Computational personalized medicine in cancer research in the -omics data era
Research and Development on Information and Communication Technology
Computational Personalized Medicine in
Cancer Research in the -Omics Data Era
Duc-Hau Le, Quynh Diep Nguyen
School of Computer Science and Engineering, Thuyloi University, Hanoi, Vietnam
Correspondence: Duc-Hau Le, [email protected]
Communication: received 29 October 2019, revised 16 April 2020, accepted 6 May 2020
Digital Object Identifier: 10.32913/mic-ict-research.v2020.n1.899
Abstract: Omics data (e.g., genomics, transcriptomics, pro-
teomics, epigenomics, etc . . . ) generated from high-throughput
next-generation sequencers in the big human genome, and
cancer genome projects have changed the way to study
personalized medicine. In the future, personalized medicine
will not be limited to diagnosis and treatment based on a
few known disease-associated mutations on some genes, but
will rely on whole molecular characteristics of patients by
integrating their –omics data. In this study, we draw a big
picture of personalized medicine research in cancer research
of the –omics data era, including –omics databases, challenges
of data fusion to solve two major problems in personalized
medicine, i.e., personalized diagnosis and treatment. These
problems are approached as patient stratification and drug
response prediction based on the –omics data by computa-
tional methods.
1 → 2% of DNA directly encodes for proteins. The major
part (98 → 99% of DNA) are non-coding regions, which
do not directly synthesize the proteins but regulate the
synthesis process as well as other biological processes and
cellular functions. Therefore, epigenome (i.e., epigenomics
data) should be taken into account for the disease etiologies.
Moreover, cells function under chemical reactions between
metabolites; thus, metabolome (i.e., metabolomics data) is
also critical. Finally, cellular components such as genes,
proteins, and metabolites,. . . do not function alone, but they
interoperate in the form of biological networks to make the
cells robust against external stimulus or mutations; thus,
interactomics data is a final “piece of cake” for a big picture
of the –omics data.
In this study, we draw a big picture of personalized
medicine in cancer research of the –omics data era. Per-
sonalized medicine is a potential approach in biomedicine
with the goal is to find the right treatment (i.e., right
drug, right dose, and right time) for each patient based
on their biological characteristics. Different patients have
different biological characteristics in terms of –omics data;
thus, the development of the right methods for diagnosis
and treatment is a challenging task due to the diversity
and complexity of the –omics data. However, to fully
understand the biological characteristics of each patient,
their –omics data should be combined together in the right
ways. Therefore, besides introducing –omics databases, we
also discuss how to integrate the –omics data for solving
the two main problems (i.e., diagnosis and treatment) in
personalized medicine.
Keywords: Personalized medicine, omics data integration,
patient stratification, drug response prediction, computational
method.
I. INTRODUCTION
In biomedicine, with high-throughput technologies such
as next-generation sequencing (NGS), we are witnessing the
exponential growth of biomedical data (e.g., –omics data).
To date, it is widely accepted for most diseases that their
genetic etiology cannot be explained with one or a few
genes/variants, but thousands of whole-genome/variome
(i.e., genomics/variomics data). Besides, it is still not
enough if the only genome is considered, but also transcrip-
tome (e.g., transcriptomics data), because this data reflects
how much change in a genome can affect the expression of
a gene in different tissues. More importantly, how the gene
expression translated into proteins is a critical issue since
protein functions decide how a cell works. Therefore, be-
sides genomics and transcriptomics, proteomics data should
also be considered when studying the disease etiologies. It
should be noted that the central dogma in biology states
protein synthesis as DNA→RNA→Protein. However, only
II. THE –OMICS DATA
With the rapid development of NGS technologies, the
cost for sequencing is significantly going down, and the
–omics data are growing exponentially. Biomedical re-
search nowadays is not only based on the genomics data,
but also on other –omics data such as transcriptomics,
1
Research and Development on Information and Communication Technology
A proteome is all proteins in a cell at a specific time of
development. Two main problems that are widely studied
for proteins are protein structure and protein interaction
network. A total of 20 types of amino acids can be
translated by the protein synthesis process. In the struc-
tural view, a protein is a sequence of amino acids (aka
polypeptide), which can be in different shapes such as linear
(level 1) and fold (level 4). The structural features decide
the protein’s functions. For the second problem, physical
interactions between proteins by chemical bonds form the
protein interaction networks or protein complexes. These
also decide the functions of cells.
A metabolome is mentioned as small chemical com-
pounds such as endogenous and exogenous metabolites.
Figure 1. The –omics layers.
Figure 1 shows that, for humans, there are about 25, 000
genes in the genome, more than 100, 000 proteins, and
5, 000 metabolites [3]. Those data generated from large-
scale biomedical projects such as human genomes, cancer
genomes, and cell lines have yielded big–omics databases.
In the next section, we are going to introduce some of such
projects briefly.
proteomics, and metabolomics [1]. Figure 1 shows the cen-
tral dogma in biology with the corresponding –omics data
layers.
A genome is understood as whole genetic information
on DNA of a species. The human genome contains about
three billion base pairs (bp), which encode for about 25, 000
genes. Coding regions occupy approximately 1 → 2% of
the genome, and the remaining is non-coding regions [2].
Change in a genome (variants) can happen at a single
point, such as SNP (Single Nucleotide Polymorphism) or
short insertion and deletion (short indel). Those changes
can be detected by traditional techniques such as Sanger or
modern ones such as microarray and NGS. If the changes
happen in protein-coding regions of DNA, they can be
synonymous, missense, nonsense, and frameshift. When
they occur in the remaining regions, those changes may
affect the level of gene expression (e.g., up and down)
or RNA splicing process. Besides such small changes,
larger changes are also observed, such as structure variants
(SV), where the insertions and deletions are found in large
regions (100 − 1000 bp). In addition, copy number variant
(CNV), inversion, and translocation are also found. For
those changes, the microarray or NGS technologies are
required for the detection.
III. THE –OMICS DATABASE
The –omics databases are generated from large-scale
projects on the human genome, such as UK’s 100, 000
Genomes [4] and GenomeAsia 100K [5]. In addition,
disease-specific projects such as those for cancer genome
have largely contributed their data to the research commu-
nity. Here are some critical projects which are widely used
in cancer research.
The Center for Cancer Genomics of National Cancer
Institute in the US had launched a number of big projects
on cancers such as TCGA (The Cancer Genome Atlas) [6],
CGCI (Cancer Genome Characterization Initiative) and
TARGET (Therapeutically Applicable Research to Generate
Effective Treatments). These projects are now managed by
gdc.cancer.gov) to foster precise medicine in cancer. The
portal has deposited data for more than 40 cancer projects
on 61 organs, with 32, 555 samples [7]. A number of
–omics data has been generated from the projects such
as single point somatic/germline mutation, structural so-
matic mutations, gene/microRNA expression by microarray,
gene expression by RNA-Seq, and DNA methylation data.
Besides, the Catalogue Of Somatic Mutations In Cancer
project (COSMIC) is the largest and most comprehensive
database of cancer somatic mutations on human [8]. Major
data of COSMIC is collected and curated by experts.
It also provides 3D structures of the mutations as well
as related genes. In addition, more than 1, 000 mutation
profiles of cell lines are also accumulated. Finally, the
A transcriptome includes all RNAs (coding RNAs such
as mRNA and non-coding RNAs) in a cell. RNA reflects
how much genetic information (i.e., embedded in genes)
from DNA is expressed. To measure the level of gene
expression, microarray and RNA-seq are widely used. The
microarray technology is used to detect known transcripts
and isoforms with established and standard pipelines/tools.
Meanwhile, the RNA-Seq, as a modern technology, can be
used to detect de novo transcripts and isoforms as well as
structural variants. However, it also yields big and complex
data, thus requires more sophisticated pipelines/tools.
2
Vol. 2020, No. 01, September
Figure 2. Integration of the –omics data in personalized medicine problems.
International Cancer Genome Consortium (ICGC) aimed to
IV. INTEGRATION OF THE –OMICS DATA
define the genomes of 25, 000 primary untreated cancers,
as its data portal [9] makes global genomic data sharing
for cancer possible, providing the international community
with comprehensive genomic data for many cancer types.
It is obvious that the future of biomedical research will
mainly rely on the –omics data because the biological
characteristics of patients are fully considered. Therefore,
with the assumption that “More is better”, recent biomed-
ical research on the –omics data is trying to combine as
many –omics data and other clinical data as possible to
solve the two major problems in personalized medicine,
i.e., diagnosis (aka patient stratification) and treatment (aka
drug response prediction) (see next section) [1, 3, 12]. In
addition, the identification of causative biomarkers (aka
predictive features) is also addressed (Figure 2).
In parallel with the abovementioned projects on cancer
tumors, projects on testing drugs/compounds on cell lines
were also launched. The outcome of those projects are
databases of the cell line’s biological characteristics and
responses of the tested drugs/compounds. These data help
foster research on pharmacogenomics, which is the most
promising direction in personalized medicine. Two typical
projects are Cancer Cell Line Encyclopedia (CCLE) [10]
and Genomics of Drug Sensitivity in Cancer (GDSC) [11].
CCLE aims to identify –omics and pharmacological prop-
erties of cancer cell lines, then developing analysis tools
for predicting drug responses based on the –omics data.
A potential application of this is to stratify patients based
on their –omics and pharmacological characteristics. CCLE
provides the –omics data including single point, indel, and
structural mutations and the gene expression of more than
1, 000 cancer cell lines as well as the pharmacological
properties of 24 drugs. Similarly, GDSC freely published
those data for about 1, 000 cancer cell lines and 265
drugs/compounds.
Data integration is approached by machine learning-
based [13] network-based [14] methods. The machine
learning-based methods rely on unsupervised and super-
vised techniques. The typical unsupervised methods, such
as matrix factorization, aim at subtyping cell lines using
the –omics data with the final goal is stratifying patients.
Meanwhile, with supervised methods, labelled data (e.g.,
patients/cell lines with known drug responses or subgroups)
is needed for training prediction models. The trained mod-
els are then used to predict for patients/cell lines with
unknown drug responses and subgroups. In parallel, the
networkbased methods often rely on the functional rela-
tionships among biomedical entities in the same or different
3
Research and Development on Information and Communication Technology
Figure 3. Patient stratification using –omics data.
–omics layers [14].
(a small number of sample, but high dimension). This raises
a lot of problems requiring computational solutions [17].
Along with the advantages of integration, there are
also some challenges in integrating the whole genome,
transcriptome, and proteome data are also faced with. An
obvious problem is “highdimensional data” since thousands
to millions of genes/proteins and variants must be investi-
gated. Therefore, techniques in dimension reduction [15] or
feature selection [16] have been proposed to select predic-
tive features. Those features are also important biomarkers
for diseases. In contrast with the high dimensionality, the
number of samples is relatively small (i.e., often thousands).
Thus the most challenging problem when dealing with the
integration of largescale –omics data is “small 푛, large 푝”
In summary, the data integration aims to systematically
assess the biological characteristics of patients; thus, the
application of identifying cancer genes [18] and cancer
diagnosis [19] in personalized medicine is approached more
precisely. In the next sections, we introduce how the two
major problems in computational personalized medicine,
i.e., diagnosis (patient stratification) and treatment (drug
response prediction), are approached using the –omics data.
4
Vol. 2020, No. 01, September
Figure 4. Drug response prediction using –omics data.
V. PATIENT STRATIFICATION USING –OMICS DATA
network clustering techniques such as module/community
detection are often used. Those techniques can be applied
to different types of biological networks such as patient
similarity network [25], protein interaction network [26],
and gene network [27]. Recent methods have relied on both
matrix and network clustering techniques [28] or matrix
clustering techniques but guided by a network [23, 29].
In this section, we introduce how to stratify patients using
the –omics data. Cancer is selected for study as it is a
complex disease. The genetic etiologies are also diversified
among different cancers/subtypes of the same cancer. Even
cells of the same cancer tumor may also act differently
(Figure 3). The patient stratification has been studied based
on molecular subtyping for some types of cancer, such as
bladder urothelial carcinoma [20] and breast cancer [21].
Although many methods have been proposed, a number
of tools have been developed for the patient stratification
problem. However, different methods could lead to dif-
ferent subtypes. This is caused by the complexity of the
–omics data (i.e., small 푛, large 푝). In addition, the cancer
tumor is heterogeneous; thus, the molecular profiles (i.e.,
the –omics data) are different for even cells derived from
the same cancer types.
As the abovementioned of the rapid growth of the
–omics data, this fosters the development of computational
methods for patient stratification based on molecular pro-
files and towards clinical application [22]. The problem
is generally solved by the following four steps: i) data
preprocessing, ii) clustering, iii) supervised classification,
and iv) subtype characterization. For the first step, raw
–omics data from high-throughput sequencers or genotyper
are qualifiedly controlled, standardized, and transformed
into the right formats for the next steps. Then, the data
clustering is performed based on either genes/proteins or
samples. If these two types of data are used simultaneously,
then biclustering techniques are used. In addition, tri-
clustering techniques are used if other factors, such as time,
are considered. In the third step, supervised classification
techniques are used to determine which subtype/subgroup
a patient belongs to. Finally, the identified subtypes are
characterized by causative biomarkers, involved pathways,
and survival time.
VI. DRUG RESPONSE PREDICTION USING
–OMICS DATA
Besides the requirement of precisely stratifying patients
based on their –omics data, personalized medicine also
requires selecting the right drugs for the right treatment
for each patient. Patients with the same conditions/cancer
types may respond differently to the same drugs due to
their different molecular profiles. Thus, the prediction of
drug response using the –omics data is an important step
for selecting the right drugs (Figure 4).
Similar to the patient stratification problem, many meth-
ods have also been proposed for predicting drug response
for patients/cell lines [30]. The drug response is measured
by dose level to inhibit 50% of the disease’s bioactivity
(IC50), or they are under the dose-response curve (AUC).
They are both continuous values. Thus, the drug response
prediction is often approached by regression techniques.
However, response values can be categorized into some
levels, such as good response, no response, and bad side
effects (Figure 4); thus, it can be formulated as a classifi-
cation problem. The main difference between the two main
Two main approaches have been proposed for this prob-
lem, including i) machine learningbased and ii) network-
based methods. A common point of the two approaches is
that they both rely on the similarity between samples (i.e.,
the similarity of patient’s molecular profiles). The similar-
ity can be represented as similarity matrices in machine
learning-based methods or similarity networks in network-
based methods. By representing as similarity matrices,
clustering techniques on a matrix such as matrix factoriza-
tion are widely used [23, 24]. For network-based methods,
5
Research and Development on Information and Communication Technology
problems in personalized medicine is that the patient strati-
fication is usually based on tumor data from tumor/patient-
based projects such as TCGA. Meanwhile, the drug re-
sponse prediction uses cell line and drug response data from
drug trial projects such as CCLE and GDSC.
–omics data algorithmically faces with the “small 푛, large
푝” problem. The object (i.e., cancer) itself is a complex
disease, which is heterogeneous between cancer types and
even cells in the same tumor. In addition, unexpected
changes in characteristics of cell lines during culture may
limit the translation of research results on cell lines to
patients. Fortunately, many big human genome and disease
genome projects have been launched and freely published
the data for the research community. In parallel, state-of-
the-art techniques in computational sciences (e.g., artifi-
cial intelligence, statistics) have fostered the application
of computational methods to study problems in medicine.
This could open a brighter future for personalized medicine
in cancer research of the –omics data era. Personalized
medicine is a broad research area and application. Indeed,
besides biological characteristics of the patients, their clin-
ical data, environment, and lifestyles are also important
factors in tailoring the individual treatments. In addition,
personalized medicine approaches are not only limited to
cancers, but also be used to diagnose and treat other
disorders such as rare diseases, which are strongly linked to
molecular alterations. Furthermore, besides the abovemen-
tioned –omics data, metagenomics and metatranscriptomics
should also be worthy of studying personalized medicine
since there exist interactions between humans and the
microbiome.
Generally, machine learning- and network-based meth-
ods are often proposed for the drug response prediction.
Network-based methods are usually based on similarity net-
works of drugs and cell lines and local [31] or global [32]
graph traversal algorithms. In contrast to a few network-
based methods, many machine learning-based methods have
been proposed for the drug response prediction problem.
Indeed, a challenge was organized for research groups over
the world [33]. Interestingly, the winner over 44 submis-
sions is a method integrating the –omics data (including
single point mutation, structural mutation, gene expression
by microarray and RNA-Seq technologies, methylation, and
protein data) using multiple kernel learning technique [34].
Other methods also show that response prediction for
multiple drugs simultaneously achieve better performance
than that for a single drug, because functional and structural
similarity among drugs is taken into account [34, 35]. In
addition, gene expression data is more dominant than the
others [33]. Finally, until now, computational methods for
the drug response prediction have been proposed mostly
for cancer cell lines. Thus to translate them to clinical
application, a recent method has built the prediction model
using the data from cell lines in GDSC, then use the
built model for predicting drug response for patients in
TCGA [36].
ACKNOWLEDGMENT
This research is funded by Vietnam National Foundation
for Science and Technology Development (NAFOSTED)
under grant number 102.01-2017.14.
VII. CONCLUSIONS
REFERENCES
Nowadays, the rapid development of high-throughput
technologies and large-scale genome projects have gen-
erated a large amount of the –omics data (i.e., the
–omics era). This has changed the ways to computationally
approach the problems in personalized medicine. To fully
understand the biological characteristics of patients, their
molecular profiles at the –ome scale has been studied.
Thus, the –omics data has been integrated into compu-
tational methods to solve the problems in personalized
medicine. The two major problems in medicine (i.e., di-
agnosis and treatment) are formulated as two problems in
computational space (i.e., patient stratification and drug
response prediction, respectively). Although current studies
of the two problems target different objects, i.e., the patient
stratification mainly focuses on patient data from the patien-
t/tumor projects; meanwhile, the drug response prediction
mostly works with artificial patients/tumors (i.e., cell lines).
However, they are both personalized based on molecular
profiles of each patient/tumor/cell line. Integration of the
[1] J. C. Venter, M. D. Adams, E. W. Myers, P. W. Li, R. J.
Mural, G. G. Sutton et al., “The sequence of the human
genome,” science, vol. 291, no. 5507, pp. 1304–1351, 2001.
[2] C. Manzoni, D. A. Kia, J. Vandrovcova, J. Hardy, N. W.
Wood, P. A. Lewis et al., “Genome, transcriptome and
proteome: The rise of omics data and their integration in
biomedical sciences,” Briefings in Bioinformatics, vol. 19,
no. 2, pp. 286–302, 2018.
[3] J. Harrow, A. Frankish, J. M. Gonzalez, E. Tapanari,
M. Diekhans, F. Kokocinski et al., “GENCODE: The refer-
ence human genome annotation for the ENCODE project,”
Genome Research, vol. 22, no. 9, pp. 1760–1774, 2012.
[4] R. P. Horgan and L. C. Kenny, “Omic technologies: Ge-
nomics, transcriptomics, proteomics and metabolomics,” The
Obstetrician & Gynaecologist, vol. 13, no. 3, pp. 189–195,
2011.
[5] G. N. Samuel and B. Farsides, “The UK’s 100,000 Genomes
Project: Manifesting policymakers’ expectations,” New Ge-
netics and Society, vol. 36, no. 4, pp. 336–353, 2017.
[6] GenomeAsia100K Consortium et al., “The GenomeAsia
100K Project enables genetic discoveries across Asia,” Na-
ture, vol. 576, no. 7785, pp. 106–111, 2019.
[7] J. N. Weinstein, E. A. Collisson, G. B. Mills, K. R. M. Shaw,
B. A. Ozenberger, K. Ellrott et al., “The cancer genome
6
Vol. 2020, No. 01, September
atlas pan-cancer analysis project,” Nature Genetics, vol. 45,
no. 10, p. 1113, 2013.
[8] S. Deorowicz, A. Danek, and M. Niemiec, “GDC 2: Com-
pression of large collections of genomes,” Scientific Reports,
vol. 5, p. 11565, 2015.
“Molecular subtyping of cancer: Current status and moving
toward clinical applications,” Briefings in Bioinformatics,
vol. 20, no. 2, pp. 572–584, 2019.
[24] M. Hofree, J. P. Shen, H. Carter, A. Gross, and T. Ideker,
“Network-based stratification of tumor mutations,” Nature
methods, vol. 10, no. 11, pp. 1108–1115, 2013.
[25] Z. He, J. Zhang, X. Yuan, Z. Liu, B. Liu, S. Tuo et al.,
“Network based stratification of major cancers by integrating
somatic mutation and gene expression data,” PloS One,
vol. 12, no. 5, 2017.
[26] B. Wang, A. M. Mezlini, F. Demir, M. Fiume, Z. Tu,
M. Brudno et al., “Similarity network fusion for aggregating
data types on a genomic scale,” Nature Methods, vol. 11,
no. 3, pp. 333–337, 2014.
[27] F. Zhang, C. Ren, K. K. Lau, Z. Zheng, G. Lu, Z. Yi et al., “A
network medicine approach to build a comprehensive atlas
for the prognosis of human cancer,” Briefings in Bioinfor-
matics, vol. 17, no. 6, pp. 1044–1059, 2016.
[9] S. A. Forbes, D. Beare, P. Gunasekaran, K. Leung, N. Bindal,
H. Boutselakis et al., “COSMIC: Exploring the world’s
knowledge of somatic mutations in human cancer,” Nucleic
Acids Research, vol. 43, no. D1, pp. D805–D811, 2015.
[10] J. Zhang, J. Baran, A. Cros, J. M. Guberman, S. Haider,
J. Hsu et al., “International Cancer Genome Consortium
Data Portal – a one-stop shop for cancer genomics data,”
Database, vol. 2011, 2011.
[11] J. Barretina, G. Caponigro, N. Stransky, K. Venkatesan,
A. A. Margolin, S. Kim et al., “The Cancer Cell Line
Encyclopedia enables predictive modelling of anticancer
drug sensitivity,” Nature, vol. 483, no. 7391, pp. 603–607,
2012.
[12] W. Yang, J. Soares, P. Greninger, E. J. Edelman, H. Light-
foot, S. Forbes et al., “Genomics of Drug Sensitivity in Can-
cer (GDSC): A resource for therapeutic biomarker discovery
in cancer cells,” Nucleic Acids Research, vol. 41, no. D1, pp.
D955–D961, 2012.
[13] S. Huang, K. Chaudhary, and L. X. Garmire, “More is better:
Recent progress in multi-omics data integration methods,”
Frontiers in Genetics, vol. 8, p. 84, 2017.
[28] M. Le Morvan, A. Zinovyev, and J.-P. Vert, “NetNorM:
Capturing cancer-relevant information in somatic exome
mutation data with gene networks for cancer stratification
and prognosis,” PLoS Computational Biology, vol. 13, no. 6,
p. e1005573, 2017.
[29] C. R. Planey and O. Gevaert, “CoINcIDE: A framework
for discovery of patient subtypes across multiple datasets,”
Genome Medicine, vol. 8, no. 1, pp. 1–17, 2016.
[14] Y. Li, F.-X. Wu, and A. Ngom, “A review on machine learn-
ing principles for multi-view biological data integration,”
Briefings in Bioinformatics, vol. 19, no. 2, pp. 325–340,
2018.
[15] J. Yan, S. L. Risacher, L. Shen, and A. J. Saykin, “Net-
work approaches to systems biology analysis of complex
disease: Integrative methods for multi-omics data,” Briefings
in Bioinformatics, vol. 19, no. 6, pp. 1370–1381, 2018.
[16] C. Meng, O. A. Zeleznik, G. G. Thallinger, B. Kuster,
A. M. Gholami, and A. C. Culhane, “Dimension reduction
techniques for the integrative analysis of multi-omics data,”
Briefings in Bioinformatics, vol. 17, no. 4, pp. 628–641,
2016.
[17] F. Rohart, B. Gautier, A. Singh, and K.-A. Lê Cao,
“mixOmics: An R package for ‘omics feature selection and
multiple data integration,” PLoS Computational Biology,
vol. 13, no. 11, p. e1005752, 2017.
[18] M. Bersanelli, E. Mosca, D. Remondini, E. Giampieri,
C. Sala, G. Castellani et al., “Methods for the integration
of multi-omics data: Mathematical aspects,” BMC Bioinfor-
matics, vol. 17, no. S2, p. S15, 2016.
[19] C. Dimitrakopoulos, S. K. Hindupur, L. Ha¨fliger, J. Behr,
H. Montazeri, M. N. Hall et al., “Network-based integration
of multi-omics data for prioritizing cancer genes,” Bioinfor-
matics, vol. 34, no. 14, pp. 2441–2448, 2018.
[20] Q. Zhao, X. Shi, Y. Xie, J. Huang, B. Shia, and S. Ma,
“Combining multidimensional genomic measurements for
predicting cancer prognosis: Observations from TCGA,”
Briefings in Bioinformatics, vol. 16, no. 2, pp. 291–303,
2015.
[30] G. Yu, X. Yu, and J. Wang, “Network-aided Bi-Clustering
for discovering cancer subtypes,” Scientific Reports, vol. 7,
no. 1, pp. 1–15, 2017.
[31] F. Azuaje, “Computational models for predicting drug re-
sponses in cancer research,” Briefings in Bioinformatics,
vol. 18, no. 5, pp. 820–829, 2017.
[32] N. Zhang, H. Wang, Y. Fang, J. Wang, X. Zheng, and
X. S. Liu, “Predicting anticancer drug responses using a
dual-layer integrated cell line-drug network model,” PLoS
Computational Biology, vol. 11, no. 9, 2015.
[33] D.-H. Le and V.-H. Pham, “Drug response prediction by
globally capturing drug and cell line information in a hetero-
geneous network,” Journal of Molecular Biology, vol. 430,
no. 18, pp. 2993–3004, 2018.
[34] J. C. Costello, L. M. Heiser, E. Georgii, M. Go¨nen, M. P.
Menden, N. J. Wang et al., “A community effort to assess
and improve drug sensitivity prediction algorithms,” Nature
Biotechnology, vol. 32, no. 12, pp. 1202–1212, 2014.
[35] M. Ammad-ud din, S. A. Khan, D. Malani, A. Muruma¨gi,
O. Kallioniemi, T. Aittokallio et al., “Drug response predic-
tion by inferring pathway-response associations with kernel-
ized Bayesian matrix factorization,” Bioinformatics, vol. 32,
no. 17, pp. i455–i463, 2016.
[36] D. Le and D. Nguyen-Ngoc, “Multi-task regression learning
for prediction of response against a panel of anti-cancer
drugs in personalized medicine,” in Proceedings of the In-
ternational Conference on Multimedia Analysis and Pattern
Recognition, Ho Chi Minh City, Vietnam, Apr. 2018.
[21] Q. Mo, F. Nikolos, F. Chen, Z. Tramel, Y.-C. Lee, K. Hayashi
et al., “Prognostic power of a tumor differentiation gene
signature for bladder urothelial carcinomas,” Journal of the
National Cancer Institute, vol. 110, no. 5, pp. 448–459,
2018.
[22] M. Cortet, A. Bertaut, F. Molinié, S. Bara, F. Beltjens,
C. Coutant et al., “Trends in molecular subtypes of breast
cancer: Description of incidence rates between 2007 and
2012 from three French registries,” BMC Cancer, vol. 18,
no. 1, p. 161, 2018.
[23] L. Zhao, V. H. Lee, M. K. Ng, H. Yan, and M. F. Bijlsma,
7
Research and Development on Information and Communication Technology
Le Duc Hau obtained his PhD degree
in Bioinformatics from University of Ul-
san, Republic of Korea in 2012. He is
now leading the Department of Compu-
tational Biomedicine, Vingroup Big Data
Institute, VietNam. He has been focus-
ing on proposing computational methods
for disease- and drug-related problems in
Quynh Diep Nguyen obtained her PhD
degree in Information Technology from the
Institute of Information Technology - The
Vietnam Academy of Science and Tech-
nology in 2015. She is a lecturer at the
School of Computer Science and Engi-
neering, Thuyloi University. She has been
focusing on computational methods for re-
personalized medicine, especially on identification of disease-
associated biomarkers, prediction of drug targets and response.
In parallel, he has been developing bioinformatics tools. So far,
he hasmore than fifty papers published in well-recognized journals
and conferences, nearly a half of those are in ISI-indexed journals.
In addition, he has been a member of program committees
and reviewer of several international conferences/journals. More-
over, he is a principal investigator and a key member of some
national/ministry-level projects. Specially, he is the principal in-
vestigator of the biggest genome project in Vietnam (i.e., building
databases of genomic variants for Vietnamese population). Finally,
he has been collaborating with some well-recognized international
research institutes.
constructing the metabolic networks. So far, she has more than
fifteen papers in journals and conferences published . Moreover,
she is a member of some national/ministry-level projects which re-
search on computational methods for uncovering latent knowledge
from high-throughput biological data.
8
8 trang
8540
Free
Bạn đang xem tài liệu "Computational personalized medicine in cancer research in the -omics data era", để tải tài liệu gốc về máy hãy click vào nút Download ở trên
File đính kèm:
- computational_personalized_medicine_in_cancer_research_in_th.pdf
