Accelerating Computation of DNA Multiple Sequence Alignment in Distributed Environment

®

http://iaesjournal.com/online/index.php/TELKOMNIKA/index

[

USER

i

r---··---·------------;
i

ＡＭｾＺＮ＠

Accelerating Computation of DNA Multiple Sequence Alignment in
Distributed Environment

ＢｾＧﾰＮ［Ｉ＠

. ｾＮ＠

•

i
.,.I

ｖｬｾｷ＠

• Sub.crlbe / U11>ub>uob"

DOl: 10.11591/telkomnika .v12il2 .6572:

Abstrdct
ｍｵｬ｜Ｑｰｾ＠

ＮｾｬＧｑｊｮ･＠
al1=;inmerrt (MS;\) ｴｾ＠ '-' t\'_etmiqul' for firdm.;i ｾＱｭｩｬｊ［ｲｴＬ＠
"' mJ- "1' lt:e ｲｾｩ［ｯｳ･､＠ｰ

t(· ｈｾ･ｬｲＧＡ＾ｴ＠

Ti-u5 tecM"l.Ub is Yer'{ •n'•IJ01tant to ｾｵｰｯｲｴ＠
ｦｲ｡ｧｭ･ｮｴｾ＠
b1nninq. Th!! ｾｩｲｮｄＱＶｴ＠
dtna1nic ｐＧ＼ｬｾｭｩｮ＿＠
tH.hn1q•Jl' 1> v"rr

.:.emputation of Star ｍｾｴＱｰｬ･＠
Ｓｾﾷ｜Ｑ･ｮ｣＠
ｴ｜ｨＮ［ｭｾｮ＠
•Js•nsi MH5111JO ｐ｡ｳｾＱｮｧ＠
lnterfaces ＨｍｾｉｩＮ＠
The
ＬＮＧｊｬｵｾｴ･､＠
by ｲＮｬＡｉｃ｜ｾｴ［ｮ＼ｊ＠
ｾｰＮｩ･､ｵ＠
E1p•mrnent '"as conducted ｵｾｭ＠
{..1 ｾＬｱｵ･ｮ｣Ｚ＠
of ｾｏ＠
bp
yiold"'d t·v randomly out from Nf"a-rencx. ｾｯｴｩｵﾷｭ＼Ｚ･＠

of ｃＭｬ［ｯｩｮ･ｲｍｸ｣ｴｭ｡ｈﾷｾＡｂ＠
ｴ｡ｾ･ｮ＠
lrom N".":51 ＨｾｊｴｬｯｲＱｩｉ＠
ＢＮＺ･ｬ｜ｴｾｲ＠
for ｂＱｴ｜｣ｾｮＮＬｬｯＧ＠
!r,forrMt•on). TM 1\'•;••ll's ＤｨｏｾＧ､＠
lhat ｴｨｾ＠
ｰｲｯｾｊ､＠
tochr.ique could \'lbtaio ｾｬＧ､ｵ＠
thr Pn
Illustration of data distribution of DNA sequence with MPI can be seen in Figure 6. In
this research, comunication of MPI used point to point communication with a blocking send and
receives operations. Parallelization scheme for MSA is to assign a computer as the data divider
and other computers as data processors. A data divider called rank 0 distributed sequence
pairs using MP/_Send() to other computer as data processors (Figure 7). Data processors
which were symbolized by the rank 1... n received sequence pairs using MPl_Recv() (Figure 8).
Next, each data processors conducted pairwise sequence alignment to compute pairwise
similarity scores in parallel. The calculation results of each data processor were transmitted to
data divider (rank 0) by using MP/_Send(). Data divider with rank 0 received the similarity
scores using MP/_Recv() command and completing the MSA process by selecting a Star

sequence and realigning all sequence to the Star sequence

8282

•

ISSN: 2302-4046
I
I

------,

I
I

I
r----------IMPl_Send(K1)/ MPl_Recv(K1)--------'f-

,--

ｾＭＬ＠

Dividersl :
data
I
: ｬＰｴｾﾱｑｩｊＵＱＺＭｍｐ｟ｒ･｣ｶＨｓｭ｡ｲｹ＠
\

I ｩｦｊｬＦｲｮＺｾ＠ｪ
I ｾＧＢＭ

MPl_Recv{Similarity Score K1) f MPl_Send(Simi!arity Score K1)

IF""'=wi

I

:
Ｕ Ｑｈｬｴｊ＠

I
I

I
ｾＭｍｐｬ｟ｓ･ｮ､ＨｋＲＩ＠

I
1

Rank 1

:

'

'-,--l- - "

I

I

i

i

l. . . . . . . . . . . . . . . . . . . . . . . .

MPl_Seod(Ko) /

L-----------------·----·--·· MPl_Recv(Similarity Score Kn) I MPl_Send(Slmllar!ty Score Kn)···

Rank2

I
I

ｍｐｬ｟ｒ･ｯｶＨｋＩＭﾷｾ＠
0

·------------:-

ｬ［ｈｩｾＱＬ＠

D

... - _,RjWl\Jl - ...

Figure 6. Illustration the distribution of sequence data with MPI
llMPI initialization
MPl_lnit(&argc, &argv);
MPl_Comm_rank (MPl_COMM_WOR.LD, &rank);//Computer init_ialization
MPl_Comm_size (MPl_COMM_WORLD, &p_size);//lnitialize the.number of computers
If rank== 0 {
c = ((js*js)'js)/2; II Calculate the many pairwise sequences
p_bts =.p_size-1; II Limit Number of Computers
p=1; II Limit Number of computers for processing data (other than rank 0)
for (i=1; i