J. Biomedical Science and Engineering, 2008, 1, 127-132
Published Online August 2008 in SciRes. http://www.srpublishing.org/journal/jbise JBiSE
Exploring diversity of different classificatory hu-
man tandem repeats
Zhong-Yu Liu1, Xu-FengLi1, Xian-PingDing1, Bo Liu1, Hao Peng2&Yi Yang1
1College of Life Sciences, Sichuan University, Chengdu, 610064, China. 2Institute of Microbiology, Chinese Academy of Sciences, Beijing, 100101, China.
Correspondence should be addressed to Z.Y. Liu ([e-mail]).
ABSTRACT
Tandem repeats (TRs) are associated with dis-
ease genes, play an important role in evolution
and are important in genomic organization and
function. Much research has been done on de-
scriptions of properties of tandem repeats, such
as copy-number, period, etc, and correlation be-
tween mutations within tandem repeats and dis-
ease. This project aims to detect some differ-
ences w hich may exist within the features of dif-
ferent tandem repeats associated w ith disease in
human whole-genome. The features of tandem
repeats associated with diabetes genes were
compared to the counterparts of non-diabetes
disease genes.
Availability: TRbase is available at http://www.trbase2.cn/
1. INTRODUCTION
Repetitive DNA sequences have been identified in large
quantities in both eukaryotic and prokaryotic genomes [2].
In some cases they can account for a large portion of the
genome, for example, in the human genome they have
been known to contribute around 40-50% of the total
DNA sequence. The existence of repetitive DNA in pro-
karyotes is limited, but it is found widely distributed
throughout a large variety of eukaryotes, and can be
found throughout the genome in both protein coding re-
gions and inergenic regions. One of the reasons tandem
repeats are of great interest is because of their nature to
expand and contract unpredictably. It has been reported
that microsatellites are biased towards expanding in
length [15]. It has also been reported that repeats within
coding regions appear to have some kind of constraint
hindering their expansion, whereas tandem repeats in
untranslated regions do not appear to have these con-
straints, therefore much higher copy numbers of these
repeats are often present [16].
Tandem repeats are known to have high mutability
rates which cause differences in repeat length between
lineages; this implies that these high mutability rates con-
tribute to overall genome evolution. The frequent changes
in tandemly repeated regions within genomes, although
caused by mutation, are more specifically assumed to be
due to slippage during DNA replication or unequal
alignment during DNA recombination [17]. However
these processes are not thought to be the sole cause of the
differences observed between lineages in the divergence
of tandem repeats within their genomes [6, 17]. Within
this current study these concepts have been expanded by
investigating the influence that disease may have had on
the evolution of tandem repeats that either cause the dis-
ease or are within disease genes. From an evolutionary
standpoint, the sequences with tandem repeats have sev-
eral interesting feature [9], Formation of the repeating
sequences is an error-prone process, with mutations in
genomic DNA repeats occurring far more frequently than
the background rate of point mutations [2]. This suggests
that repetitive sequences evolve more quickly than non-
repetitive sequences. In general, one of the most interest-
ing features of prokaryotic and eukaryotic genomes (both
coding and non-coding regions) is the presence of rela-
tively short perfect tandemly repeated DNA sequences.
These repeated DNA sequences are distributed almost at
random throughout the genome [7, 8, 13]. Much research
indicates that at least ten kinds of inherited neurological
disease including Huntington’s disease and spinocerebel-
lar ataxia, as well as many less serious diseases such as
epilepsy and deafness, are known to be the product of
tandem repeat expansions (http://tandem.bu.edu) [10, 18].
Diabetes mellitus is characterized by abnormally high
levels of sugar (glucose) in the blood. The most common
forms of diabetes are type 1 diabetes (5%), which is an
autoimmune disorder, and type 2 diabetes (90%), which
is associated with obesity (http://www.ncbi.nlm.nih.gov/
books/bookres.fcgi/diabetes/pdf_ch1.pdf). The vast ma-
jority of diabetes cases fall into the categories of type 1
and type 2 diabetes. However, up to 5% of cases have
other specific causes and include diabetes that results
from the mutation of a single gene. About 18 regions of
the genome have been linked with influencing type 1 dia-
betes risk. These regions, each of which may contain sev-
eral genes, have been labeled IDDM1 to IDDM18. In rare
forms of diabetes, mutations of one gene can result in
disease. However, in type 2 diabetes, many genes are
thought to be involved. “Diabetes genes” may show only
as subtle variation in the gene sequence, and these varia-
tions may be extremely common. The difficulty lies in
linking such common gene variations, known as single
nucleotide polymorphisms (SNPs), with an increased risk
of developing diabetes. One method of finding the diabe-
tes susceptibility genes is by whole-genome linkage stud-
ies.
SciRes Copyright © 2008

128 Z. Y. Liu et al. / J. Biomedical Science and Engineering 1 (2008) 127-132
SciRes Copyright © 2008 JBiSE
This study specifically concentrated on the statistical
comparisons of some features (copy number, percentage
match, period, indels, and %GC ) of tandem repeats asso-
ciated with disease; its aim was to find some differences
among features of tandem repeats between one specific
set of disease genes (in this case, these are diabetes genes)
and other disease genes. If those differences existed, fur-
ther research would be carried out to explore the relation-
ship between those diversities. This paper will present an
inchoative detection of those features in terms of com-
parison of features of tandem repeats associated with dia-
betes genes and non-diabetes disease genes selected ran-
domly from all disease genes (excluding diabetes genes)
on chromosome 12.
2. IMPLEMENTATION AND METHODS
2.1. TRbase extension
All the disease genes and relevant data about the features
of tandem repeats of disease genes were retrieved from a
web-accessible relational tandem repeats database
TRbase that relates tandem repeats to gene locations and
disease genes of the human genome [1]. DNA sequences
and annotations were retrieved for the completed chromo-
somes 4, 5, 6, 14, 16, 18, 19, 20, 21, and 22 [1]; however
this project required data on all those disease genes and
their relevant information in whole-human genomes, in-
dicating that TRbase need extending to all human chro-
mosomes prior to data preparations. DNA sequences and
annotations of the remaining chromosomes (1, 2, 3, 7, 8,
9, 10, 11, 12, 13, 15, 17, X and Y) were downloaded from
GenBank (http://www.ncbi.nlm.nih.gov/mapview/maps.
cgi?TAXID=9606&MAPS=ideogr,cntg-r,ugHs,genes&C
HR=1), all tandem repeats were detected using the TRF
program (version 3.01; Benson, 1999) with parameters as
in [1] applied to DNA sequences extracted from GenBank
in the FASTA format using the Seqret program of
EMBOSS [12].
2.2. Representative chromosome generation
It has been proposed that vertebrate genomes, including
human, are made up of compositionally homogeneous
DNA segments based on G+C content [16]. These re-
gions, known as isochores, have been studied diabetes
genes using density gradient centrifugation on mechani-
cally sheared DNA in the range of 50-100 kb [16] since
their discovery in the 1970s [17]. Isochores are biologi-
cally interesting due to the association between increasing
G+C content and high gene density [18, 19, 20 ].
According to Bernardi’s theories, there are five fami-
lies of isochores, each having a different level of cytosine
and guanine (C and G, respectively) as described in Ta-
ble 1. There are two G+C-poor isochore families L1 and
L2 that make up approximately 60% of the humangenome.
The isochore family L1 is defined to be regions corre-
sponding to less than 37% G+C content; L2 is defined to
be regions containing between 37% and 41% G+C. The
Table 1. Isochore classifications. Isochore classifications are the
GC ranges for each of the five isochore classifications as defined
by Bernardi (2000). ANote that the L1 and L2 isochore classes
together represent 60% of the human genome.
Isochore Class Percent (G+C) Percent of Genome
L1 0-37
L2 37-41
60.0A
H1 41-46 24.0
H2 41-46 7.5
H3 53-100 4.7
isochore family H1 forms 24% of the human genome and
corre-sponds to regions between 41% and 46% G+C. The
other G+C rich isochore family H2 forms 7.5% of the
human genome and corresponds to those regions contain-
ing between 46% and 53% G+C. The final isochore fam-
ily, H3 forms almost 5% of the genome and corresponds
to those very G+C rich regions which are greater than
53% G+C.
Since the overall composition of the human genome is
approximately 60% AT and 40% GC, the L1 and L2
families correspond to isochore regions containing less
than average G+C content while the H1, H2, and H3
families correspond to isochore regions containing higher
than average G+C content. Columns 1-4 in Table 2 were
created using these guidelines to split the histograms for
75 kb fragments for the various chromosomes into densi-
ties of 60%, 84%, and 91.5%, which would theoretically
find the isochore boundaries. The first three columns in
Table 2 were retrieved from an article of [14]. X1, X2 ….
X9 stands for the substantial value of each feature in each
row in Table 2. In each row, Y = (X1- mean) 2 + (X2-
mean) 2…. + (X9- mean) 2 (mean is the value in bottom
row in Table 2). The lowest value in column 11 corre-
sponded to chromosome 12, which indicated that chro-
mosome 12 was the most representative one in all human
chromosomes.
2.3. Diabetes genes and control variables genera-
tion and parameters selection
All the diabetes genes were detected by the Search Dis-
ease Information at TRbase website (http://trbase2.cn) [1],
which provided information of their location within the
human genome. The tandem repeats in diabetes genes
were identified at the advanced composite search page at
the TRbase website. When selecting diabetes genes, the
parameter settings are: High Stringency detection parame-
ter was used; the tandem repeat copy number ranged from
1.9 to 13086.4; the percentage match to the consensus
sequence was kept above 70%; the tandem repeat unit
length varied from 1 to 1000. The preceding parameter
settings were applied to tandem repeats within intron,
exon and intergene regions. All diabetesgenes were iden
tified one by one using the above parameters, and each
result formed a table with many columns. The col-

Z. Y. Liu et al. / J. Biomedical Science and Engineering 1 (2008) 127-132 129
SciRes Copyright © 2008 JBiSE
umns(excluding labeled gene, copy Number, pe-
riod,%Match, Indels and Consensus) would be deleted in
the combined table. A new column composed of the %
G+C of each consensus was inserted into the table with
deletion of consensus column. Alternatively, diabetes
genes data were gained from the MySQL TRbase, using
MySQL commands equivalent to the processes stated
above.
The non-diabetes disease genes [with five features
(copy Number, period, % Match, indels and % G+C)] for
two control groups were selected randomly from all dis-
eases genes (excluding four diabetes genes) on chromo
some 12; the number of diseases genes selected randomly
in the two control groups was equal to the number of dia-
betes genes in the whole genome. The method to prepare
the diabetes gene table was applied to generate control 1
Table 2. Representative chromosome dependent on all those data. Columns 1-4 shows that those data associated with G+C contents
and the breakpoints of 60%, 84%, and 91.5% indicate breakpoints for the defined isochore classes L2-H1, H1-H2, and H2-H3 (Bernardi,
2000). The data in columns 5-10 retrieved (July 23, 2006) from http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=genome&cmd= Re-
trieve&dopt= Overview&list_uids=2). Y = (X1- mean) 2 + (X2- mean) 2…. + (X9- mean) 2 (mean is the value in bottom row in Table
2).The number of genes and protein coding, the nucleotide length, structural RNAs, Pseudogenes, contigs of each chromosome were
obtained from Entrez Genome in NCBI (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=genome&cmd=Retrieve&dopt=Over-
view&list_uids=1.
Isochore Boundary locations based on totalercent of all frag-
ments
Some Features of each chromosome retrieved from Entrez ge-
nome in NCBI
Chromosome
60% of all
fragment
L2-H1
Boundary
(X1)
84% of all
fragment H1-
H2 Boundary
(X2)
91.5%of all
fragment
H2-H3
Boundary
(X3)
Gene
den-
sity
(/Mb)
(X4)
Pro-
tein
coding
den-
sity
(/Mb)
(X5)
Struc-
tural
RNAs
(X6)
Pseud
o
genes
(X7)
Con-
tigs
(X8)
%co
ding
(X9) Y
1 44% 49% 51% 11.2313.65180 364 39 1
4.32E+0
4
2 44% 47% 49% 7.68 8.52 49 255 18 1 4043.4
3 41% 47% 49% 7.38 8.47 39 223 5 0 1545.2
4 40% 43% 45% 6.08 6.25 52 207 14 0 366.41
5 41% 44% 46% 7.08 8.26 48 188 6 0 335.33
6 39% 43% 45% 8.91 8.96 197 302 9 1
2.94E+0
4
7 46% 51% 52% 9.28 10.2979 240 12 1 2376.7
8 42% 45% 49% 7.01 7.82 47 151 10 0 2567.7
9 47% 53% 54% 8.62 9.74 46 218 39 1 1670.6
10 44% 48% 49% 8.08 10.1833 157 18 1 2346.5
11 46% 52% 55% 13.6913.8567 371 8 1
1.15E+0
4
12 44% 48% 50% 10.2411.8 55 195 7 1 105.03
13 41% 44% 47% 4.87 4.43 39 119 5 0 6306.9
14 43% 51% 55% 11.478.55 102 248 1 0 4524.5
15 43% 46% 47% 9.58 11.17115 152 11 1 4334
16 47% 51% 55% 12.4716.6357 132 5 1 3928.4
17 49% 52% 54% 18.3122.4797 146 10 2 3563.5
18 41% 44% 46% 7.75 6.23 6 79 5 0 16587
19 51% 54% 55% 25.4529.2395 133 4 3 5274.3
20 47% 50% 53% 11.4813.7920 127 6 1 6416.8
21 50% 55% 56% 7.82 9.2 11 71 4 0 17912
22 50% 54% 56% 15.2116.0622 98 9 1 10970
X 40% 43% 45% 8.68 9.28 69 287 17 0 8778.3
Y 39% 42% 43% 5.57 2.65 21 184 17 0 2060.6
mean 43% 48% 51% 9.68 11.1464
193.6
3 11.63 1 7.92E+0
3

130 Z. Y. Liu et al. / J. Biomedical Science and Engineering 1 (2008) 127-132
SciRes Copyright © 2008 JBiSE
Figure 1. Frequency distributions of copy number of tandem repeats in three data group: Diabetes, Control 1, Control 2.
Figure 2. Frequency distributions of percentage of match of tandem repeats in three data group: Diabetes, Control 1, Control 2.
Figure 3. Frequency distributions of period of tandem repeats in three data group: Diabetes, Control 1, Control 2.
Figure 4. Frequency distributions of indels of tandem repeats in three data group: Diabetes, Control 1, Control 2.

Z. Y. Liu et al. / J. Biomedical Science and Engineering 1 (2008) 127-132 131
SciRes Copyright © 2008 JBiSE
Figure 5. Frequency distributions of %G+C of tandem repeats in
three data group: Diabetes, Control 1, Control 2.
and control 2 tables in which the tandemrepeats from
control 1 and control 2 groups respectively were identi-
fied at TRbase website.
After preparation of the three tables (diabetes genes,
control 1, control 2), distributions, chi-square test and
independent-sample t-test were performed for the data
in the three tables.
3. RESULTS AND DISCUSSION
3.1. Quantitative comparison between tandem
repeats in diabetes genes and non-diabetes dis-
ease gene
In order to identify whether differences exist in quantity
between tandem repeat-containing genes and non-
tandem repeat-containing genes of diabetes and non-
diabetes disease genes, Table 3 was created.
χ2 tests were performed between (1) diabetes genes
and control 1; (2) diabetes genes and control 2; (3) con-
trol 1 and control 2. The results of the χ2 tests are re-
spectively: χ2 (1)=10.5, p(1)=0.01; χ2 (2)=3.4 ,
Table 3. Comparison of the 3 group data.
ChromosomeTR genes Non-TR
genes
Total
Diabetes
genes
All 9 38 47
Control 1 12 23 24 47
Control 2 12 17 30 47
Total 49 92 141
Table 4. P-value of t-test. p >0.05 means not significantly differ-
ent; p < 0.05 means significantly different.
Copy Num-
ber
%match Period Indels %G+C
Diabetes vs
control 1
0.080 0.000 0.000 0.000 0.056
Diabetes vs
control 2
0.128 0.000 0.530 0.000 0.422
Control 1 vs
control 2
0.830 0.139 0.001 0.540 0.557
p(2)=0.1; χ2 (3)=2.12 , p(3)=0.2. The result between
diabetes genes and control 1 is inconsistent with the re-
sult between dia-betes genes and control 2, which indi-
cates that the quantitative distribution of TRs versus non-
TRs in diseases genes is irregular.
3.2. Property of tandem repeats in diabetes genes
and non-diabetes diseases genes
Frequency distributions were plotted for the five features
of tandem repeats. In Figure 1-5 are respectively the
histograms of frequency distributions of the five features
of tandem repeats within the diabetes genes, control1
and control 2 data groups. In Figure 1 as well as Figure
5, the frequency distributions of experimental (diabetes)
genes, contro1 1 and control 2 are very similar; the three
frequency distributions of period in Figure 3 are not
obviously different from each other. In the Figure 2 and
Figure 4, the key two items of %match and indels of
tandem repeats in diabetes genes differ from in control
1and control 2 genes. This means that the significant
differences exist between the data in diabetes genes and
non-diabetes disease genes.
3.3. Independent sample t-test
The data were compared pairwise between each feature
of tandem repeats in diabetes genes, control 1 and con-
trol 2, Using independent sample t-tests. The results
(shown in Table 4) of t-test manifest that only percent
age match and indels of tandem repeats have significant
differences between diabetes genes and non-diabetes
disease genes.
4. CONCLUSION AND FUTURE PER-
SPECTIVE
TRbase extended provides a platform to study the asso-

132 Z. Y. Liu et al. / J. Biomedical Science and Engineering 1 (2008) 127-132
SciRes Copyright © 2008 JBiSE
ciations between disease genes and previously uncharac-
terized tandem repeats in whole human genomes. In all
features (copy number, percentage match, period, indels,
%G+C) of tandem repeats associated with disease genes,
statistically significant differences only exist for %match
and indels features of tandems repeats associated with
different disease genes. Currently, just a very prelimi-
nary research work has been done in mining of those
differences, further investigations are being conducted,
for example, correlations, regressions and modelling of
the differences will be performed, and use machine
learning methods to train and test the model.
ACKNOWLEDGEMENTS
The authors wish to express sincere appreciation to Dr. A.-M. Patch for
instruction in extending TRbase. In addition, special thanks are due to
Dr S. J. Aves, whose critical eye, and enlightened mentoring were
instrumental and inspiring. I wish to acknowledge my gratitude to Dr Z.
R. Yang, whose instructions and inspiration to me were helpful
throughout the whole project. I also thank Robin Batten for his assis-
tance in source data support.
REFERENCES
[1] Buard, J. and Vergnaud, G. (1994) Complex recombination events
at the hypermutable minisatellite CEB1 (D2S90). EMBO J. 13,
3203-3210.
[2] Van Belkum, A., Scherer, S., and Verbrugh, H. (1998) Short-
sequences DNA repeats in prokaryotic genomes, Microbiol.
Mol.Biol.Rev. 62, 275-293.
[3] Rubinsztein. D.C., Amos, B, and Cooper, G. (1999) Microsatellite
and trinucleotide-repeat evolution: evidence for mutational bias and
different rates of evolution in different lineages. Phil. Trans. R. Soc.
Lond. B 354, 1095-1099.
[4] Sutherland, G..R.and Richards, R.I. (1995) Simple tandem DNA
repeats and human genetic disease. Proc. Natl. Acad. Sci. USA 92
3636-3641.
[5] Tóth,G., Gaspari,Z. and Jurka,J. (2000) Microsatellites in different
eukaryotic genomes: survey and analysis. Genome Res., 10, 967–
981.
[6] Hancock, J.M., Worthey, E.A. and Santibez-Koref, M.F. (2001) A
Role for Selection in Regulating the Evolutionary Emergence of
Disease-Causing and Other Coding CAG Repeats in Humans and
Mice. Mol. Biol. Evol. 18, 1014-1023.
[7] Kajava .A.V (2001) Review: Proteins with repeated sequence –
structural prediction and modeling. Jour. Stru. Biol. 134, 132-144.
[8] Huang C., Lin Y., Yang Y., Huang S. and Chen C. (1998). The
telomeres of Streptomyces chromosomes contain conserved palin-
dromic sequences with potential to form complex secondary struc-
tures. Mol. Microbiol. 28, 905–916.
[9] Richard G. F., Hennequin C., Thierry A. and Dujon B. (1999) Tri-
nucleotide repeats and other microsatellites in yeasts. Res. Micro-
biol. 150, 589–602.
[10] Heslop-Harrison J. S. (2003( Tandemly repeated DNA sequences
and centromeric chromosomal regions of Arabidopsis species.
Chromosome Res. 11, 241–253.
[11] Lalioti M. D., Scott H. S., Buresi C., Bottani A., Norris M. A.,
Malafosse A. and Antonarakis S. E. (1997) Dodecamer repeat ex-
pansion in cystatin B gene in progressive myoclonus epilepsy. Na-
ture 386, 847–852.
[12] Wren J. D., Forgacs E., Fondon J. W., Pertsemlidis A., Cheng S. Y.
and Gallardo T. et al. 2000 Repeat polymorphisms within gene re-
gions: phenotypic and evolutionary implications. Am. J. Hum.
Genet. 67, 345–356..
[13] Boby T.., Patch A.M., Aves S.J. (2005) TRbase: a database relat-
ing tandem repeats to disease genes for the human genome. Bioin-
formatics, 21(6):811-6.
[14] Benson,G. (1999) Tandem Repeats Finder: a program to analyze
DNA sequences.Nucleic Acids Res., 27, 573–580.
[15] Rice. p., Longdnen, I. and Bleasby, A. (2000) Mini- and microsa-
tellite expansions: the recombination connection. EMBO Rep., 1,
122-126.
[16] Bernardi, G. (1993) "The isochore organization of the human
genome and its evolutionary history – a review."Gene, 135:57-66.
[17] Macaya, G., Thiery, J.P., Bernardi, G. (1976) "An approach to the
organization of eukaryotic genomes at a macromolecular level."
Journal of Molecular Biology,108(1): 237-254.
[18] Mouchiroud, D., D'Onofrio, G., Aissani, B, Macaya, G., Gautier, C.
Bernardi, G. (1991) "The distribution of genes in the human ge-
nome." Gene, 100:181-187.
[19] Gardiner,K. (1996) "Base composition and gene distribution:
critical patterns in mammalian genome organization." Trends in
Genetics, 12(12):519-524.
[20] Zoubak, S., Clay, O., Bernardi, G. (1996) "The gene distribution of
the human genome." Gene, 174:95-102.
[21] Rouchka, E.C., States, D.J. (2002) .Compositional Analysis of
Homogeneous Regions in Human Genomic DNA. Technical Re-
port. Washington University Department of Computer Science,
WUCS-2002-2