Content uploaded by Haeseung Lee
Author content
All content in this area was uploaded by Haeseung Lee on Jan 18, 2021
Content may be subject to copyright.
REVIEW
In silico drug repositioning: from large-scale transcriptome data
to therapeutics
Ok-Seon Kwon
1,3
•Wankyu Kim
2
•Hyuk-Jin Cha
1,3
•Haeseung Lee
2
Received: 30 January 2019 / Accepted: 26 July 2019
ÓThe Pharmaceutical Society of Korea 2019
Abstract Drug repositioning is an attractive alternative to
conventional drug development when new beneficial
effects of old drugs are clinically validated because phar-
macokinetic and safety profiles are generally already
available. Since *30% of drugs newly approved by the
US food and drug administration (FDA) are developed
through drug repositioning, identifying novel usage for
existing drugs is an emerging strategy for developing dis-
ease treatments. With advances in next-generation
sequencing technologies, available transcriptome data
related to diseases have expanded rapidly. Harnessing these
resources enables a better understanding of disease mech-
anisms and drug mode of action (MOA), and moves toward
personalized pharmacotherapy. In this review, we briefly
outline publicly available large-scale transcriptome data-
bases and tools for drug repositioning. We also highlight
recent approaches leading to the discovery of novel drug
targets, drug response biomarkers, drug indications, and
drug MOA.
Keywords Drug repositioning In silico drug
repositioning Transcriptome Pharmacogenomics Big
data
Introduction
Despite major scientific and technological advances in not
only basic research but also drug discovery and develop-
ment, the number of new drugs approved by the US Food
and Drug Administration (FDA) has steadily declined. This
trend is called Eroom’s Law, and is the reverse of the more
familiar Moore’s Law that refers to the exponential
increase in the number of transistors in a dense integrated
circuit (Scannell et al. 2012). To improve the productivity
and success rate of drug discovery, many novel strategies
have been developed, including target structure-based drug
design (Chen and Butte 2016), disease modeling with stem
cell technology (Xia and Wong 2012; Kondo et al. 2017),
and drug repositioning (or repurposing) (Hernandez et al.
2017).
Among these strategies, drug repositioning could
potentially overcome the challenges of drug discovery
(Hernandez et al. 2017; Novac 2013). In this approach,
toxicity, pharmacokinetics, and pharmacodynamics profiles
of a given drug are fully characterized during clinical trials.
Once the efficacy of old drugs or drug candidates on new
indications is established, successful therapeutic interven-
tion can be anticipated with less risk from failure during
clinical trials. There are many examples of serendipitous
success for drug repositioning, as exemplified by thalido-
mide (originally developed for sedation and nausea, and
now prescribed for multiple myeloma), sildenafil (origi-
nally developed for angina but now used to treat male
erectile dysfunction), and raloxifene (originally developed
for breast cancer and now an established treatment for
osteoporosis) (Pushpakom et al. 2018). Currently, *30%
of drugs newly approved by the FDA are derived from drug
repositioning. Encouraged by these successes, more sys-
tematic approaches are being applied to identify new
&Haeseung Lee
haeseung@ewha.ac.kr
1
Research Institute of Pharmaceutical Sciences, Seoul
National University, Seoul 08826, Republic of Korea
2
Ewha Research Center for Systems Biology, Department of
Life Sciences, Ewha Womans University, Seoul 03760,
Republic of Korea
3
College of Pharmacy, Seoul National University,
Seoul 08826, Republic of Korea
123
Arch. Pharm. Res. Online ISSN 1976-3786
https://doi.org/10.1007/s12272-019-01176-3 Print ISSN 0253-6269
indications for old drugs, drug candidates, and drugs
withdrawn from the market via drug repositioning and
repurposing (Kim et al. 2016).
Recent advances in next-generation sequencing and
high-throughput technologies have rapidly expanded
available biological and chemical datasets, ushering in the
era of big data (Costa 2014). Furthermore, computational
methods taking advantage of these datasets have been
actively developed to couple diseases with novel thera-
peutics (Jin and Wong 2014). In particular, data-driven
approaches based on large-scale transcriptome data have
accelerated the discovery of candidate drugs across a wide
range of diseases (Chen and Butte 2016), some of which
are enrolled in clinical trials (Jahchan et al. 2013). Tran-
scriptional profiling, especially of mRNA, provides a
comprehensive view of biological changes that reflect the
overall consequences of multiple genetic variations. Thus,
disease mechanisms and drug mode of action (MOA) can
often be elucidated based on altered transcriptome profiles.
In this review, we describe publicly available resources
widely used for transcriptome-based drug repositioning,
and focus on recently developed computational approaches
to identify new drug targets, drug response biomarkers,
drug indication, and drug MOA that utilize these databases.
Public resources for in silico drug repositioning
A number of notable reference datasets widely used in
transcriptome-based drug repositioning have recently been
published and updated (Table 1) (Chen and Butte 2016;
Kannan et al. 2016). These are divided into (i) disease-
based, (ii) drug-based, and (iii) knowledge-based datasets
depending on the biological perspective that the data
describe. Disease-based datasets such as The Cancer
Genome Atlas (TCGA) and Cancer Cell Line Encyclopedia
(CCLE) include gene expression profiles and detailed
information on clinical or preclinical samples. Drug-based
datasets such as CMap, LINCS, and CTRP include gene
expression profiles of drug perturbation, drug efficacy,
known targets, and other drug-related features. Knowledge-
based datasets such as Gene Ontology, MSigDB, and
KEGG include gene or protein functional annotations that
are used to understand or interpret mechanisms of disease
or drug action based on a set of genes. Datasets listed
herein have been thoroughly investigated for hypothesis
testing and discovery in the pharmacogenomics field, but
we selectively describe their applications in identifying
drug targets, drug response biomarkers, drug indication,
and drug MOA (Fig. 1).
Clinical and preclinical transcriptome data
With rapid advances in systemic approaches, integrative
analysis of multiple-layer omics data from various sources
is widely used not only to identify drug–disease relation-
ships, but also to discover optimal drug targets and/or drugs
for diseases (Kannan et al. 2016). Thus, it is of the utmost
importance to develop better integrative platforms and
databases for disease-related information that are publi-
cally accessible.
Patient-derived transcriptome data
The Cancer Genome Atlas (TCGA; https://cancergenome.
nih.gov) project, a joint collaboration between the National
Cancer Institute (NCI) and the National Human Genome
Research Institute (NHGRI), is one of the largest public
resources for multi-layer cancer genomics, with over
11,000 patient profiles representing 36 cancer types, and 15
genomic assays per tumor type. TCGA data contain
information on tumors such as gene expression, copy
number variation, somatic mutations, single-nucleotide
polymorphisms (SNPs), and clinical outcomes with
pathological annotation. Even though TCGA database
contains comprehensive information of cancer, TCGA has
not been complete in the aspect of missing information
such as transcriptome from normal tissue or drug treatment
history. In this line, considering the difference of analytical
breadth such as incomplete information, and emergent
themes across cancer type and organ of origins, TCGA
launched the Pan-Cancer analysis project to provide com-
prehensive information about cancer. Through TCGA Pan-
Cancer Atlas project, comprehensive database of 12 dif-
ferent tumor type including a total of 5074 tumor sample
has been assessed for clinical, genomic, epigenomic,
transcriptional and proteomic data on at least one platform
each (Cancer Genome Atlas Research et al. 2013). More-
over, Pan-Cancer Atlase reclassifies human tumor into
three major categories based on molecular similarities:
cell-of-origin pattern, oncogenic processes and signaling
pathways (Sanchez-Vega et al. 2018). The integrative data
from TCGA Pan-Cancer Atlas, this is a powerful emerging
resource as we enter a new era of cancer treatment.
Cell line-derived transcriptome data
The Cancer Cell Line Encyclopedia (CCLE; https://portals.
broadinstitute.org/ccle) database is a large-scale genomic
dataset of gene expression, copy number, and DNA
sequencing data from 1457 human cancer cell lines,
encompassing 36 tumor types. Similarly, transcriptome
data from in vitro cancer cell lines are provided in the
Genomics of Drug Sensitivity in Cancer (GDSC) database,
O.-S. Kwon et al.
123
which includes over 1000 cancer cell lines, and NCI-60,
which covers 60 cancer cell lines. Moreover, by compiling
cell line and compound sensitivity data using CCLE,
GDSC, and NCI-60, a profound understanding of the
connections between pharmacological vulnerability and
molecular signature of a responsive cancer cell line can be
gleaned (Barretina et al. 2012; Cancer Cell Line Encyclo-
pedia and Genomics of Drug Sensitivity in Cancer 2015).
However, standardization with additional curation and
processing to combine information of cell line and drug
Table 1 Publically accessible databases widely used in transcriptomic-based in silico drug repositioning
Category Name Description (as of December 2018) URL
Omics data
repositories
GEO Raw and processed transcriptome data from multiple platforms https://www.ncbi.nim.nih.
gov/geo/
SRA Sequencing data from multiple platforms https://www.ncbi.nlm.nih.
gov/sra
ArrayExpress Raw and processed transcriptome data from multiple platforms https://www.ebi.ac.uk/
arrayexpress/
Disease-
based
ICGC Genomic, transcriptomic, epigenomic and clinical data from [24,000 tumours
(22 different tumour types)
https://icgc.org/
TCGA Genomic, transcriptomic, epigenomic and clinical data from [11,000 tumours
(33 different tumour types)
http://tcga-data.nci.nih.gov/
CCLE Genomic, transcriptomic and epigenomic data from [1000 cancer cell lines http://www.broadinstitute.
org/ccle
GDSE Genomic, transcriptomic and epigenomic data from [1000 cancer cell lines https://wwwcancerrxgene.
org/
Durg-based CMap Gene expression profiles for 1309 chemical compounds in 5 cancer cell lines https://portals.broadinstitute.
org/cmap/
LINCS Gene expression profiles for perturbagens (20,413 chemicals and 2119 genetic
knockdown/overexpression) across 77 cell lines
https://clue.io/
NCI60 Drug response data (GI50, LC50 values) of 60 cancer cell lines for 45,449
compounds
https://dtp.cancer.gov/
discovety_development/
nci-60/
CTRP Drug response data (AUC, EC50 values) of 860 cancer cell lines for 481
compounds
https://portals.broadinstitute.
org/ctrp/
CCLE Drug response data (AUC, IC50 values) of 504 cancer cell lines for 24
compounds
http://www.broadinstitute.
org/ccle
GDSE Drug response data (AUC, IC50 values) of 714 cancer cell lines for 142
compounds
https://www.cancerrxgene.
org/
NCl-
ALMANAC
Therapeutic activity for pairwise combinations ([5000 pairs) of 104 FDA-
approved anticancer drugs against NCI-60 cell lines
https://dtp.cancer.
govincialmanac
PubChem
Bioassay
Chemical compound screening data, including [3.4 M unique chemical
compounds, [12 K protein targets, and [1 M assays.
https://pubchem.ncbi.nlm.
nih.gov/
ChEMBL Chemical compound screening data, including [2.2 M unique chemical
compounds, [12 K protein targets, and [1 M assays
https://www.ebi.ac.uk/
chembl/
Knowledge-
based
Gene
ontology
Database collection of over 15,000 genes with gene-ontology, including 13,212
biological process, 1547 cellular components and 4162 molecular functions
http://www.geneontology.
org/
MsigDB Data repository, contained 17,810 genes sets, with 8 major collection Database
collection of over 2300 biological pathways for 25 different species
http://software.broadinstitute.
org/gsea/msigdb
http://www.wikipathways.
org
KEGG Database collection for genomes, pathways, disease and compounds
information, including 3947 genes, 200 pathway and 9324 gene-pathway
association
http://www.genome.jp/kegg
BioCarta Database collection of 1396 genes with 254 pathway and 4417 gene-pathway
association
http://www.biocarta.com
Reactome Database collection of 7535 genes with 1638 pathway and 83,680 gene-pathway
association
http://www.reactome.org
In silico drug repositioning: from large-scale transcriptome data to therapeutics
123
treatment is a challenge that is required to enable integra-
tive analysis.
Transcriptome data following treatment
The connectivity map (build 02)
The connectivity map (CMap) is a collection of genome-
wide gene expression data from five human cancer cell
lines treated with 1309 compounds obtained using the
Affymetrix microarray platform (Lamb et al. 2006). The
concept of CMap is to establish a comprehensive reference
database of drug-induced gene expression profiles to
compare with a set of genes representing the biological
state of interest, and to discover functional connections
between them. It provides a web-based tool that performs
simple pattern matching analysis with CMap reference data
based on a user-submitted gene list, but is no longer
updated or modified (https://portals.broadinstitute.org/
cmap/).
Library of integrated network-based cellular signatures
(LINCS) L1000
LINCS L1000, also referred to as LINCS, or an extended
version of CMap, is a resource containing 1.3 million gene
expression profiles associated with 20,413 chemical per-
turbagens (e.g., small molecules or drugs) and *5000
genetic perturbagens (e.g., single-gene knockdown or
overexpression) (Subramanian et al. 2017). Data were
acquired using the L1000 assay developed by the Broad
Institute CMap team to facilitate rapid high-throughput
gene expression profiling at low cost. The L1000 assay
measures the expression of 978 landmark genes, and
expression values for remaining genes are estimated by a
linear model using a diverse collection of transcriptome
data from Affymetrix microarray data in Gene Expression
Omnibus (GEO). LINCS L1000 datasets are fully down-
loadable from GEO (accession: GSE92742) and are easily
accessible via the cloud-based software platform CLUE
(https://clue.io/).
Fig. 1 Public databases utilized
in drug repositioning pipelines
O.-S. Kwon et al.
123
Knowledge-based gene annotations
In drug development pipelines, the knowledge base, which
includes information on drugs, biological implications of
drugs, and clinical outcomes, can reveal associations and
thereby provide integrative implications (Fotis et al. 2018).
To provide biological insight relevant to drug development,
utilizing molecular interaction data gathered from various
knowledge bases is a potentially powerful method. Below,
gene annotation databases that illuminate the biological
background by exploring molecular mechanisms and
molecular interactions are briefly described.
Gene annotation databases such as the Kyoto Encyclo-
pedia of Genes and Genomes (KEGG), Gene Ontology
(GO), and the Molecular Signatures Database (MSigDB)
provide diverse types of interaction models, including
signaling pathways, metabolic networks, and regulatory
interactions, based on transcriptome data. The KEGG
database collection integrates genomic and chemical
information. In terms of systemic information, the KEGG
database includes KEGG Pathway containing pathway
maps, KEGG Disease comprising disease entries, and
KEGG Drug that includes comprehensive information on
drugs, approved in Japan, the USA, and Europe. In par-
ticular, KEGG Pathway, which contains manually drawn
pathway maps, provides intuitive information on interac-
tions between genes and proteins (Kanehisa et al. 2018).
By contrast, the GO project aims to provide ontologies of
genes defined with their own properties. GO provides
ontologies and annotation information for three domains:
cellular component (CC), biological process (BP), and
molecular function (MF) (Zhang et al. 2014; Rhee et al.
2008). Meanwhile, MSigDB, developed for gene set
enrichment analysis (GSEA), covers a large number of
gene sets with annotations and links from external
resources including KEGG, GO, GEO, and ArrayExpress
(Liberzon et al. 2011). Together, these knowledge-based
databases provide a foundation for computational drug
repositioning based on transcriptome analysis, and collate
valuable information such as target identification and MOA
of drugs.
Web-based drug repositioning tools
Exploring complex large data sets described above often
requires high-performance computing resources but access
is difficult without proficient computer skills. A number of
user-friendly interface-based web tools that assist research
in drug repositioning have lowered this barrier for all sci-
entists regardless of their computational backgrounds (Sam
and Athri 2019). Since most transcriptome-based studies
initiate hypothesis testing on the sets of differentially
expressed genes (DEGs) that represent the biological state
of interest, various web-based analytic tools have been
developed to associate these DEGs with drugs.
CLUE (https://clue.io/l1000-query) provides a cloud-
based query tool to find positive or negative connections
between a user-submitted gene set and all the signatures in
LINCS L1000 (Subramanian et al. 2017). The term sig-
nature here refers to a vector of differential gene expression
values (Z score) induced by individual perturbagen in
LINCS L1000. CLUE returns a list of approximately
50,000 unique perturbagens, including small molecules,
single-gene knockdown and overexpression, with a score
based on the amount of inducing expressional changes of
the input genes.
L1000CDS
2
(http://amp.pharm.mssm.edu/L1000CDS2/)
is another LINCS L1000 signature search engine (Duan
et al. 2016). It processed LINCS L1000 data to define the
signatures using the characteristic direction method (Clark
et al. 2014). Predictive performance of L1000CDS
2
was
tested on expression signatures from human cells infected
with Ebola virus. Based on these signatures, kenpaullone, a
GSK3B/CDK2 inhibitor was predicted and its dose-de-
pendent efficacy in inhibiting Ebola infection was validated
in vitro.
DeSigN (http://design-v2.cancerresearch.my/query)
associates drug efficacy with a user-submitted gene set by
comparing it against drug response-related gene expression
signatures for 140 drugs (Lee et al. 2017). The individual
expression signature of a drug was defined as a differential
gene expression profile derived by using its drug sensitivity
(IC50) and baseline gene expression against cancer cell
lines in GDSC data. DeSigN was validated using four
different drug sensitivity studies deposited in the GEO
database. In addition, bosutinib, a src tyrosine kinase
inhibitor, was predicted as a sensitive drug for oral squa-
mous cell carcinoma (OSCC) and its efficacy was
demonstrated by in vitro viability assay.
Prediction of novel drug–target interactions
Drug polypharmacology (Hopkins 2008), in which a single
drug acts on multiple targets, implies the therapeutic
potential of a drug for new indications, and thus facilitates
innovative and successful drug repositioning (Reddy and
Zhang 2013). Drug-induced transcriptome data reflect the
combined effects of multiple targets of a drug, providing
insight into its MOA or unintended off-targets. CMap and
LINCS are the most comprehensive resources for exploring
novel drug–target interactions (DTIs). From CMap data,
high correlations among gene expression changes caused
by drugs sharing the same target have been systematically
shown (Wang et al. 2013). Several methods have been
developed to expand known drug–target relationships
In silico drug repositioning: from large-scale transcriptome data to therapeutics
123
based on drug similarity at the gene expression level
(Hizukuri et al. 2015; Iwata et al. 2017). On the other hand,
drug-induced differentially expressed genes (DEGs) com-
prise only a small proportion of known target genes, but are
distributed close to targets in the functional protein–protein
interaction (PPI) network (Isik et al. 2015). Based on these
observations, a target prediction model was developed that
integrates drug-induced DEGs and the network topology of
PPIs.
Genetically perturbed transcriptome data can also be
utilized to seek novel DTIs using drug-induced transcrip-
tome data. Importantly, novel connections between a drug
and its target gene can be inferred from common expres-
sion signatures shared by both drug treatments and loss of
gene function in yeast systems (Hughes et al. 2000). This
idea was applied in human cancer cells using LINCS
L1000 data, which led to the discovery of compound BRD-
1868 that targets Casein Kinase 1A1, which is related to
drug resistance in lung cancer (Lantermann et al. 2015;
Subramanian et al. 2017). Another similar approach com-
prehensively predicted novel DTIs between 1124 drugs and
829 target proteins by correlating gene expression patterns
caused by chemical and genetic perturbations (Sawada
et al. 2018). Notably, this approach distinguished predicted
DTIs by inhibitory and activatory interactions, depending
on whether a genetic perturbation directly compared with a
drug is knockdown or overexpression.
Identification of drug response biomarkers
In drug development pipelines, most drugs are developed
based on the molecular features of a given disease. In terms
of drug repositioning, identification of indicators or
biomarkers of repurposed drugs is critical to match the
appropriate drug with the right patient based on predicted
drug responses (Kelloff and Sigman 2012). With great
advances in sequencing technologies, large-scale tran-
scriptome data and pharmacogenomics-based disease
models have emerged that aid the identification of
biomarkers and the prediction of drug responses. In the
Cancer Therapeutics Response Portal (CTRP) database,
transcriptome-based biomarkers of drug sensitivity have
been identified by integrating drug response profiles for
481 anticancer drugs across 860 cancer cell lines (Cancer
Cell Line Encyclopedia and Genomics of Drug Sensitivity
in Cancer 2015). Drug response profiles from CTRP can be
utilized to predict drug responses in cell lines, which have
particular disease features or defined gene signatures,
suggesting that drugs may sensitize certain disease fea-
tures. For example, sensitivity patterns of 481 chemical
compounds were correlated with *19,000 basal transcript
levels across 823 different human cancer cell lines, and this
demonstrated that analyzing the basal gene expression
profile of cell lines can predict drug responses and illu-
minate the mechanisms of small molecules (Rees et al.
2016). Furthermore, based on validation with previously
annotated targets and drugs, as exemplified by BCL2 and
ABT-199 (Rees et al. 2016) and SLC35F2 and YM-155
(Winter et al. 2014; Rees et al. 2016), ML239 was newly
identified, after being originally identified by phenotypic
screening to selectively eliminate epithelial breast cancer
cells, and found to activate fatty acid desaturase 2 (FADS2)
(Rees et al. 2016). Furthermore, chemoresistance score was
defined, which is strongly correlated with mesenchymal
cancer traits, by leveraging integrative transcriptome data
from both CTRP and CCLE (Hong et al. 2018). Further-
more, analyzing the association between drug response
profiles and genome-wide RNAi screening data in the
Achilles project (Tsherniak et al. 2017) identified ITGB3,
highly expressed in mesenchymal-type lung cancer cell
lines (Bae et al. 2016; Hong et al. 2016), as an Achilles’
heel for chemoresistant cancer cells with mesenchymal
traits. In conclusion, dependency on ITGB3 was considered
to be one of the major factors determining the responses of
most chemotherapeutic drugs (Hong et al. 2018). Thus,
leveraging publicly available pharmacogenomics data
linked to diseases offers a promising approach for identi-
fying drug biomarkers with statistical reliability.
Discovery of novel drug indications
Systems biology approaches have utilized large-scale
pharmacogenomics data to identify previously unrecog-
nized relationships between diseases and drugs. These
approaches generally begin by defining a gene expression
signature (e.g., a collection of genes representing a disease
state) and comparing it directly against compound signa-
tures in reference databases such as CMap or CTRP. This
query signature can be derived from a disease, drug per-
turbation, or genetic perturbation, and used to perform
(i) drug–disease, (ii) drug–drug, or (iii) drug–gene com-
parisons (Fig. 2). Below, several studies that have discov-
ered new drug indications through such comparisons are
described.
The first case, the most prevalent approach, typically
defines a disease signature as a set of DEGs obtained by
comparing disease and corresponding control (healthy)
states, and seeks a drug whose perturbation reverses the
disease signature. For example, disease signatures were
generated for 100 diseases using microarray data from
GEO, and each disease signature was mapped to 164 drug
signatures in CMap (build 01) (Dudley et al. 2011; Sirota
et al. 2011). Among the highly anti-correlated disease–drug
pairs, many known disease–drug relationships were
O.-S. Kwon et al.
123
recovered, along with new associations including cime-
tidine (a histamine H
2
receptor antagonist for antiulcer
treatment) for the treatment of lung adenocarcinoma, and
topiramate (a voltage-gated sodium and calcium channel
blocker as an anticonvulsant) for the treatment of inflam-
matory bowel disease. A similar systematic approach using
a small cell lung cancer (SCLC) expression signature found
that antidepressant drugs (imipramine, a tricyclic antide-
pressant; promethazine, a histamine H
1
receptor antagonist
for allergies; and bepridil, an amine calcium channel
blocker) are potent inducers of apoptosis in SCLC (Jahchan
et al. 2013). These findings led to the enrolment into
clinical trials of a related molecule, the tricyclic antide-
pressant desipramine, for the treatment of SCLC
(NCT01719861, phase IIa clinical trials). In another
example, comparison of a metastatic colon signature
against compound signatures in CMap (build 02) resulted
in the identification of citalopram (a selective serotonin
reuptake inhibitor and antidepressant), troglitazone (a
ligand mimetic of PPARcand antihyperglycemic agent),
and enilconazole (a fungicide) drugs for the treatment of
colorectal cancer metastasis (van Noort et al. 2014). A
common assumption of these studies is that a strong anti-
correlation between a disease and drug signatures indicates
that the drug may potentially have a therapeutic effect on
the disease.
Disease signatures can also be used to characterize
disease states for other biological systems (e.g., cancer
cells or organoids), and may be associated with drug
activity such as IC50, EC50, and AUC values. For exam-
ple, a mesenchymal score was calculated using a mes-
enchymal signature for each cancer cell line available in
CTRP, and correlated with cell line sensitivity against 481
compounds (Viswanathan et al. 2017). The authors found
that ferroptosis inducers (e.g., RSL3, ML210, and ML162)
were selectively potent against mesenchymal cancer cells
via inhibition of a lipid peroxidase pathway. A similar
approach using a YM155-resistant signature led to the
Fig. 2 Three signature types
(disease, drug perturbation, and
genetic perturbation) used to
compare compound signatures
in pharmacogenomics databases
for identifying novel
relationships between diseases
and drugs
In silico drug repositioning: from large-scale transcriptome data to therapeutics
123
discovery of BCL2 homology 3 mimetics (ABT-263, ABT-
737, and WEHI-539) that selectively ablate abnormal
human embryonic stem cells (hESCs) resistant to YM155,
which specifically eliminates undifferentiated hESCs (Lee
et al. 2013; Cho et al. 2018).
Drug–drug comparisons can be used to extrapolate
knowledge on a given drug to other drugs based on simi-
larity, assuming that drugs whose perturbations cause
similar gene expression changes may have similar thera-
peutic effects. Indeed, drugs with similar MOAs were
significantly enriched in the sub-modules of a large-scale
drug association network constructed based on drug-in-
duced transcriptional similarity from CMap data (Iorio
et al. 2010). In this network, fasudil, a Rho-kinase inhibitor
and vasodilator, was clustered with well-known autophagy
inducers, and its effect on autophagy enhancement was
validated.
LINCS contains drug-induced transcriptome data for
additional perturbagens causing perturbations 15-fold
beyond the range included in CMap, providing an excellent
opportunity for exploring candidate compounds. One
approach retrieved LINCS data to identify drugs whose
signatures (i.e., DEGs from comparisons before and after
drug treatment) are similar to those of known glioblastoma
(GBM) drugs (Lee et al. 2016). By integrating this signa-
ture similarity with other features such as drug targets and
chemical structures, 14 drugs were predicted for the
treatment of GBM, and more than half displayed anti-
proliferative activity against patient-derived GBM cells.
In the final case, a disease is linked to a drug based on
similarity between transcriptomic signatures generated
from a genetic perturbation (e.g., knockdown or overex-
pression of a disease biomarker) and a drug. This concept
was first applied as an alternative to targeting the poorly
druggable gene encoding integrin beta 3 (ITGB3), respon-
sible for chemoresistance in mesenchymal lung cancer
(Hong et al. 2018). From LINCS data, atorvastatin (a
HMG-CoA reductase inhibitor for anti-dyslipidemia)
mimicked expression changes caused by knockdown of
ITGB3 and was identified as a chemosensitizer. A similar
approach was performed for the N-acetylgalactosaminyl-
transferase 14 (GALNT14) protein, the expression of which
is strongly correlated with lung cancer recurrence and
metastasis (Lee et al. 2008; Kwon et al. 2015). Due to a
lack of feasible drugs that directly inhibit the GALNT14
protein, the authors generated the gene expression signa-
ture of shRNA-mediated GALNT14 depletion in metastatic
lung cancer, and identified bortezomib (the first-in-class
proteasome inhibitor for multiple myeloma) (Argyriou
et al. 2008), which likely reverses GALNT14-dependent
gene expression (Kwon et al. 2018).
All the above studies successfully identified drugs by
matching transcriptomic signatures of drugs and
therapeutic targets, rather than attempting to inhibit these
protein targets directly. Given that many potential molec-
ular targets identified from cancer genomic profiling are
undruggable (Lazo and Sharlow 2016), this approach could
prove to be a viable strategy in cancer pharmacology.
Identification of drug mode of action
Identification of the molecular pathways and adverse
effects of a compound are crucial for drug repositioning.
Traditionally, the MOA of a drug has been predicted based
on analysis of chemical structure, gene expression profiles
following drug treatment (Lamb 2007), and side effect
similarity (Campillos et al. 2008). Furthermore, most of
these approaches are only applied to drugs that are well-
characterized based on the available structure and docu-
mented side effect (Iorio et al. 2010). Thus, when prior
information on drugs is lacking, gene signature-based
methods are the most cost-effective approach for eluci-
dating the MOA (Lamb et al. 2006; Lamb 2007). In this
regard, a subset of differential gene expression data fol-
lowing treatment can provide profound information on
connections between drugs, pathways, and diseases
through pharmacogenomics (e.g., CMap and LINCS) and
pathway (e.g., KEGG and GO) databases. As an example
of the power of identifying the MOA of a drug, fasudil was
newly identified to induce cellular autophagy through
network analysis, and could therefore be applicable for
neurodegenerative disorders (Iorio et al. 2010). Moreover,
they provide their approach for discovering MOA with
publically accessible tool, named as Mode of Action by
NeTwoRk Analysis (MANTRA, http://mantra.tigem.it).
Similarly, another study analyzed 16,268 compound and 68
human cell lines, gathered from LINCS, and performed
pathway enrichment analysis to reveal active pathways
(Iwata et al. 2017). By mapping onto KEGG biological
pathways using genes up- or down-regulated by the com-
pound, the authors proposed a computational approach to
identify not only active pathways, but also target proteins
and therapeutic indications. In another example, borte-
zomib was found to interrupt tumor metastasis in lung
cancer (Kwon et al. 2018), caused by an unexpected off-
target effect within cells, independent of proteasome inhi-
bition. The authors subsequently discovered the MOA of
bortezomib on metastasis by conducting KEGG and GSEA
analyses with combined transcriptome data from borte-
zomib-treated cells and GALNT14 expressing lung cancer
patients from TCGA. It was concluded that bortezomib
suppresses the TGFb-dependent gene signature, and
thereby inhibits tumor metastasis in lung cancer (Kwon
et al. 2018).
O.-S. Kwon et al.
123
Conclusions and future directions
With rapid advances in technology, complex transcriptome
datasets that reflect disease systems, ranging from single
cells to patients, are emerging and expanding. Diverse
transcriptome datasets enable researchers to achieve effi-
cient drug repositioning by prediction of drug side effects,
drug indications, and drug MOA. However, transcriptome
data has inherent challenging limitations. First, integrating
gene expression data from different platforms (e.g.,
microarray, RNA-seq, or L1000 assay) is intriguing as the
range of genes measured varies depending on the platform.
Second, since changes in mRNA levels are relatively more
sensitive than changes in DNA molecules, the effects of
various environmental factors that induce gene expression
changes are comprehensively reflected in transcriptome
data. In particular, large-scale data such as CMap generated
over a long period of time contain considerable experi-
mental variations or noise due to batch effects. Additional
pre-processing steps are therefore required to yield reliable
results. Third, in contrast to detecting genetic variations
compared to a reference genome, there is no representative
reference to define whether gene expression levels are high
or low. To determine if gene expression has changed under
a certain condition, there should be comparable gene
expression data under a control condition. The former two
limitations can be overcome by applying sophisticated
normalization methods, but the last one is still challenging.
Despite these limitations, systems biology approaches
leveraging these datasets can reveal previously unrecog-
nized relationships between drugs and diseases, and pro-
vide alternative strategies for associating drugs with newly
identified targets, regardless of their druggability. Since
drugging undruggable molecular targets represents a major
hurdle for traditional drug discovery methods, new thera-
peutic indications for existing drugs can prove crucial for
certain diseases. In particular, for personalized pharma-
cotherapy, an emerging approach for disease treatment that
takes into account individual characteristics of each
patient, discovering therapeutic indications from matched
drugs may be of great benefit to drug development and
drug repositioning. In this article, we reviewed publicly
available large-scale transcriptome databases and multi-
disciplinary methods utilizing the data within them. These
big data approaches hold great promise for overcoming the
limitations of traditional drug discovery pipelines and
supporting the field of precision pharmacology.
Acknowledgements This work was supported by the National
Research Foundation of Korea (NRF) via grants funded by the Korea
government (MSIT; NRF-2017R1A6A3A11030794, NRF-
2017M3C9A5028690, and NRF-2019R1C1C1008710) to HSL,
WKK, and OSK.
Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of
interest.
References
Argyriou AA, Iconomou G, Kalofonos HP (2008) Bortezomib-
induced peripheral neuropathy in multiple myeloma: a compre-
hensive review of the literature. Blood 112:1593–1599
Bae GY, Hong SK, Park JR, Kwon OS, Kim KT, Koo J, Oh E, Cha HJ
(2016) Chronic TGFbeta stimulation promotes the metastatic
potential of lung cancer cells by Snail protein stabilization
through integrin beta3-Akt-GSK3beta signaling. Oncotarget
7:25366–25376
Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA,
Kim S, Wilson CJ, Lehar J, Kryukov GV, Sonkin D, Reddy A,
Liu M, Murray L, Berger MF, Monahan JE, Morais P, Meltzer J,
Korejwa A, Jane-Valbuena J, Mapa FA, Thibault J, Bric-Furlong
E, Raman P, Shipway A, Engels IH, Cheng J, Yu GK, Yu J,
Aspesi P Jr, De Silva M, Jagtap K, Jones MD, Wang L, Hatton
C, Palescandolo E, Gupta S, Mahan S, Sougnez C, Onofrio RC,
Liefeld T, Macconaill L, Winckler W, Reich M, Li N, Mesirov
JP, Gabriel SB, Getz G, Ardlie K, Chan V, Myer VE, Weber BL,
Porter J, Warmuth M, Finan P, Harris JL, Meyerson M, Golub
TR, Morrissey MP, Sellers WR, Schlegel R, Garraway LA
(2012) The cancer cell line encyclopedia enables predictive
modelling of anticancer drug sensitivity. Nature 483:603–607
Campillos M, Kuhn M, Gavin AC, Jensen LJ, Bork P (2008) Drug
target identification using side-effect similarity. Science
321:263–266
Cancer Cell Line Encyclopedia, Genomics of Drug Sensitivity in
Cancer (2015) Pharmacogenomic agreement between two cancer
cell line data sets. Nature 528:84–87
Cancer Genome Atlas Research, Weinstein JN, Collisson EA, Mills
GB, Shaw KR, Ozenberger BA, Ellrott K, Shmulevich I, Sander
C, Stuart JM (2013) The cancer genome atlas pan-cancer
analysis project. Nat Genet 45:1113–1120
Chen B, Butte AJ (2016) Leveraging big data to transform target
selection and drug discovery. Clin Pharmacol Ther 99:285–297
Cho SJ, Kim KT, Jeong HC, Park JC, Kwon OS, Song YH, Shin JG,
Kang S, Kim W, Shin HD, Lee MO, Moon SH, Cha HJ (2018)
Selective elimination of culture-adapted human embryonic stem
cells with bh3 mimetics. Stem Cell Rep 11:1244–1256
Clark NR, Hu KS, Feldmann AS, Kou Y, Chen EY, Duan Q, Ma’ayan
A (2014) The characteristic direction: a geometrical approach to
identify differentially expressed genes. BMC Bioinform 15:79
Costa FF (2014) Big data in biomedicine. Drug Discov Today
19:433–440
Duan Q, Reid SP, Clark NR, Wang Z, Fernandez NF, Rouillard AD,
Readhead B, Tritsch SR, Hodos R, Hafner M, Niepel M, Sorger
PK, Dudley JT, Bavari S, Panchal RG, Ma’ayan A (2016)
L1000CDS(2): LINCS L1000 characteristic direction signatures
search engine. NPJ Syst Biol Appl 2:16015
Dudley JT, Sirota M, Shenoy M, Pai RK, Roedder S, Chiang AP,
Morgan AA, Sarwal MM, Pasricha PJ, Butte AJ (2011)
Computational repositioning of the anticonvulsant topiramate
for inflammatory bowel disease. Sci Transl Med 3:96ra76
Fotis C, Antoranz A, Hatziavramidis D, Sakellaropoulos T, Alex-
opoulos LG (2018) Network-based technologies for early drug
discovery. Drug Discov Today 23:626–635
Hernandez JJ, Pryszlak M, Smith L, Yanchus C, Kurji N, Shahani
VM, Molinski SV (2017) Giving drugs a second chance:
In silico drug repositioning: from large-scale transcriptome data to therapeutics
123
overcoming regulatory and financial hurdles in repurposing
approved drugs as cancer therapeutics. Front Oncol 7:273
Hizukuri Y, Sawada R, Yamanishi Y (2015) Predicting target proteins
for drug candidate compounds based on drug-induced gene
expression data in a chemical structure-independent manner.
BMC Med Genom 8:82
Hong SK, Park JR, Kwon OS, Kim KT, Bae GY, Cha HJ (2016)
Induction of integrin beta3 by sustained ERK activity promotes
the invasiveness of TGFbeta-induced mesenchymal tumor cells.
Cancer Lett 376:339–346
Hong SK, Lee H, Kwon OS, Song NY, Lee HJ, Kang S, Kim JH, Kim
M, Kim W, Cha HJ (2018) Large-scale pharmacogenomics
based drug discovery for ITGB3 dependent chemoresistance in
mesenchymal lung cancer. Mol Cancer 17:175
Hopkins AL (2008) Network pharmacology: the next paradigm in
drug discovery. Nat Chem Biol 4:682–690
Hughes TR, Marton MJ, Jones AR, Roberts CJ, Stoughton R, Armour
CD, Bennett HA, Coffey E, Dai H, He YD, Kidd MJ, King AM,
Meyer MR, Slade D, Lum PY, Stepaniants SB, Shoemaker DD,
Gachotte D, Chakraburtty K, Simon J, Bard M, Friend SH (2000)
Functional discovery via a compendium of expression profiles.
Cell 102:109–126
Iorio F, Bosotti R, Scacheri E, Belcastro V, Mithbaokar P, Ferriero R,
Murino L, Tagliaferri R, Brunetti-Pierri N, Isacchi A, Di
Bernardo D (2010) Discovery of drug mode of action and drug
repositioning from transcriptional responses. Proc Natl Acad Sci
USA 107:14621–14626
Isik Z, Baldow C, Cannistraci CV, Schroeder M (2015) Drug target
prioritization by perturbed gene expression and network infor-
mation. Sci Rep 5:17417
Iwata M, Sawada R, Iwata H, Kotera M, Yamanishi Y (2017)
Elucidating the modes of action for bioactive compounds in a
cell-specific manner by large-scale chemically-induced tran-
scriptomics. Sci Rep 7:40164
Jahchan NS, Dudley JT, Mazur PK, Flores N, Yang D, Palmerton A,
Zmoos AF, Vaka D, Tran KQ, Zhou M, Krasinska K, Riess JW,
Neal JW, Khatri P, Park KS, Butte AJ, Sage J (2013) A drug
repositioning approach identifies tricyclic antidepressants as
inhibitors of small cell lung cancer and other neuroendocrine
tumors. Cancer Discov 3:1364–1377
Jin G, Wong ST (2014) Toward better drug repositioning: prioritizing
and integrating existing methods into efficient pipelines. Drug
Discov Today 19:637–644
Kanehisa M, Sato Y, Furumichi M, Morishima K, Tanabe M (2018)
New approach for understanding genome variations in KEGG.
Nucleic Acids Res 47:D590–D595
Kannan L, Ramos M, Re A, El-Hachem N, Safikhani Z, Gendoo DM,
Davis S, Gomez-Cabrero D, Castelo R, Hansen KD, Carey VJ,
Morgan M, Culhane AC, Haibe-Kains B, Waldron L (2016)
Public data and open source tools for multi-assay genomic
investigation of disease. Brief Bioinform 17:603–615
Kelloff GJ, Sigman CC (2012) Cancer biomarkers: selecting the right
drug for the right patient. Nat Rev Drug Discov 11:201–214
Kim RS, Goossens N, Hoshida Y (2016) Use of big data in drug
development for precision medicine. Expert Rev Precis Med
Drug Dev 1:245–253
Kondo T, Imamura K, Funayama M, Tsukita K, Miyake M, Ohta A,
Woltjen K, Nakagawa M, Asada T, Arai T, Kawakatsu S, Izumi
Y, Kaji R, Iwata N, Inoue H (2017) iPSC-based compound
screening and in vitro trials identify a synergistic anti-amyloid
beta combination for Alzheimer’s disease. Cell Rep
21:2304–2312
Kwon OS, Oh E, Park JR, Lee JS, Bae GY, Koo JH, Kim H, Choi YL,
Choi YS, Kim J, Cha HJ (2015) GalNAc-T14 promotes
metastasis through Wnt dependent HOXB9 expression in lung
adenocarcinoma. Oncotarget 6:41916–41928
Kwon O-S, Lee H, Kong H-J, Park JE, Lee W, Kang S, Kim M, Kim
W, Cha H-J (2018) In silico Drug Repositioning of bortezomib
to reverse metastatic effect of GALNT14 in lung cancer. https://
doi.org/10.1101/394163
Lamb J (2007) The connectivity map: a new tool for biomedical
research. Nat Rev Cancer 7:54–60
Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ,
Lerner J, Brunet JP, Subramanian A, Ross KN, Reich M,
Hieronymus H, Wei G, Armstrong SA, Haggarty SJ, Clemons
PA, Wei R, Carr SA, Lander ES, Golub TR (2006) The
connectivity map: using gene-expression signatures to connect
small molecules, genes, and disease. Science 313:1929–1935
Lantermann AB, Chen D, Mccutcheon K, Hoffman G, Frias E, Ruddy
D, Rakiec D, Korn J, Mcallister G, Stegmeier F, Meyer MJ,
Sharma SV (2015) Inhibition of casein kinase 1 alpha prevents
acquired drug resistance to erlotinib in EGFR-mutant non-small
cell lung cancer. Cancer Res 75:4937–4948
Lazo JS, Sharlow ER (2016) Drugging undruggable molecular cancer
targets. Annu Rev Pharmacol Toxicol 56:23–40
Lee ES, Son DS, Kim SH, Lee J, Jo J, Han J, Kim H, Lee HJ, Choi
HY, Jung Y, Park M, Lim YS, Kim K, Shim Y, Kim BC, Lee K,
Huh N, Ko C, Park K, Lee JW, Choi YS, Kim J (2008)
Prediction of recurrence-free survival in postoperative non-small
cell lung cancer patients by using an integrated model of clinical
information and gene expression. Clin Cancer Res 14:7397–7404
Lee MO, Moon SH, Jeong HC, Yi JY, Lee TH, Shim SH, Rhee YH,
Lee SH, Oh SJ, Lee MY, Han MJ, Cho YS, Chung HM, Kim KS,
Cha HJ (2013) Inhibition of pluripotent stem cell-derived
teratoma formation by small molecules. Proc Natl Acad Sci
USA 110:E3281–E3290
Lee H, Kang S, Kim W (2016) Drug repositioning for cancer therapy
based on large-scale drug-induced transcriptional signatures.
PLoS ONE 11:e0150460
Lee BK, Tiong KH, Chang JK, Liew CS, Abdul Rahman ZA, Tan
AC, Khang TF, Cheong SC (2017) DeSigN: connecting gene
expression with therapeutics for drug repurposing and develop-
ment. BMC Genom 18:934
Liberzon A, Subramanian A, Pinchback R, Thorvaldsdottir H,
Tamayo P, Mesirov JP (2011) Molecular signatures database
(MSigDB) 3.0. Bioinformatics 27:1739–1740
Novac N (2013) Challenges and opportunities of drug repositioning.
Trends Pharmacol Sci 34:267–272
Pushpakom S, Iorio F, Eyers PA, Escott KJ, Hopper S, Wells A, Doig
A, Guilliams T, Latimer J, Mcnamee C, Norris A, Sanseau P,
Cavalla D, Pirmohamed M (2018) Drug repurposing: progress,
challenges and recommendations. Nat Rev Drug Discov 18:41
Reddy AS, Zhang S (2013) Polypharmacology: drug discovery for the
future. Expert Rev Clin Pharmacol 6:41–47
Rees MG, Seashore-Ludlow B, Cheah JH, Adams DJ, Price EV, Gill
S, Javaid S, Coletti ME, Jones VL, Bodycombe NE, Soule CK,
Alexander B, Li A, Montgomery P, Kotz JD, Hon CS, Munoz B,
Liefeld T, Dancik V, Haber DA, Clish CB, Bittker JA, Palmer
M, Wagner BK, Clemons PA, Shamji AF, Schreiber SL (2016)
Correlating chemical sensitivity and basal gene expression
reveals mechanism of action. Nat Chem Biol 12:109–116
Rhee SY, Wood V, Dolinski K, Draghici S (2008) Use and misuse of
the gene ontology annotations. Nat Rev Genet 9:509–515
Sam E, Athri P (2019) Web-based drug repurposing tools: a survey.
Brief Bioinform 20:299–316
Sanchez-Vega F, Mina M, Armenia J, Chatila WK, Luna A, La KC,
Dimitriadoy S, Liu DL, Kantheti HS, Saghafinia S, Chakravarty
D, Daian F, Gao Q, Bailey MH, Liang WW, Foltz SM,
Shmulevich I, Ding L, Heins Z, Ochoa A, Gross B, Gao J, Zhang
H, Kundra R, Kandoth C, Bahceci I, Dervishi L, Dogrusoz U,
Zhou W, Shen H, Laird PW, Way GP, Greene CS, Liang H, Xiao
Y, Wang C, Iavarone A, Berger AH, Bivona TG, Lazar AJ,
O.-S. Kwon et al.
123
Hammer GD, Giordano T, Kwong LN, Mcarthur G, Huang C,
Tward AD, Frederick MJ, Mccormick F, Meyerson M, Cancer
Genome Atlas Research, Van Allen EM, Cherniack AD, Ciriello
G, Sander C, Schultz N (2018) Oncogenic signaling pathways in
the cancer genome atlas. Cell 173(321–337):e10
Sawada R, Iwata M, Tabei Y, Yamato H, Yamanishi Y (2018)
Predicting inhibitory and activatory drug targets by chemically
and genetically perturbed transcriptome signatures. Sci Rep
8:156
Scannell JW, Blanckley A, Boldon H, Warrington B (2012)
Diagnosing the decline in pharmaceutical R&D efficiency. Nat
Rev Drug Discov 11:191–200
Sirota M, Dudley JT, Kim J, Chiang AP, Morgan AA, Sweet-Cordero
A, Sage J, Butte AJ (2011) Discovery and preclinical validation
of drug indications using compendia of public gene expression
data. Sci Transl Med 3:96ra77
Subramanian A, Narayan R, Corsello SM, Peck DD, Natoli TE, Lu X,
Gould J, Davis JF, Tubelli AA, Asiedu JK, Lahr DL, Hirschman
JE, Liu Z, Donahue M, Julian B, Khan M, Wadden D, Smith IC,
Lam D, Liberzon A, Toder C, Bagul M, Orzechowski M, Enache
OM, Piccioni F, Johnson SA, Lyons NJ, Berger AH, Shamji AF,
Brooks AN, Vrcic A, Flynn C, Rosains J, Takeda DY, Hu R,
Davison D, Lamb J, Ardlie K, Hogstrom L, Greenside P, Gray
NS, Clemons PA, Silver S, Wu X, Zhao WN, Read-Button W,
Wu X, Haggarty SJ, Ronco LV, Boehm JS, Schreiber SL,
Doench JG, Bittker JA, Root DE, Wong B, Golub TR (2017) A
next generation connectivity map: L1000 platform and the
FIRST 1,000,000 profiles. Cell 171(1437–1452):e17
Tsherniak A, Vazquez F, Montgomery PG, Weir BA, Kryukov G,
Cowley GS, Gill S, Harrington WF, Pantel S, Krill-Burger JM,
Meyers RM, Ali L, Goodale A, Lee Y, Jiang G, Hsiao J, Gerath
WFJ, Howell S, Merkel E, Ghandi M, Garraway LA, Root DE,
Golub TR, Boehm JS, Hahn WC (2017) Defining a cancer
dependency map. Cell 170(564–576):e16
Van Noort V, Scholch S, Iskar M, Zeller G, Ostertag K, Schweitzer C,
Werner K, Weitz J, Koch M, Bork P (2014) Novel drug
candidates for the treatment of metastatic colorectal cancer
through global inverse gene-expression profiling. Cancer Res
74:5690–5699
Viswanathan VS, Ryan MJ, Dhruv HD, Gill S, Eichhoff OM,
Seashore-Ludlow B, Kaffenberger SD, Eaton JK, Shimada K,
Aguirre AJ, Viswanathan SR, Chattopadhyay S, Tamayo P,
Yang WS, Rees MG, Chen S, Boskovic ZV, Javaid S, Huang C,
Wu X, Tseng YY, Roider EM, Gao D, Cleary JM, Wolpin BM,
Mesirov JP, Haber DA, Engelman JA, Boehm JS, Kotz JD, Hon
CS, Chen Y, Hahn WC, Levesque MP, Doench JG, Berens ME,
Shamji AF, Clemons PA, Stockwell BR, Schreiber SL (2017)
Dependency of a therapy-resistant state of cancer cells on a lipid
peroxidase pathway. Nature 547:453–457
Wang K, Sun J, Zhou S, Wan C, Qin S, Li C, He L, Yang L (2013)
Prediction of drug-target interactions for drug repositioning only
based on genomic expression similarity. PLoS Comput Biol
9:e1003315
Winter GE, Radic B, Mayor-Ruiz C, Blomen VA, Trefzer C,
Kandasamy RK, Huber KVM, Gridling M, Chen D, Klampfl T,
Kralovics R, Kubicek S, Fernandez-Capetillo O, Brummelkamp
TR, Superti-Furga G (2014) The solute carrier SLC35F2 enables
YM155-mediated DNA damage toxicity. Nat Chem Biol
10:768–773
Xia X, Wong ST (2012) Concise review: a high-content screening
approach to stem cell research and drug discovery. Stem Cells
30:1800–1807
Zhang J, Xing Z, Ma M, Wang N, Cai YD, Chen L, Xu X (2014)
Gene ontology and KEGG enrichment analyses of genes related
to age-related macular degeneration. Biomed Res Int
2014:450386
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
In silico drug repositioning: from large-scale transcriptome data to therapeutics
123