Prediction forPotential Risk of Drug Combination

stalliongrapevineBiotechnology

Oct 1, 2013 (3 years and 8 months ago)

79 views

Prediction forPotential Risk of Drug Combination
Wang Xiaoqian,
Zhang Jianing, Tian Luyi

Department of Bioinformatics, Zhejiang University School of Life Science, 388 Yu Hang Tang
Road, Hangzhou 310058, PR China


Abstract

Background

Drug combination makes a novel way for research and development of new drugs. In the study of
drug combination, the potential risk is a significant aspect to be taken into consideration.
Nowadays, most reported studies about side effect of drug combination

are based on wet labs,
which leaves many things to be discovered. Side effect results from the drug binding to an
off
-
target, which is related with a certain unwanted biological process.

Methodology/Principal Findings

In this article, we describe a
bioinf
ormatics

method
to predict side effect of drug combination.
The
prediction
is based

on the

biological process related with each component in the combination
.
We
built the relationship between side effect and biological process, with which we can make
predi
ction for drug combination side effect.
We
constructed a database with the data we collected
from Drugbank, Gene Ontology, SIDER, Connectivity Map and our prediction for drugs in
DCDB.

Conclusions/Significance

The database provides necessary information of

drug and side effect. Besides, the prediction
resultscan serve

as
assistance in further studyof side effect
and
offer reference frame for
new
candidate
drug combination
for experimental verification.


Keywords

Drug combination, side effect, biological pr
ocess, prediction, database


1.

Introduction

Expenditures on pharmaceuticals have grown
faster than other major components of the
health care system since the late 1990s
[1]
.

On
average, the cost for one new drug is about
one billion dollars per year
[1, 2]
. However,
new registered compounds showed declining
current since 1990s even the expenditure grew
fast.


The study of

drug combination makes a novel
way for research and development (R&D) of
new drugs. Different aspects of the research of
drug combination have attracted scientists’
interests
[3]
, among which the study

of side
effect played an important role. Recent studies
for drug combination are mostly based on
experiments
[4
-
6]
, some studies use
bioinformatics method to make analysis
[7]
,
which still leaves many things to be
discovered.


A drug binds to its receptor and affects to the
biological process, which brings about its
healing effect. However, if it binds to an
unwanted target, side effect will come
along.Recently, there were
researches
identifying targets of known drugs based on
side effect information
[8]
. For example, in the
research of Campillos group
[9]
, they

used
phenotypic side
-
effect similarities to infer
whether two drugs share a target.
In the
research of
Keiser

group
[10]
, they found new
target for drugs with the similarity in chem
ical
structure.


Researches about binding target had got good
results.However, apart from that, the influence
of biological process has to be taken into
consideration.As is in the research of
Lee,
S.
group
[8, 11]
, they
made t
he multi
-
level
network (the proc
ess
-
drug
-
side effect network)
built

by merging the drug
-
biological process
network a
nd the drug
-
side effect network,
with whichthey built up

the relationship
between side effect and biological process.


In this article, w
e described a bioinformatics
method for predicting the side effect of drug
combination

based on biological process
.
With data colle
cted from Drugbank, Gene
Ontology, SIDER, Co
nnectivity Map and the
additional file in
Lee, S.
group’s research
[11]
,
we constructed relations between side effect
and biological process. Then we made our
prediction for drug combination side effect

based on the relations we get. With these
results, we built up a database called Drug
Side Effect Database, where there was basic
information for drug, side effect, and our
prediction results for combination in DCDB.


2.
Results and Discussions

2.1 Data P
rocessing

We use
d

connectivity map
and data from
Lee,
S.
group’s additional file
[11]

to const
ruct a
database

relating drug with its biological
process
. The connectivity map
is

a collection
of genome
-
wide

transcriptional expression
data from cultured human cells treated with
bioactive small molecules
[12]
.


With data from SIDER, we constructed the
relationship between drug and side
effect.SIDER is t
he r
ecent development of a
database of the side effects,
making

a first step
to provide the relationship between drugs and
side effects.

This relationship we got can be
connected with the former one using
Drugbank ID as a primary key.


2.2

Making Prediction

With these two relations above, we built the
side effect
-
drug
-
biological process network
(Fig. 1), this network comes from
Lee,
S.
group’s paperwork
[8]
.
As is shown below,
drug informationcan be found in Drugbank;
side effect and the drug & side effect
relationship information can be found in
SIDER; biological process information found
in GO and si
de effect & biological process
found in connectivity map. For every drug,
since we knew its relevant side effect and
biological process, we could find the
duplicated ones and build link between them
(Fig. 2).
Pre
dictions were based on
relationship between
side effect and biological
process. Drug combination was got from
DCDB
[13]
, which was a database for
combination of drugs. For every combination,
we found each component’s biological process,
where the

duplicated one represented their
common ground in action. With the duplicated
biological process, we could find the relevant
side effect and then got the prediction for side
effect of drug combination (Fig. 3).


Fig. 1


Fig.
2Fig. 3

2.3 Database Construction

With the data we collected and processed,
along with the prediction results for the side
effect of drug combination, we built a
database named Drug Side Effect Database
(Fig. 4). The database was c
onstructed with
php, MySQL and Apache, using javascript to
embed script.


Fig. 4

In the database, there is query for drug basic
information, including
Drugbank_ID, Generic
name, Brands, Molecular weight,

PKA, CID
(Side effect ID number), Side effect,
Indication. There is also query for side effect
information, including CID, Side effect,
GO_ID, process name, namespace and define.
The query results can be downloaded as
required.


Besides, there is uploadin
g function for drug
information (Fig. 5), where the DB_ID, CID
and GO_ID are necessary for uploading.
Downloading is also available for the basic
information and our prediction result (Fig. 6,
7).


Fig. 5
Fig. 6


Fig. 7

3.

Conclusions

We have developed a
bioinformatics
method

to make prediction for side effect of drug
combinations.The results are based on
biological process analysis, which can be used
in further studyof side effect
and
offer
reference frame for
new candidate
drug
combination
for experimental verification
.


4.

Mat
erial and Methods

Thedrug data was be collected from Drugbank
(
www.drugbank.ca/
), where we downloaded
the file drugbank.xml.zip and extracted the
information of all approved drugs with perl.
The features we selected
wereDrugbank_ID,
Generic name, Brands, IUPAC NAME,
Chemical formula, Molecular weight, PKA,
ALOGPS_Solubility, Indication
. There were
36from 1424drug information in all
.


The side effect data was collected from
SIDER(http://
sideeffects.embl.de
), where we
d
ownloaded
adverse_effects_raw.tsv.gz

and
extracted information with perl. The features
selected were Drugbank_ID, Drug name,
SIDER_ID, Side effect. There were 168 from
740 information in all. With DB_ID as a key,
the table of drug and side effect got conne
cted
and made a new table for drug basic
information.

The biological process information was
downloaded from Gene Ontology
(
www.geneontology.org
). We got the file of
OBO v1.2.obo, extracting the features of
GO_ID
, Name, Namespace and Define. There
were 2202 from 36778 GO sum in all.


For the relationship between drug and
biological process, we searched Connectivity
Map and downloaded
rankMatrix.txt.zip
. With
the assistance of
Lee, S.
group
’s additional
file
[11]
, webuilt the relationship between drug
and biological process with perl, and this gave
out a new table for side effect, including the
information of side effect and biological
process as well. Then
we did data mining to
find the relationship between side effect and
biological process.


Based on the results we had got, we made
prediction for combinations in DCDB
(
http://www.cls.zju.edu.cn/rlibs
). The
pre
diction was done by perl. For every
component in one combination, we found its
biological process and relevant side effect, and
then we looked for biological process that
occurring mostly and had the highest score(for
one certain biological process, some s
ide
effect is related with its upregulating, others
are related with its downregulating). Then we
put out the relevant side effect and it made the
prediction results. There are 153 predictions in
all.


References:


[1].

DiMasi, J.A., R.W.
Hansen and H.G.
Grabowski, The price of innovation: new
estimates of drug development costs. Journal of
Health Economics, 2003. 22(2): p. 151
-
185.


[2].

Adams, C.P. and V.V. Brantner, Estimating
The Cost Of New Drug Development: Is It
Really $802 Million?
Health Affairs, 2006. 25
(2): p. 420
-
428.


[3].

Chou, T., Theoretical Basis, Experimental
Design, and Computerized Simulation of
Synergism and Antagonism in Drug
Combination Studies. Pharmacological Reviews,
2006. 58 (3): p. 621
-
681.


[4].

Cardona, F., et
al., [Bisphosphonate
-
related
osteonecrosis of the jaws. An update]. An Sist
Sanit Navar, 2009. 32(3): p. 413
-
21.


[5].

Patsalos, P.N. and E. Perucca, Clinically
important drug interactions in epilepsy:
interactions between antiepileptic drugs and
other dru
gs. The Lancet Neurology, 2003. 2(8):
p. 473
-
481.


[6].

Domagala, J.M., Structure
-
activity and
structure
-
side
-
effect relationships for the
quinolone antibacterials. Journal of
Antimicrobial Chemotherapy, 1994. 33 (4): p.
685
-
706.


[7].

Chou, T., Drug Combi
nation Studies and
Their Synergy Quantification Using the
Chou
-
Talalay Method. Cancer Research, 2010.
70 (2): p. 440
-
446.


[8].

Lee, S., M. Song and D. Lee. Discovering
biological processes and side effects relationship
using the process
-
drug
-
side effect n
etwork. in
Proceedings of the ACM fourth international
workshop on Data and text mining in biomedical
informatics. 2010. Toronto, ON, Canada: ACM.


[9].

Campillos, M., et al., Drug Target
Identification Using Side
-
Effect Similarity.
Science, 2008. 321 (588
6): p. 263
-
266.

[10].

Keiser, M.J., et al., Predicting new
molecular targets for known drugs. Nature, 2009.
462(7270): p. 175
-
181.

[11].

Lee, S., et al., Building the
process
-
drug
-
side effect network to discover the
relationship between biological Processe
s and
side effects. BMC Bioinformatics, 2011.
12(Suppl 2): p. S2.

[12].

Lamb, J., et al., The Connectivity Map:
Using Gene
-
Expression Signatures to Connect
Small Molecules, Genes, and Disease. Science,
2006. 313(5795): p. 1929
-
1935.

[13].

Liu, Y., et al.,
DCDB: Drug combination
database. Bioinformatics, 2010. 26 (4): p.
587
-
588.


Acknowledgments:

We thank Dr. Chen Xin for his instructions;

Dr. Liu Yanbin for his help in finding a way
for predictions.

All this project is accomplished in
bioinformatics la
boratory 210 room and 402
room.