, Sanz F, Furlong LI: The DisGeNET knowledge
platform for disease genomics: 2019 update. Nucleic Acids
Res 2019, 48:D845–D855, https://doi.org/10.1093/nar/gkz1021.
19. Martin FJ, Amode MR, Aneja A, Austine-Orimoloye O, Azov AG,
Barnes I, Becker A, Bennett R, Berry A, Bhai J, et al.: Ensembl
2023. Nucleic Acids Res 2022, 51:D933–D941, https://doi.org/
10.1093/nar/gkac958.
20. Heath AP, Ferretti V, Agrawal S, An M, Angelakos JC, Arya R,
Bajari R, Baqar B, Barnowski JHB, Burt J, et al.: The NCI
genomic data commons. Nat Genet 2021, 53:257–262, https://
doi.org/10.1038/s41588-021-00791-5.
21. Seal RL, Braschi B, Gray K, Jones TEM, Tweedie S, Haim-
Vilmovsky L, Bruford EA: Genenames.org: the HGNC re-
sources in 2023. Nucleic Acids Res 2022, 51:D1003–D1009,
https://doi.org/10.1093/nar/gkac888.
22. Luck K, Kim D-K, Lambourne L, Spirohn K, Begg BE, Bian W,
Brignall R, Cafarelli T, Campos-Laborie FJ, Charloteaux B, et al.:
A reference map of the human binary protein interactome.
Nature 2020, 580:402–408, https://doi.org/10.1038/s41586-020-
2188-x.
23. Breuer K, Foroushani AK, Laird MR, Chen C, Sribnaia A, Lo R,
Winsor GL, Hancock REW, Brinkman FSL, Lynn DJ: InnateDB:
systems biology of innate immunity and beyond-recent up-
dates and continuing curation. Nucleic Acids Res 2012, 41:
D1228–D1233, https://doi.org/10.1093/nar/gks1147.
24. del Toro N, Shrivastava A, Ragueneau E, Meldal B, Combe C,
Barrera E, Perfetto L, How K, Ratan P, Shirodkar G, et al.: The
IntAct database: efficient access to fine-grained molecular
interaction data. Nucleic Acids Res 2021, 50:D648–D653,
https://doi.org/10.1093/nar/gkab1006.
25. Kanehisa M, Sato Y, Kawashima M: KEGG mapping tools for
uncovering hidden features in biological data. Protein Sci
2022, 31:47–53, https://doi.org/10.1002/pro.4172.
26. Calderone A, Iannuccelli M, Peluso D, Licata L: Using the MINT
database to search protein interactions. Current Protocols in
Bioinformatics 2020, 69:e93, https://doi.org/10.1002/cpbi.93.Current Opinion in Structural Biology 2024, 88:102881
10 Biophyiscal Methods (2024): Protein Networks in Health and Disease27. Brown GR, Hem V, Katz KS, Ovetsky M, Wallin C,
Ermolaeva O, Tolstoy I, Tatusova T, Pruitt KD, Maglott DR,
Murphy TD: Gene: a gene-centered information resource at
NCBI. Nucleic Acids Res 2014, 43:D36–D42, https://doi.org/
10.1093/nar/gku1055.
28. Popescu V-B, Sánchez-Martín JA, Schacherer D, Safadoust S,
Majidi N, Andronescu A, Nedea A, Ion D, Mititelu E, Czeizler E,
Petre I: NetControl4BioMed: a web-based platform for
controllability analysis of protein-protein interaction net-
works. Bioinformatics 2021, 37:3976–3978, https://doi.org/
10.1093/bioinformatics/btab570.
29. Ochoa D, Hercules A, Carmona M, Suveges D, Baker J,
Malangone C, Lopez I, Miranda A, Cruz-Castillo C, Fumis L, et al.:
The next-generation Open Targets Platform: reimagined,
redesigned, rebuilt. Nucleic Acids Res 2022, 51:D1353–D1359,
https://doi.org/10.1093/nar/gkac1046.
30. Whirl-Carrillo M, Huddart R, Gong L, Sangkuhl K, Thorn CF,
Whaley R, Klein TE: An evidence-based framework for eval-
uating pharmacogenomics knowledge for personalized
medicine. Clin Pharmacol Therapeut 2021, 110:563–572, https://
doi.org/10.1002/cpt.2350.
31. Milacic M, Beavers D, Conley P, Gong C, Gillespie M, Griss J,
Haw R, Jassal B, Matthews L, May B, et al.: The reactome
pathway knowledgebase 2024. Nucleic Acids Res 2023, 52:
D672–D678, https://doi.org/10.1093/nar/gkad1025.
32. Lo Surdo P, Iannuccelli M, Contino S, Castagnoli L, Licata L,
Cesareni G, Perfetto L: SIGNOR 3.0, the SIGnaling network
open resource 3.0: 2022 update. Nucleic Acids Res 2022, 51:
D631–D637, https://doi.org/10.1093/nar/gkac883.
33
* *
. Consortium TU: UniProt: the universal protein knowledgebase
in 2023. Nucleic Acids Res 2022, 51:D523–D531, https://doi.org/
10.1093/nar/gkac1052.
The latest version contains structure prediction using Alpha Fold, which
is a Deep Mind product, for more than 85% proteins.
34. Agrawal A, Balci H, Hanspers K, Coort SL, Martens M,
Slenter DN, Ehrhart F, Digles D, Waagmeester A, Wassink I,
et al.:WikiPathways 2024: next generation pathway database.
Nucleic Acids Res 2023, 52:D679–D689, https://doi.org/10.1093/
nar/gkad960.
35. Galindez G, Sadegh S, Baumbach J, Kacprowski T, List M:
Network-based approaches for modeling disease regulation
and progression. Comput Struct Biotechnol J 2022, 21:
780–795, https://doi.org/10.1016/j.csbj.2022.12.022.
36
*
. Veleiro U, de la Fuente J, Serrano G, Pizurica M, Casals M,
Pineda-Lucena A, Vicent S, Ochoa I, Gevaert O, Hernaez M:
GeNNius: an ultrafast drug-target interaction
inference method based on graph neural networks. Bioinfor-
matics 2023, 40, btad774, https://doi.org/10.1093/bioinformatics/
btad774.
Identify drug-target connections as predicted by a neural network
containing SAGEConv layers.
37. Huang Y, Wuchty S, Zhou Y, Zhang Z: SGPPI: structure-aware
prediction of protein-protein interactions in rigorous condi-
tions with graph convolutional network. Briefings Bioinf 2023,
24, bbad020, https://doi.org/10.1093/bib/bbad020.
38
*
. Sun F, Sun J, Zhao Q: A deep learning method for
predicting metabolite-disease associations via graph neural
network. Briefings Bioinf 2022, 23, bbac266, https://doi.org/
10.1093/bib/bbac266.
Prediction of metabolite–disease associations using a graph neural
network, but also similarities of metabolites, of diseases and hetero-
geneous networks.
39
*
. Wang W, Zhang L, Sun J, Zhao Q, Shuai J: Predicting the po-
tential human lncRNA-miRNA interactions based on graph
convolution network with conditional random field. Briefings
Bioinf 2022, 23:bbac463, https://doi.org/10.1093/bib/bbac463.
GCN-based lncRNA-miRNA prediction, whose embeddings are
updated relative to a conditional random field, and weights are updated
as a result of an attention mechanism.
40. Zhang X, Guo H, Zhang F, Wang X, Wu K, Qiu S, Liu B, Wang Y,
Hu Y, Li J: HNetGO: protein function prediction via hetero-
geneous network transformer. Briefings Bioinf 2023, 24,
bbab556, https://doi.org/10.1093/bib/bbab556.Current Opinion in Structural Biology 2024, 88:10288141. Gysi DM, Barabási A-L: Noncoding RNAs improve the pre-
dictive power of network medicine. Proc Natl Acad Sci USA
2023, 120, e2301342120, https://doi.org/10.1073/
pnas.2301342120.
42. Guthrie J, Bal SK, Lombardo SD, Müller F, Sin C, Hütter CV,
Menche J, Boztug K: AutoCore: a network-based definition of
the core module of human autoimmunity and auto-
inflammation. Sci Adv 2023, 9, eadg6375, https://doi.org/
10.1126/sciadv.adg6375.
43. Meng P, Wang G, Guo H, Jiang T: Identifying cancer driver
genes using a two-stage random walk with restart on a gene
interaction network. Comput Biol Med 2023, 158, 106810,
https://doi.org/10.1016/j.compbiomed.2023.106810.
44. Xu G, Meng L: A novel algorithm for identifying influential
nodes in complex networks based on local propagation
probability model. Chaos, Solit Fractals 2023, 168, 113155,
https://doi.org/10.1016/j.chaos.2023.113155.
45. Payra AK, Saha B, Ghosh A: MM-CCNB: essential protein
prediction using MAX-MIN strategies and compartment of
common neighboring approach. Comput Methods Progr
Biomed 2023, 228, 107247, https://doi.org/10.1016/
j.cmpb.2022.107247.
46. Staus LP, Komusiewicz C, Morawietz N, Sommer F: Exact al-
gorithms for group closeness centrality. In SIAM conference
on applied and computational discrete algorithms (ACDA23).
SIAM; 2023:1–12, https://doi.org/10.1137/1.9781611977714.1.
47. Grady SK, Abu-Khzam FN, Hagan RD, Shams H, Langston MA:
Domination based classification algorithms for the control-
lability analysis of biological interaction networks. Sci Rep
2022, 12, 11897, https://doi.org/10.1038/s41598-022-15464-4.
48. Wong R, Chang W-L, Chung W-Y, Vasilakos AV: Biomolecular
and quantum algorithms for the dominating set problem in
arbitrary networks. Sci Rep 2023, 13:4205, https://doi.org/
10.1038/s41598-023-30600-4.
49. Parmer T, Rocha LM, Radicchi F: Influence maximization in
Boolean networks. Nat Commun 2022, 13:3457, https://doi.org/
10.1038/s41467-022-31066-0.
50. Someya W, Akutsu T, Schwartz J-M, Nacher JC: Measuring
criticality in control of complex biological networks. npj
Systems Biology and Applications 2024, 10:9, https://doi.org/
10.1038/s41540-024-00333-9.
51. Li S, Liu W, Wu R, Li J: An adaptive attack model to network
controllability. Reliab Eng Syst Saf 2023, 235, 109252, https://
doi.org/10.1016/j.ress.2023.109252.
52
*
. Manicka S, Johnson K, Levin M, Murrugarra D: The nonlinearity
of regulation in biological networks. npj Systems Biology and
Applications 2023, 9:10, https://doi.org/10.1038/s41540-023-
00273-w.
The authors analyze the property of biological system components to
be (non)linearly regulated using Boolean network models.
53. Wang P, Wen X, Li H, Lang P, Li S, Lei Y, Shu H, Gao L, Zhao D,
Zeng J: Deciphering driver regulators of cell fate decisions
from single-cell transcriptomics data with CEFCON. Nat
Commun 2023, 14:8459, https://doi.org/10.1038/s41467-023-
44103-3.
54. Peng W, Yu P, Dai W, Fu X, Liu L, Pan Y: A graph convolution
network-based model for prioritizing personalized cancer
driver genes of individual patients. IEEE Trans Nano-
Bioscience 2023, 22:744–754, https://doi.org/10.1109/
TNB.2023.3277316.
55. Duan Z, Dai Y, Hwang A, Lee C, Xie K, Xiao C, Xu M, Girgenti MJ,
Zhang J: iHerd: an integrative hierarchical graph represen-
tation learning framework to quantify network changes and
prioritize risk genes in disease. PLoS Comput Biol 2023, 19:
1–20, https://doi.org/10.1371/journal.pcbi.1011444.
56. Wu Y, Gao M, Zeng M, Zhang J, Li M: BridgeDPI: a novel Graph
Neural Network for predicting drug-protein interactions. Bio-
informatics 2022, 38:2571–2578, https://doi.org/10.1093/bioin-
formatics/btac155.
57
* *
. Theodoris CV, Xiao L, Chopra A, Chaffin MD, Al Sayed ZR,
Hill MC, Mantineo H, Brydon EM, Zeng Z, Liu XS, Ellinor PT:www.sciencedirect.com
Connecting the dots Siminea et al. 11Transfer learning enables predictions in network biology.
Nature 2023, 618:616–624, https://doi.org/10.1038/s41586-023-
06139-9.
A deep learning model trained on a large corpus of transcriptome data
to predict key genes or drug targets in cases where there is little data.
58. Pan Y, Wang Y, Guan J, Zhou S: PCGAN: a generative
approach for protein complex identification from protein
interaction networks. Bioinformatics 2023, 39, btad473, https://
doi.org/10.1093/bioinformatics/btad473.
59. Li J, Mi L, Ran B, Sui C, Zhou L, Li F, Dionigi G, Sun H, Liang N:
Identification of potential diagnostic and prognostic bio-
markers for papillary thyroid microcarcinoma (PTMC) based
on TMT-labeled LC-MS/MS and machine learning. J Endocrinol
Invest 2023, 46:1131–1143, https://doi.org/10.1007/s40618-022-
01960-x.
60. Cai Y, Su H, Si Y, Ni N: Machine learning-based prediction of
diagnostic markers for Graves’ orbitopathy. Endocrine 2023,
81:277–289, https://doi.org/10.1007/s12020-023-03349-z.
61. Su Y, Tian X, Gao R, Guo W, Chen C, Chen C, Jia D, Li H, Lv X:
Colon cancer diagnosis and staging classification based on
machine learning and bioinformatics analysis. Comput Biol
Med 2022, 145, 105409, https://doi.org/10.1016/
j.compbiomed.2022.105409.
62. Nazari E, Pourali G, Khazaei M, Asadnia A, Dashtiahangar M,
Mohit R, Maftooh M, Nassiri M, Hassanian SM, Ghayour-
Mobarhan M, et al.: Identification of potential biomarkers in
stomach adenocarcinoma using machine learning ap-
proaches. Curr Bioinf 2023, 18:320–333, https://doi.org/10.2174/
1574893618666230227103427.
63. Meng J, Du H, Lv H, Lu J, Li J, Yao J: Identification of the
osteoarthritis signature gene PDK1 by machine learning and
its regulatory mechanisms on chondrocyte autophagy and
apoptosis. Front Immunol 2023, 13, https://doi.org/10.3389/
fimmu.2022.1072526.
64
*
. Bai Y, Zhou L, Zhang C, Guo M, Xia L, Tang Z, Liu Y, Deng S:
Dual network analysis of transcriptome data for discovery of
new therapeutic targets in non-small cell lung cancer.
Oncogene 2023, 42:3605–3618, https://doi.org/10.1038/s41388-
023-02866-5.
A network study combining different network analysis methods to
identify a new drug for nonsmall cell lung cancer.
65. Chirom K, Malik MZ, Mangangcha IR, Somvanshi P, Singh RKB:
Network medicine in ovarian cancer: topological properties
to drug discovery. Briefings Bioinf 2022, 23, bbac085, https://
doi.org/10.1093/bib/bbac085.
66. Ahmed F, Yang YJ, Samantasinghar A, Kim YW, Ko JB, Choi KH:
Network-based drug repurposing for HPV-associated cervi-
cal cancer. Comput Struct Biotechnol J 2023, 21:5186–5200,
https://doi.org/10.1016/j.csbj.2023.10.038.
67. Gan X, Shu Z, Wang X, Yan D, Li J, Ofaim S, Albert R, Li X, Liu B,
Zhou X, Barabási A-L: Network medicine framework reveals
generic herb-symptom effectiveness of traditional Chinese
medicine. Sci Adv 2023, 9, eadh0215, https://doi.org/10.1126/
sciadv.adh0215.
68. Dwivedi PSR, Shastry CS: System biology mediated assess-
ment of molecular mechanism for sinapic acid against breast
cancer: via network pharmacology and molecular dynamic
simulation. Sci Rep 2023, 13, 21982, https://doi.org/10.1038/
s41598-023-47901-3.
69. Zhai Y, Pribis JP, Dooling SW, Garcia-Villada L, Minnick P, Xia J,
Liu J, Mei Q, Fitzgerald DM, Herman C, et al.: Drugging evolu-
tion of antibiotic resistance at a regulatory network hub. Sci
Adv 2023, 9, eadg0188, https://doi.org/10.1126/sciadv.adg0188.
70. Mark KG, Kolla S, Aguirre JD, Garshott DM, Schmitt S,
Haakonsen DL, Xu C, Kater L, Kempf G, Martínez-González B,
et al.: Orphan quality control shapes network dynamics and
gene expression. Cell 2023, 186:3460–3475.e23, https://
doi.org/10.1016/j.cell.2023.06.015.
71. Chang R, Trushina E, Zhu K, Zaidi SSA, Lau BM, Kueider-
Paisley A, Moein S, He Q, Alamprese ML, Vagnerova B,
et al.: Predictive metabolic networks reveal sex- and APOE
genotype-specific metabolic signatures and drivers forwww.sciencedirect.comprecision medicine in Alzheimer’s disease. Alzheimer’s
Dementia 2023, 19:518–531, https://doi.org/10.1002/
alz.12675.
72. Cheng Y, Yin Y, Zhang A, Bernstein AM, Kawaguchi R, Gao K,
Potter K, Gilbert H-Y, Ao Y, Ou J, et al.: Transcription factor
network analysis identifies REST/NRSF as an intrinsic regu-
lator of CNS regeneration in mice. Nat Commun 2022, 13:4418,
https://doi.org/10.1038/s41467-022-31960-7.
73. Shah AM, Myhre PL, Arthur V, Dorbala P, Rasheed H,
Buckley LF, Claggett B, Liu G, Ma J, Nguyen NQ, et al.: Large
scale plasma proteomics identifies novel proteins and pro-
tein networks associated with heart failure development. Nat
Commun 2024, 15:528, https://doi.org/10.1038/s41467-023-
44680-3.
74. Humphrey J, Venkatesh S, Hasan R, Herb JT, de Paiva Lopes K,
Küçükali F, Byrska-Bishop M, Evani US, Narzisi G, Fagegaltier D,
et al.: Integrative transcriptomic analysis of the amyotrophic
lateral sclerosis spinal cord implicates glial activation and
suggests new risk genes. Nat Neurosci 2023, 26:150–162,
https://doi.org/10.1038/s41593-022-01205-3.
75. van Oostrum M, Blok TM, Giandomenico SL, tom Dieck S,
Tushev G, Fürst N, Langer JD, Schuman EM: The proteomic
landscape of synaptic diversity across brain regions and cell
types. Cell 2023, 186:5411–5427, https://doi.org/10.1016/
j.cell.2023.09.028. e23.
76. Berry JS, Tarn J, Casement J, Duret P-M, Scott L, Wood K,
Johnsen S-J, Nordmark G, Devauchelle-Pensec V, Seror R,
et al.: Examining the biological pathways underlying clinical
heterogeneity in Sjogren’s syndrome: proteomic and
network analysis. Ann Rheum Dis 2024, 83:88–95, https://
doi.org/10.1136/ard-2023-224503.
77. Kong J, Ha D, Lee J, Kim I, Park M, Im S-H, Shin K, Kim S:
Network-based machine learning approach to predict
immunotherapy response in cancer patients. Nat Commun
2022, 13:3703, https://doi.org/10.1038/s41467-022-31535-6.
78. Franz M, Lopes CT, Fong D, Kucera M, Cheung M, Siper MC,
Huck G, Dong Y, Sumer O, Bader GD: Cytoscape.js 2023
update: a graph theory library for visualization and analysis.
Bioinformatics 2023, 39:btad031, https://doi.org/10.1093/bioin-
formatics/btad031.
79. Freeman TC, Horsewell S, Patir A, Harling-Lee J, Regan T,
Shih BB, Prendergast J, Hume DA, Angus T: Graphia: a plat-
form for the graph-based visualisation and analysis of high
dimensional data. PLoS Comput Biol 2022, 18:1–17, https://
doi.org/10.1371/journal.pcbi.1010310.
80. Csárdi G, Nepusz T, Müller K, Horvát S, Traag V, Zanini F,
Noom D: Igraph for R: R interface of the igraph library for graph
theory and network analysis. 2024, https://doi.org/10.5281/
ZENODO.7682609.
81. Pillich RT, Chen J, Churas C, Fong D, Gyori BM, Ideker T,
Karis K, Liu SN, Ono K, Pico A, Pratt D: NDEx IQuery: a multi-
method network gene set analysis leveraging the Network
Data Exchange. Bioinformatics 2023, 39:btad118, https://doi.org/
10.1093/bioinformatics/btad118.
82. OuestWare, retina. 2022.
83
* *
. Mohseni Behbahani Y, Saighi P, Corsi F, Laine E, Carbone A:
LEVELNET to visualize, explore, and compare protein-
protein interaction networks. Proteomics 2023, 23, 2200159,
https://doi.org/10.1002/pmic.202200159.
LEVELNET is a new versatile tool for visualizing, analyzing and
comparing protein–protein interaction networks that integrates data
from different sources. It represents networks as multilayered graphs to
facilitate the comparison of their subnetworks and their biological
interpretation.
84. Doncheva NT, Morris JH, Holze H, Kirsch R, Nastou KC, Cuesta-
Astroz Y, Rattei T, Szklarczyk D, von Mering C, Jensen LJ:
Cytoscape stringApp 2.0: analysis and visualization of het-
erogeneous biological networks. J Proteome Res 2023, 22:
637–646, https://doi.org/10.1021/acs.jproteome.2c00651.
85. Zhao T, Wang Z: GraphBio: a shiny web app to easily perform
popular visualization analysis for omics data. Front Genet
2022, 13, https://doi.org/10.3389/fgene.2022.957317.Current Opinion in Structural Biology 2024, 88:102881
12 Biophyiscal Methods (2024): Protein Networks in Health and Disease86. Raredon MSB, Yang J, Garritano J, Wang M, Kushnir D,
Schupp JC, Adams TS, Greaney AM, Leiby KL, Kaminski N,
et al.: Computation and visualization of cell-cell signaling
topologies in single-cell systems data using Connectome.
Sci Rep 2022, 12:4187, https://doi.org/10.1038/s41598-022-
07959-x.
87. Maitra C, Seal DB, De RK: NeuroDAVIS: a neural network
model for data visualization. Neurocomputing 2024, 573,
127182, https://doi.org/10.1016/j.neucom.2023.127182.
88. Gao K, Chen L, Li D, Liu V, Wang X, Zhang R, Lu L: Demo:
NetVision: efficient visualization front-end for packet-level
discrete-event network simulation. In Proceedings of the ACM
SIGCOMM 2023 conference, ACM SIGCOMM ’23. New York,
NY, USA: Association for Computing Machinery; 2023:
1179–1181, https://doi.org/10.1145/3603269.3610851.
89. Helmy M, Mee M, Ranjan A, Hao T, Vidal M, Calderwood MA,
Luck K, Bader GD: OpenPIP: an open-source platform for
hosting, visualizing and analyzing protein interaction data.
J Mol Biol 2022, 434, 167603, https://doi.org/10.1016/
j.jmb.2022.167603, computation Resources for Molecular
Biology.
90. Koh CWT, Ooi JSG, Ong EZ, Chan KR: STAGEs: a web-based
tool that integrates data visualization and pathway enrich-
ment analysis for gene expression studies. Sci Rep 2023, 13:
7135, https://doi.org/10.1038/s41598-023-34163-2. URL:.
91. Eberhard P, Kern M, Aichem M, Borlinghaus H, Klein K, Delp J,
Suciu I, Moser B, Dietrich D, Leist M, Schreiber F: Path-
wayNexus: a tool for interactive metabolic data analysis.
Bioinformatics 2024, btae310, https://doi.org/10.1093/bioinfor-
matics/btae310.
92. Ma A, Wang X, Li J, Wang C, Xiao T, Liu Y, Cheng H, Wang J,
Li Y, Chang Y, et al.: Single-cell biological network inference
using a heterogeneous graph transformer. Nat Commun 2023,
14:964, https://doi.org/10.1038/s41467-023-36559-0.
93. Guo MG, Sosa DN, Altman RB: Challenges and opportunities
in network-based solutions for biological questions. Briefings
Bioinf 2022, 23. https://10.1093/bib/bbab437.
94. Cvijovic M, Polster A: Network medicine: facilitating a new
view on complex diseases. Frontiers in Bioinformatics 2023, 3,
https://doi.org/10.3389/fbinf.2023.1163445.
95. Zhang L, Wang S, Liu J, Chang X, Lin Q, Wu Y, Zheng Q: MuL-
GRN: multi-level graph relation network for few-shot node
classification. IEEE Trans Knowl Data Eng 2023, 35:
6085–6098, https://doi.org/10.1109/TKDE.2022.3176880.
96. Zhao Z, Qian P, Yang X, Zeng Z, Guan C, Tam WL, Li X:
SemiGNN-PPI: self-ensembling multi-graph neural network
for efficient and generalizable protein-protein interaction
prediction. In Proceedings of the thirty-second international joint
conference on artificial intelligence, vol. 23. IJCAI; 2023:
4984–4992, https://doi.org/10.24963/ijcai.2023/554.
97
* *
. Gao Z, Jiang C, Zhang J, Jiang X, Li L, Zhao P, Yang H, Huang Y,
Li J: Hierarchical graph learning for protein-protein interac-
tion. Nat Commun 2023, 14:1093, https://doi.org/10.1038/
s41467-023-36736-1.
A hierarchical graph learning model is introduced to predict
protein–protein interactions. The model examines both outside-of-
protein and inside-of-protein of the human interactome to establish a
robust machine understanding of PPIs.
98. Jha K, Saha S, Singh H: Prediction of protein-protein inter-
action using graph neural networks. Sci Rep 2022, 12:8360,
https://doi.org/10.1038/s41598-022-12201-9.Current Opinion in Structural Biology 2024, 88:10288199. Zhao B-W, Su X-R, Hu P-W, Huang Y-A, You Z-H, Hu L:
iGRLDTI: an improved graph representation learning method
for predicting drug-target interactions over heterogeneous
biological information network. Bioinformatics 2023, 39,
btad451, https://doi.org/10.1093/bioinformatics/btad451.
100. Li M, Cai X, Xu S, Ji H: Metapath-aggregated heterogeneous
graph neural network for drug-target interaction prediction.
Briefings Bioinf 2023, 24, bbac578, https://doi.org/10.1093/bib/
bbac578.
101. Gu Y, Zheng S, Yin Q, Jiang R, Li J: REDDA: integrating mul-
tiple biological relations to heterogeneous graph neural
network for drug-disease association prediction. Comput Biol
Med 2022, 150, 106127, https://doi.org/10.1016/
j.compbiomed.2022.106127.
102. Yang Z, Zhong W, Lv Q, Dong T, Yu-Chian Chen C: Geometric
interaction graph neural network for predicting protein-
ligand binding affinities from 3D structures (GIGN). J Phys
Chem Lett 2023, 14:2020–2033, https://doi.org/10.1021/
acs.jpclett.2c03906.
103. Zhao B-W, Wang L, Hu P-W, Wong L, Su X-R, Wang B-Q, You Z-
H, Hu L: Fusing higher and lower-order biological information
for drug repositioning via graph representation learning.
IEEE Transactions on Emerging Topics in Computing 2024, 12:
163–176, https://doi.org/10.1109/TETC.2023.3239949.
104. Lv Q, Chen G, Yang Z, Zhong W, Chen CY-C: Meta learning
with graph attention networks for low-data drug discovery.
IEEE Transact Neural Networks Learn Syst 2023:1–13, https://
doi.org/10.1109/TNNLS.2023.3250324.
105. Jumper J, Evans R, Pritzel A, Green T, Figurnov M,
Ronneberger O, Tunyasuvunakool K, Bates R, Zídek A,
Potapenko A, et al.: Highly accurate protein structure predic-
tion with AlphaFold. Nature 2021, 596:583–589, https://doi.org/
10.1038/s41586-021-03819-2.
106. Wang K, Zhou R, Tang J, Li M: GraphscoreDTA: optimized
graph neural network for protein-ligand binding affinity pre-
diction. Bioinformatics 2023, 39, btad340, https://doi.org/
10.1093/bioinformatics/btad340.
107. Wicky BIM, Milles LF, Courbet A, Ragotte RJ, Dauparas J,
Kinfu E, Tipps S, Kibler RD, Baek M, DiMaio F, et al.: Halluci-
nating symmetric protein assemblies. Science 2022, 378:
56–61, https://doi.org/10.1126/science.add1964.
108. Guo H-B, Perminov A, Bekele S, Kedziora G, Farajollahi S,
Varaljay V, Hinkle K, Molinero V, Meister K, Hung C, et al.:
AlphaFold2 models indicate that protein sequence de-
termines both structure and dynamics. Sci Rep 2022, 12,
10696, https://doi.org/10.1038/s41598-022-14382-9.
109. Gao M, Nakajima An D, Parks JM, Skolnick J: AF2Complex
predicts direct physical interactions in multimeric proteins
with deep learning. Nat Commun 2022, 13:1744, https://doi.org/
10.1038/s41467-022-29394-2.
110. Elfmann C, Stülke J: PAE viewer: a webserver for the inter-
active visualization of the predicted aligned error for multi-
mer structure predictions and crosslinks. Nucleic Acids Res
2023, 51:W404–W410, https://doi.org/10.1093/nar/gkad350.
111. Sayin AZ, Abali Z, Senyuz S, Cankara F, Gursoy A, Keskin O:
Conformational diversity and protein-protein interfaces in
drug repurposing in Ras signaling pathway. Sci Rep 2024, 14:
1239, https://doi.org/10.1038/s41598-023-50913-8.www.sciencedirect.com