Journal of Theoretical Biology 267 (2010) 95–105

Contents lists available at ScienceDirect

Journal of Theoretical Biology

journal homepage: www.elsevier.com/locate/yjtbi

A study of entropy/clarity of genetic sequences using

metric spaces and fuzzy sets
D.N. Georgiou a, T.E. Karakasidis b,, Juan J. Nieto c, A. Torres d
a
University of Patras, Department of Mathematics, 265 00 Patras, Greece
b
University of Thessaly, Department of Civil Engineering, 383 34 Volos, Greece
c
Departamento de Análisis Matemático, Facultad de Matemáticas, Universidad de Santiago de Compostela, 15782, Spain
d
Departamento de Psiquiatrı́a Radiologı́a y Salud Pública, Facultad de Medicina, Universidad de Santiago de Compostela, 15782, Spain

a r t i c l e in fo abstract

Article history: The study of genetic sequences is of great importance in biology and medicine. Sequence analysis and
Received 21 January 2010 taxonomy are two major fields of application of bioinformatics. In the present paper we extend the
Received in revised form notion of entropy and clarity to the use of different metrics and apply them in the case of the Fuzzy
22 July 2010
Polynuclotide Space (FPS). Applications of these notions on selected polynucleotides and complete
Accepted 6 August 2010
Available online 11 August 2010
genomes both in the I12  k space, but also using their representation in FPS are presented. Our results
show that the values of fuzzy entropy/clarity are indicative of the degree of complexity necessary for
Keywords: the description of the polynucleotides in the FPS, although in the latter case the interpretation is
Polynucleotides slightly different than in the case of the I12  k hypercube. Fuzzy entropy/clarity along with the use of
Fuzzy sets
appropriate metrics can contribute to sequence analysis and taxonomy.
Metric spaces
& 2010 Elsevier Ltd. All rights reserved.
Entropy
Clarity

1. Introduction Fuzzy Polynucleotide Space (FPS) based on the principle of the

Bioinformatics is a relatively new discipline (see Jamshidi
et al., 2001; Morgenstern, 2002; Paun et al., 1998; Percus, 2002;
Tang, 2000) where Mathematics play an important role in the
analysis of genetic sequences. The genetic material of living
organisms consist of nucleic acids DNA and RNA. The analysis of
the genetic material is of great importance for diagnosis and
taxonomy reasons. In this course there are two basic strategies
that are commonly used: (a) sequence analysis, i.e. determination
of the building blocks of a nucleic acid (nucleotides) and their
order in the molecular chain, and (b) sequence comparison used
to identify the degree of difference/similarity between polynuclo-
tides, e.g in order to identify similarity with known viruses.
DNA and RNA are made of triplets XYZ of codons each of them
having the possibility to be one of four nucleotides {U, C, A, G} in
the case of DNA and {T, C, A, G} in the case of RNA (A ¼Adenine,
C¼Cytosine, G ¼Guanine, T¼Thymine, U¼Uracil) (Freland and
Hurst, 1998). Sadegh-Zadeh (see Sadegh-Zadeh, 2000) showed
that the genetic code can be represented in a 12-dimensional
space because a triplet codon XYZ has a 3  4 ¼ 12 dimensional
fuzzy code (a1,y,a12) and it is a point in the 12-dimensional fuzzy
polynucleotide space ½0,112 as a subspace of the real space
½0,112 . Sadegh-Zadeh (see Sadegh-Zadeh, 2000) introduced the
 Corresponding author.
E-mail address: thkarak@uth.gr (T.E. Karakasidis).
fuzzy hypercube (Kosko, 1992). In this notation a polynucleotide
consisting of a sequence of k triplets XYZ is a point in a I12  k
space. However, Torres and Nieto (see Torres and Nieto, 2003)
mapped a polynucleotide on a I12 space by considering the
frequencies of the nucleotides at the three base sites of a codon in
the coding sequence. In that work using a metric motivated by
publications of Lin (1997) and Sadegh-Zadeh (see Sadegh-Zadeh,
2000), they calculated distances between nucleotides. They also
applied their algorithm for the comparison of complete genomes
(for example M. tuberculosis and E. coli). Further work has been
recently performed using the idea of Nieto et al. (see Nieto et al.,
2006) in which the influence of several metrics have been
examined. The advantages of this methodology are:
(a) one can compare polynucleotides of very big length in a very
efficient computationally way and
(b) one can apply the algorithm in order to compare polynucleo-
tides of different length as it is the case for genomes of
different organisms.
We point that metrics play an important role on computational
biology. Different metrics have been used to study secondary
structures (see Moulton et al., 2000) or biopolyment contact
structures (see Liabres and Rossello, 2004).
It is very important to be in a position to determine how close
two genetic sequences are since there are many important biological
and medical implications (see Stephen and Freeland, 2008;

0022-5193/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jtbi.2010.08.010
96 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105
Kedarisetti et al., 2006; DasGupta et al., 1998; Chechetkin, 2003;
Chou, 1995, 2000b; Foster et al., 1999; Gusev et al., 1999; Jiang et al.,
2002; Liben-Nowell, 2001; Li et al., 2001; Mocz, 1995). The
biological distance among the 20 amino acids can be calculated
according to their classification results. Since the concept of pseudo
amino acid composition was proposed by Chou (2001), many efforts
have been made trying to use various quantities to represent the 20
native amino acids in order to better reflect the sequence-order
effects through the vehicle of pseudo amino acid composition
(PseAA), along with work in order to choose effective properties for
such procedures (Kawashima et al., 1999; Kawashima and Kanehisa,
2000; Trinquier and Sanejouand, 1998). In an earlier paper (Chou,
2000a), the physicochemical distance among the 20 amino acids
(Schneider and Wrede, 1994) was adopted to define PseAA.
Subsequently, some investigators used complexity measure factor
(Xiao et al., 2005a), some used the values derived from the cellular
automata (Xiao et al., 2005b,c, 2006a,b), some used hydrophobic
and/or hydrophilic values (Wolfenden, 2007; Zhang et al., 2001;
Chou, 2005b; Feng, 2002; Wang et al., 2006, 2004; Gao et al., 2005;
Chen et al., 2006a; Kurgan et al., 2007; Kurgan and Chen, 2007) and
some were through Fourier transform (Liu et al., 2005; Perez-
Montoto et al., 2009; Guo et al., 2006), as well as trough cellular
automaton approach (Xiao et al., 2009b) The pseudo amino acid
composition was originally introduced to improve the prediction
quality for protein subcellular localization and membrane protein
type (Chou and Cai, 2005, 2006; Chou, 2001; Chou and Elrod, 1999a,
2002, 2003; Shen and Chou, 2009a, b), as well as for enzyme
functional class (Chou, 2005b; Chou and Elrod, 1999b). Work using
pseudo amino acid composition has also been performed (Xiao et al.,
2008a,b, 2009a; Zhang et al., 2008; Lin and Pan, 2001). The pseudo
amino acid composition can be used to represent a protein sequence
with a discrete model yet without completely losing its sequence-
order information (Chou and Shen, 2007a,b,d), and hence is
particularly useful for analyzing a large amount of complicated
protein sequences by means of the taxonomic approach. Actually, it
has been widely used to study various protein attributes, such as
protein structural class (Chen et al., 2006a,b; Lin and Li, 2007a; Ding
et al., 2007; Gu and Chen, 2009; Homaeian et al., 2007), protein
subcellular localization (Chou and Shen, 2008, 2007a,b,Shen and
Chou, 2007a), protein subnuclear localization (Shen and Chou,
2005a,b; Mundra et al., 2007) protein submitochondria localization
(Du and Li, 2006), protein oligomer type (Chou and Cai, 2003),
conotoxin superfamily classification (Mondal et al., 2006; Lin and Li,
2007b) membrane protein type (Liu et al., 2005; Shen and Chou,
2005a; Wang et al., 2006; Shen et al., 2006; Chou and Shen, 2007c)
apoptosis protein subcellular localization (Chen and Li, 2007a,b)
enzyme functional classification (Chou, 2005a; Chou and Cai, 2004;
Zhou et al., 2007; Shen and Chou, 2007b) protein fold pattern (Shen
and Chou, 2006), and signal peptide (Chou et al., 2006; Chou and
Shen, 2007e; Shen and Chou, 2007c). Recent research works on the
extension of these kind of parameters in the form of Markov Chain
invariants of 2D graph or networks representation of aminoacid,
DNA, and RNA sequences to codify pseudo-aminoacid and pseudo-
nucleotide bases composition (Aguero-Chapin et al., 2008; Gonza-
lez-Diaz et al., 2007c, 2007b, 2007a; Aguero-Chapin et al., 2006;
Vilar et al., 2009; Georgiou et al., 2009) as well as more complex
work such as Xiao and Lin, 2009, Xiao et al., 2010. The reader can
also consult some recent reviews which made a discussion of many
of these previous results (Gonzalez-Diaz et al., 2008; Chou, 2009; Lin
et al., 2009).
In the present paper we present some new results concerning
the notions of entropy and clarity of a nucleotide that can be used
in order to estimate the fuzziness of a polynucleotide. We
compare the results with that obtained in Sadegh-Zadeh (2000).
We note that it is possible to compare sequences using a
minimum entropy principle (Sadovsky, 2003). More precisely
we focus on the use of different metrics in the calculation of the
entropy and clarity of a polynucleotide in conjunction with the
use of FPS which can be used in order to reduce the information
necessary for the representation of large polynucloetides.
The structure of the paper is as follows. In Section 2 we present
the notion of the Fuzzy Polynucleotide Space (FPS) and the
entropy concept and give some applications on polynucleotides
and selected genomes. We compare some of the results using our
entropy definitions with results obtained in Menconi (2005)
where the notion of computable complexity of several
complete genomes is analyzed and compared with the classical
entropy results. In Section 3, clarity of a polynucleotide is
considered and results on several polynucleotides are presented.
Finally in Section 4 the conclusions of the present work are
summarized.
2. Entropy and fuzzy polynucleotide space
2.1. Fuzzy sets and fuzzy hypercube
Let X be a set. A is a fuzzy subset of X if there is a function mA
such that (for details see Bardossy and Duckstein, 1995; Bezdek,
1981; Klir and Yuan, 1995; Hashimoto, 1983; Terrano et al., 1992;
Zimmermann, 1991)
(1) mA : X-½0,1.
(2) A ¼ fðx, mA ðxÞÞ : x A Xg, that is A is the set of all pairs ðx, mA ðxÞÞ
such that x A X and mA ðxÞ is the degree of its membership in A.
In what follows if X¼{x1,x2, y,xn} and
A ¼ fðx1 , mA ðx1 ÞÞ, . . . ,ðxn , mA ðxn ÞÞg,
then we write
A ¼ ðmA ðx1 Þ, . . . , mA ðxn ÞÞ:
Let A and B two fuzzy sets of a set X.
Then by A4B we denote the fuzzy set for which the member-
ship function mA4B : X-½0,1 is defined as following
mA4B ðxÞ ¼ minfmA ðxÞ, mB ðxÞg,
for every x A X.
Also by A3B we denote the fuzzy set for which the member-
ship function mA3B : X-½0,1 is defined as following
mA3B ðxÞ ¼ maxfmA ðxÞ, mB ðxÞg,
for every x A X.
For A a fuzzy set, the fuzzy complement Ac is defined by
Ac(x)¼ 1 A(x), x A X.
Kosko (1992) introduced a geometrical interpretation of fuzzy
sets as points in a hypercube. Indeed, for a given set X ¼{x1,x2,
y,xn}, the set of all fuzzy subsets (of X) is precisely the unit
hypercube
In ¼ ½0,1n ,
since any fuzzy subset A determines a point P A In given by
P ¼ ðmA ðx1 Þ, . . . , mA ðxn ÞÞ:
Reciprocally, any point P ¼ ða1 , . . . ,an Þ A In generates a
fuzzy subset A of X defined by the map mA : X-½0,1 such that
mA ðxi Þ ¼ ai , i ¼1,2,y,n.
Nonfuzzy or crisp subsets of X¼{x1,y,xn} are given by
mappings
m : X-f0,1g
from the set X into the set {0,1} and they are located at the 2n
corners of the n-dimensional unit hypercube In. So, the ground set
 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105 97

X¼{x1,y,xn} is itself the fuzzy set ð1,1, . . . ,1Þ A In . Also, the empty Table 1
fuzzy set is the fuzzy set ð0,0, . . . ,0Þ A In , denoted by |. (a) Number of nucleotides at the three base sites of a codon in the s1 sequence. and
(b) Fractions of nucleotides at the three base sites of a codon in the coding
Hypercubical calculus is developed in Zaus (1999), and some
sequence of s1
applications of the fuzzy unit hypercube are given in Nieto and
Torres (2003), Sadegh-Zadeh (1999) and Hegalson and Jobe U C A G Total
(1998). In this context a codon corresponds to a corner of the
12-dimensional unit hypercube I12. Any element of I12 may be (a)
First base 2 0 0 0 2
viewed as a fuzzy codon. Second base 0 0 1 1 2
DNA and RNA can be treated as a language written using an Third base 1 1 0 0 2
alphabet of strings. The role of strings is played by several
U C A G
chemical compounds. In fact the alphabet for DNA is {T,C,A,G}
while for RNA {U,C,A,G} where A,C,G,T and U stand for Adenine, (b)
Cytosine, Guanine, Thymine and Uracil, respectively. In this First base 1 0 0 0
context in the case of RNA alphabet if U is the first letter of this Second base 0 0 0.5 0.5
Third base 0.5 0.5 0 0
alphabet one codes it as ð1,0,0,0Þ : 1 because the first letter U is
present, 0 since the second letter does not appear, 0 since the example, Sadegh-Zadeh, 2000). Since the notion of entropy is
third letter is not present and 0 since the fourth letter G does not related also to the calculation of distances between points in
appear. In a similar way C is represented as (0,1,0,0), A as (0,0,1,0) the FPS, we describe briefly in the next section the notion of
and G as (0,0,0,1). So if we have a nucleotide described by the distance and then pass to the concepts of entropy and clarity.
codon UCG (serine) this would be written in the I12 hypercube as
ð1,0,0,0,0,1,0,0,0,0,0,1Þ:

2.2. Metrics—distances
There are cases where the exact chemical structure of the
sequence is not known for the complete sequence. In this case
Consider the n-dimensional unit hypercube In.
some components of its fuzzy code being neither 0 or 1 but a
If p ¼ ðp1 , . . . ,pn Þ, q ¼ ðq1 , . . . ,qn Þ A In are two different fuzzy
value in the interval (0,1) and are sequences not necessarily at a
polynucleotides, then we consider the following distances
corner of the hypercube. If for example we have a codon
between the elements p and q:
ð0:3,0:4,0:2,0:1,0,0,1,0,1,0,0,0Þ
This stands for XAU. The first letter X is unknown and corresponds
to: U to extent 0.3, C to extent 0.4, A to extent 0.2 and G to (1) Euclidean distance
extend 0.1 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
When we have a polynuclotide which is a sequence of k u n
uX
triplets, one would need a k  12 hyperspace. For example if we l ðp,qÞ ¼ t
2
ðpi qi Þ2 :
i¼1
have the polynucleotide described by the sequence UACUGU
(tyrosin/cysteine), it is a point in I212 ¼ I24 and represented by
(2) Hamming distance
s1 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0,1,0,0,0,0,0,0,1,1,0,0,0Þ
X
n
l1 ðp,qÞ ¼ jpi qi j:
However, if one considers the frequencies of the nucleotides of i¼1
the alphabet at the three base sites of a codon in the coding
sequence it may be viewed as a point in the hypercube I12. (3) Nieto–Torres Vazquez–Trasande (NTV) metric
Taking into account the number of nucleotides at the three Pn
base sites of a codon in the s1 sequence (see Table 1a) as well as jp q j
lðp,qÞ ¼ Pn i ¼ 1 i i :
the fractions of nucleotides at the three base sites of a codon in i¼1 maxfp i ,qi g
the coding sequence of s1 (See Table 1b) sequence s1 can be
written in the I12 space as
f ðs1 Þ ¼ ð1,0,0,0,0,0,0:5,0:5,0:5,0:5,0,0Þ: Also, if p ¼ q ¼ | ¼ ð0, . . . ,0Þ, then lð|,|Þ ¼ 0 (see Nieto et al.,
2003).
The distance l is motivated by publications Lin (1997) and
In the case of complete genome the frequencies of the Sadegh-Zadeh (2000). We know that l is a metric Nieto et al.
nucleotides at the three base sites of a codon in the codon (2003) and has already been employed in Torres and Nieto (2003),
sequence are considered. This can be viewed as point in the Torres and Nieto (2006) and Nieto et al. (2006). In Dress and Lokot
hypercube I12. (2003) (see, also Dress et al., 2004) it is proposed to call this
This idea has been applied to the genomes of M. tuberculosis, metric as the NTV metric. Applications of metrics are also
E. coli, and A. aeolicus to obtain their fuzzy set of frequencies presented in (Karakasidis and Georgiou, 2004; Samaras et al.,
and calculate their corresponding distance in the Fuzzy 2001).
Polynucleotide Space (FPS) (see Torres and Nieto, 2003; Nieto
et al., 2006).
When dealing with genetic sequences it is of interest: 2.3. The entropy of a polynucleotide

(i) to be able to describe how different two sequences are. For
this reason the notion of distance is used (see Nieto et al.,
2003, 2006; Nieto and Torres, 2003), and
(ii) to know how much ordered the sequence is. In this direction
the notion of entropy and clarity is employed (see, for
Let X¼{x1,y,xn} be a set and
A ¼ fðx1 , mA ðx1 Þ ¼ a1 Þ, . . . ,ðxn , mA ðxn Þ ¼ an Þg  ða1 , . . . ,an Þ,
where ai A ½0,1, a fuzzy set of X. Then by c(A) (see, for example,
Sadegh-Zadeh, 2000) we denote the number
98 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105

X
n X
n entropyl ðAÞ ¼ 1lðA,CÞ,
mA ðxi Þ ¼ ai :
i¼1 i¼1 where C is the fuzzy set (0.5,y,0.5) of In.

Proof. It is known that (see Sadegh-Zadeh, 2000) cðCÞ ¼ l1 ðC,|Þ.

In the crisp case, c(A) is the cardinality of A. Thus
In the following we propose a new definition of entropy
l1 ðC,AÞ
Definition 1. Let (In,d) be a metric space (see, for example, entropyl1 ðAÞ ¼ 1
l1 ðC,|Þ
Engelking, 1977), X ¼{x1,y,xn} a set, C ¼ ð0:5, . . . ,0:5Þ A In , l1 ðC,AÞ
| ¼ ð0, . . . ,0Þ A In , and F(2X) the fuzzy power set of X. The map ¼ 1
cðCÞ
entropyd : Fð2X Þ-½0,1, cðCÞl1 ðA,CÞ
¼ :
cðCÞ
dðA,CÞ
entropyd ðAÞ ¼ 1 ,
dðC,|Þ
Obviously,
for every A ¼ ða1 , . . . ,an Þ A In , is called fuzzy entropy map with
n
respect to the metric d. cðCÞ ¼ 0:5 þ 0:5 þ    þ0:5 ¼ n  0:5 ¼ ,
2
Remark. De Luca and Termini (1972) first axiomatized nonprob- and we have
abilistic entropy in the setting of fuzzy sets theory (see also Fan
and Ma, 2002). We adopt them here. Let X be a set and let E be a cðCÞl1 ðA,CÞ
entropyl1 ðAÞ ¼
set-to-point map cðCÞ
n 1
E : Fð2X Þ-½0,1, l ðA,CÞ
¼2 n
where F(2X) is the fuzzy power set of X. Hence E is a fuzzy set
2
defined on fuzzy sets of X. E is an entropy measure if it satisfies n2l1 ðA,CÞ
the four De Luca–Termini axioms: ¼ :
n

(DT1) E(A) ¼0 if A A 2X (A nonfuzzy), where 2X is the power

set of X. Also
(DT2) E(A) ¼1 if A(x)¼0.5, for every x A X. 1 pffiffiffi
l2 ðC,|Þ ¼  n,
(DT3) EðAÞ rEðBÞ if AðxÞ rBðxÞ when BðxÞ r0:5 and BðxÞ rAðxÞ 2
when BðxÞ Z 0:5.
and
(DT4) E(A) ¼e(Ac).
l2 ðC,AÞ
entropyl2 ðAÞ ¼ 1
l2 ðC,|Þ
It is possible that for some metric d the fuzzy entropy map
l2 ðC,AÞ
entropyd with respect to the metric d to satisfy the De Luca and ¼ 1
1 pffiffiffi
Termini axioms. So, for example for the metrics l1 and l2 it is clear  n
that the maps pffiffiffi 2
n2lðA,CÞ
¼ pffiffiffi :
l1 ðA,CÞ n
E1 ðAÞ  entropyl1 ðAÞ ¼ 1
l1 ðC,|Þ
and Now, for the NTV metric we have
2
l ðA,CÞ n  0:5
E2 ðAÞ  entropyl2 ðAÞ ¼ 1 lðC,|Þ ¼ ¼ 1:
l2 ðC,|Þ n  0:5
satisfy the above four De Luca–Termini axioms.
Also, it is possible that for some metric d the map entropyd does
not satisfy the De Luca and Termini axioms. So, for example for Thus
the NTV metric l the map
lðA,CÞ
entropyl ðAÞ ¼ 1 ¼ 1lðA,CÞ:
lðA,CÞ lðC,|Þ
E0 ðAÞ  entropyl ðAÞ ¼ 1
lðC,|Þ
does not satisfy the above four De Luca–Termini axioms (see This complete the proof. &
Example 2 below).
In the following we present some theorems concerning the
specification of the formulas used to calculate entropies in the FPS Example 1. Let X¼{x1,x2} be a set and A ¼(0.4,0.8) a fuzzy set of X.
space for specific metrics. We consider the metric space (I2, l1).

Theorem 1. Let X ¼{x1,y,xn} and A ¼(a1,y,an) a fuzzy set of X. Using the definition of entropy given in Sadegh-Zadeh (2000)
Then, the following statements are true: we have (see page 23 of Sadegh-Zadeh, 2000):
3
cðCÞl1 ðA,CÞ n2l1 ðA,CÞ entðAÞ ¼ ¼ 0:4286:
entropyl1 ðAÞ ¼ ¼ , 7
cðCÞ n
pffiffiffi
n2l2 ðA,CÞ Using the above definition we have
entropyl2 ðAÞ ¼ pffiffiffi ,
n 22l1 ðA,CÞ
entropyl1 ðAÞ ¼
and 2
 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105 99

22ðj0:40:5j þj0:80:5jÞ Now cðCÞ ¼ n=2. Thus,

¼
2
cðA4Ac Þ 2  cðA4Ac Þ
20:8 1:2 entropyl1 ðAÞ ¼ ¼ :
¼ ¼ ¼ 0:6: cðCÞ n
2 2
We thus observe that the entropy of Sadegh-Zadeh does not
coincide with the Hamming entropy Finally, by the fact that
entðAÞ ¼ 0:4286a entropyl1 ðAÞ ¼ 0:6: cðA4Ac Þ ¼ l1 ðA4Ac ,|Þ:
it follows that
Remark. According to the Definition 1 and Theorem 1 we have a 2  l1 ðA4Ac ,|Þ
entropyl1 ðAÞ ¼ :
geometrical intepretation of entropy is illustrated in Fig. 1. The n
entropyl1 of a fuzzy set A is 1 minus the Euclidean distance
a ¼ l2 ðA,CÞ divided by the Euclidean distance b ¼ l2 ðC,|Þ, that is When some of the ai are less than or equal 0.5 and others
a greater than 0.5, the proof is analogous.
entropyl2 ðAÞ ¼ 1 :
b In the following we present some examples of applications of
the use of the entropy definitions in various cases of polynucleo-
tides from relatively small ones up to large ones.
Theorem 2. Let X ¼{x1,y,xn} be a set and A¼ (a1,y,an), where
ai A ½0,1, a fuzzy set of X. Let Ac ¼ ð1a1 , . . . ,1an Þ A In . Then Example 2. Let (I2,d) be a metric space, X¼{x1,x2} a set, and
A¼C ¼(0.5,0.5) a fuzzy set of X. Then, we have
cðA4Ac Þ 2  cðA4Ac Þ 2  l1 ðA4Ac ,|Þ
entropyl1 ðAÞ ¼ ¼ ¼ , dðC,CÞ
cðCÞ n n entropyd ðAÞ ¼ 1 ¼ 10 ¼ 1:
dðC,|Þ
where C is the fuzzy set C ¼ ð0:5, . . . ,0:5Þ A In .

Proof. Suppose that ai r 0:5 for every i¼ 1,2,y,n  1, and an 4 0:5.

Then, we have Example 3. Let (I2,d) be a metric space, X ¼{x1,x2} a set and
A¼(0,1), B ¼(1,0) and D ¼(1,1) three fuzzy sets of X. Then, we have
A4Ac ¼ ðminfa1 ,1a1 g, . . . ,minfan1 1,1an1 g,minfan ,1an gÞ
¼ ða1 , . . . ,an1 ,1an Þ: & dðC,AÞ
entropyd ðAÞ ¼ 1 ¼ 11 ¼ 0,
dðC,|Þ
where d¼ l1 or d ¼l2.

By the above we have Similarly

1
l ðA,CÞ entropyd ðBÞ ¼ entropyd ðDÞ ¼ 0,
entropyl1 ðAÞ ¼ 1
l1 ðC,|Þ where d¼ l1 or d ¼l2.
Pn
j0:5ai j For the NTV metric l we have
¼ 1 i ¼ 1
n  0:5 lðC,AÞ 2 1
Pn1
j0:5ai j þ j0:5an j entropyl ðAÞ ¼ 1 ¼ 1 ¼ ,
¼ 1 i ¼ 1 lðC,|Þ 3 3
n  0:5
Pn1
i ¼ 1 0:5ai þ an 0:5 1
¼ 1 entropyl ðBÞ ¼ ,
n  0:5 3
n  0:5ða1 þa2 þ    þ an1 Þ þ an 1
¼ 1 and
n  0:5
n  0:5n  0:5 þ ða1 þ a2 þ    þ an1 Þan þ 1 lðC,DÞ 1 1
¼ entropyl ðDÞ ¼ 1 ¼ 1 ¼ :
n  0:5 lðC,|Þ 2 2
a1 þa2 þ    þ an1 þ1an
¼
n  0:5 We see that for points at the corners of I2, NTV metric does not
cðA4Ac Þ result zero values as is the case for metrics l1 or l2.
¼ :
cðCÞ
Example 4. Consider the sequences employed also by Sadegh-
Zadeh (2000):
(0,0,1) (0,1,1)
s1 ¼ UACUGU tyrosine=cysteine

This point belongs to the 24-dimensional unit cube and it

(1,0,1) corresponds to a corner in I24. Following the methodology of
(1,1,1)
Torres and Nieto (2003) we calculate the frequencies (fractions) of
a
the nucleotide at the three base sites in order to obtain their fuzzy
A
representation in the I12 hyperspace. The corresponding results
C appear in Tables 1a and 1b. Note that the entropy in I24 is
b
entropyl1 ðs1 Þ ¼ entropyl2 ðs1 Þ ¼ 0
O (0,1,0)
and
entropyl ðs1 Þ ¼ 0:2:

(1,1,0)
(1,0,0) In the I12 space the frequencies give a point in the I12 space :
Fig. 1 f ðs1 Þ ¼ ð1,0,0,0,0,0,0:5,0:5,0:5,0:5,0,0Þ:
100 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105

Note that now we identify s1 in I24 and f(s1) in I12. Table 3

If C ¼ ð0:5, . . . ,0:5Þ A I12 , then, we have the following entropies Calculated entropy values for sequences s7, s8 and s9 using the metrics l, l1 and l2.
for the Euclidean metric
pffiffiffi Metric s7 s8 s9
l2 ðs1 ,CÞ 2
entropyl2 ðs1 Þ ¼ 1 2 ¼ 1 pffiffiffi  0:183503, l2 0 0.087129 0.183503
l ðC,|Þ 3
l1 0 0.166667 0.333333
Hamming metric, l 0.2 0.285714 0.384615

l1 ðs1 ,CÞ 1
entropyl1 ðs1 Þ ¼ 1 ¼  0:333333,
l1 ðC,|Þ 3 Example 6. Now consider the following sequences:
and NTV metric
s7 ¼UACUAC
lðs1 ,CÞ 5
entropyl ðs1 Þ ¼ 1 ¼  0:384615: s8 ¼UAGUAU
lðC,|Þ 13
s9 ¼UACUCG

We can see that there is a difference in the results when which correspond in the FPS, respectively, to
dealing with FPS. This subtlety will be analyzed with further
s7 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0Þ
results.

Example 5. Consider the sequences: s8 ¼ ð1,0,0,0,0,1,0,0,0:5,0,0,0:5Þ,

s2 ¼CACUGU histidine/cysteine s9 ¼ ð1,0,0,0,0,0:5,0:5,0,0,0,0:5,0:5Þ:

s3 ¼CUCUGU leucine/cysteine The corresponding entropy values appear in Table 3. Note in the
s4 ¼CAUUGU histidine/cysteine first one, s7, the same triplet UAC is repeated all along the
s5 ¼CAGUGU glutamine/cysteine sequence. In the second one, s8, the dinucleotide UA is repeated at
s6 ¼CAAUGU glutamine/cysteine the same base positions.

The program to compute the entropy using these three metrics

These are points in a 24-dimensional unit cube since they are is available on request from the authors.
made of 2 triplets. Following the methodology of Torres and Nieto As we know from physics, entropy is a measure of the order/
(2003) we calculated the frequencies (fractions) of the nucleotides disorder of the system, which represents the degree of complexity
at the three base sites in order to obtain their fuzzy representation in order to describe the system. In the case where we have
in the I12 hyperspace. The entropy in the I24 is again zero as there repetition of the same triplet (as it is the case for UAC in
is no uncertainty concerning the chemical composition. However, sequence s7) we have maximum order resulting in zero entropy.
when dealing with FPS results will be different. In the case of FPS In the case where we have repetition of the same dyad like UA in
zero entropy means maximum order we have the same triplet all sequence s8 we have a slightly higher entropy and in other cases
along the genetic sequence. like sequence s9 entropy increases further.
The above sequences are represented in the I12 space as (see We remind here that there is a probabilistic definition of
Nieto et al., 2006): entropy for signals related to thermodynamics see Shannon,
s2 ¼ ð0:5,0:5,0,0,0,0,0:5,0:5,0:5,0:5,0,0Þ, 1948). This classical measure of entropy is defined as
X
H¼ pj log2 ðpj Þ
s3 ¼ ð0:5,0:5,0,0,0:5,0,0,0:5,0:5,0:5,0,0Þ,
where pj are the non-zero probabilities of a signal to have a given
s4 ¼ ð0:5,0:5,0,0,0,0,0:5,0:5,1,0,0,0Þ,
value.
s5 ¼ ð0:5,0:5,0,0,0,0,0:5,0:5,0:5,0,0,0:5Þ Applying this definition in the case of selected polynucleotides
as represented in the FPS the role of pj is played by the non-zero
and coordinates of the polynucleotide. In this case for the entropy of
s6 ¼ ð0:5,0:5,0,0,0,0,0:5,0:5,0:5,0,0:5,0Þ: s7, s8 and s9 we have
Hðs7 Þ ¼ 0,
The results of entropy using the various metrics are summar-
ized in Table 2. Hðs8 Þ ¼ 1
s2, s3, s4, s5 and s6 present the same entropy results although and
the exact value changes depending on the metric used and only s4
presents different entropy which is lower. In fact s2, s3, s5 and s6, Hðs9 Þ ¼ 2:
have the same number of coordinates being equal to 0.5 and all
the others 0, while s4 has only four coordinates equal to 0.5, one What is of interest is that sequence s7 which corresponds to
equal to 1 and all the others equal to 0. the most ordered sequence results in zero entropy. In the other
two sequences we have increasing entropy as in the case of
Table 2 definitions based on metrics given above. It is also remarkable
Entropy values for sequences s2, s3, s4, s5, s6 using the metrics l, l1 and l2 calculated that the ratio of entropies of sequences s8 and s9 equals 2 in both
in FPS.
cases: probabilistic entropy and entropy based on metrics.
Metric s2 s3 s4 s5 s6
Example 7. Consider the following sequences with three triplets
2
l 0.292893 0.292893 0.183503 0.292893 0.292893 r1 ¼ UACUACUAC
l1 0.5 0.5 0.333333 0.5 0.5
l 0.5 0.5 0.384615 0.5 0.5
r2 ¼ UACUACUAG
 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105 101

r3 ¼ UACCAAUAG In the probabilistic definition of entropy, for the point E we have

312 36
represented in the I ¼I space as HðEÞ ¼ 6
r1 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0,1,0,0,0,0,1,0,0,0,0,1,0,1,0,0,0,0,1,0,0,0,0,1,0Þ
which is the maximum possible value in I12.
(2) If A ¼ ða1 , . . . ,a12 Þ A I12 , C ¼ ð0:5, . . . ,0:5Þ A I12 , and | ¼
r2 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0,1,0,0,0,0,1,0,0,0,0,1,0,1,0,0,0,0,1,0,0,0,0,0,1Þ
12
ð0, . . . ,0Þ A I , then
r3 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0,0,0,1,0,0,1,0,0,0,1,0,0,1,0,0,0,0,1,0,0,0,0,0,1Þ:
l2 ðC,AÞ
entropyl2 ðAÞ ¼ 1
l2 ðC,|Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The entropy in the I36 space is ða1 0:5Þ2 þ    þða12 0:5Þ2
¼ 1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
entropyd ðr1 Þ ¼ entropyd ðr2 Þ ¼ entropyd ðr3 Þ ¼ 0, ð0:50Þ2 þ    þð0:50Þ2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where d ¼l1 or d ¼l2, and
ða1 0:5Þ2 þ    þða12 0:5Þ2
entropyl ðr1 Þ ¼ entropyl ðr2 Þ ¼ entropyl ðr3 Þ ¼ 0:2: ¼ 1 pffiffiffi :
3

However, in the FPS representation the entropy of the

sequences would not result zero values. In fact zero values would In FPS we have that
be reproduced only for the sequences where the same triplet is a1 þ    þ a12 ¼ 3:
repeated all along the sequence, for the second where we have
repetition of the dyad UA at the same base positions entropy This comes out from the fact that each ai corresponds to the
increases and is higher in the third case. Following the frequency of appearance of a letter of the DNA (or RNA) alphabet
methodology of Torres and Nieto (2003) we calculated the at each triplet base (first, second, third). For each base these
frequencies (fractions) of the nucleotides at the three base sites probabilities which correspond to a quadruplet of ai sums to 1, so
in order to obtain their fuzzy representation in the I12 hyperspace: for the three bases this results in 3.
Using a simple program of Mathematica (see Appendix
r1 ¼ ð1,0,0,0,0,1,0,0,0,0,1,0Þ
Appendix A) we see that the map
r2 ¼ ð1,0,0,0,0,1,0,0,0,0,2=3,1=3Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ða1 0:5Þ2 þ    þ ða12 0:5Þ2
entropyl2 ðAÞ ¼ 1 pffiffiffi
r3 ¼ ð2=3,1=3,0,0,0,1,0,0,0,1=3,1=3,1=3Þ 3
with the restriction
The corresponding entropy values are summarized in Table 4.
a1 þ    þ a12 ¼ 3:
Remarks. (1) In the case of large sequences the maximum
we have a maximum of the entropy at the point E.
entropy would correspond to the characteristic case in which
As above, we can see that
we have equiprobable distribution of all alphabet letters at all
three bases. This would correspond to the point. l1 ðC,AÞ ja1 0:5jþ    þ ja12 0:5j
entropyl1 ðAÞ ¼ 1 ¼ 1
E ¼ ð0:25,0:25,0:25,0:25,0:25,0:25,0:25,0:25,0:25,0:25,0:25,0:25Þ l1 ðC,|Þ 6

in the FPS representation. Its entropy in the metric l2 is with the restriction
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l2 ðE,CÞ 12  0:252 a1 þ    þ a12 ¼ 3:
entropyl2 ðEÞ ¼ 1 2 ¼ 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 0:5:
l ðC,|Þ 12  0:52 have a maximum (of entropy) at the point E.
The entropies obtained by the use of metrics l1 and l are
Example 8. Following the methodology of Torres and Nieto
l1 ðE,CÞ
entropyl1 ðEÞ ¼ 1 ¼ 0:5 (2003), we consider the point
l1 ðC,|Þ
ð0:1632,0:3089,0:1724,0:3556,0:2036,0:3145,0:1763,0:3056,
and
0:1645,0:3461,0:1593,0:3302Þ A I12 :
lðE,CÞ
entropyl ðEÞ ¼ 1 ¼ 0:5: which corresponds to the fuzzy set of frequencies of the genome
lðC,|Þ
of M. tuberculosis (see Torres and Nieto, 2003), the point
ð0:1605,0:2420,0:2600,0:3374,0:3116,0:2286,0:2846,0:1752,
A point in the FPS corresponds to a corner of the hypercube
0:2619,0:2568,0:1831,0:2981Þ A I12 :
when we have the same triplet all along the sequence. If we have
maximum order, the point occupies a corner of the hypercube. which corresponds to the fuzzy set of frequencies of the genome
The bigger the distance from the corners, the bigger the entropy, of E. coli, and the point
and thus the bigger the complexity to describe the sequence. ð0:1706,0:1605,0:3241,0:3446,0:3282,0:1735,0:3478,0:1504,
0:2139,0:2455,0:3052,0:2352Þ A I12
Table 4
Calculated entropy values for sequences r1, r2 and r3 using the metrics l, l1 and l2.
which corresponds to the fuzzy set of frequencies of the genome
of A. aeolicus (see Nieto et al., 2006).
Metric r1 r2 r3
We also compare the entropies of the Mycoplasma Pneumoniae
2
l 0 0.077042 0.206508 using our definitions of entropy. In Tables 5a and 5b we
l1 0 0.111111 0.277778
l 0.2 0.255814 0.35
present the results concerning the representation of Mycoplasma
Pneumoniae in FPS.
102 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105

Thus, the genome of M. pneumoniae is represented in the I12 Table 6

hypercube by the point Entropy values for sequences M. tuberculosis, E. coli, A. aeolicus and M. pneumoniae
using the metrics l, l1 and l2 calculated in FPS.
ð0:2038,0:1769,0:327,0:2923,0:3054,0:1936,0:3601,0:1408,
0:3212,0:2285,0:26,0:1902Þ A I12 : Metric M. tuberculosis E. coli A. aeolicus M. pneumoniae

l2 0.475876 0.488814 0.478769 0.482055

The corresponding entropies of all long polynucleotides appear l1 0.500033 0.499967 0.499917 0.499967
l 0.500033 0.499967 0.499917 0.499967
in Table 6. We observe that metrics give the same numerical
results while the NTV metric gives different results and can
differentiate in a more clear way the complexity of polynucleo-
tides in their FPS representation. Table 7
Comparison of results obtained using entropyl , entropyl1 , entropyl2 with the results
Remark. In Table 7 we compare the results obtained using the of complexity K and probabilistic entropy H1 of Menconi (2005) in the case of A.
three above entropy definitions for the A. aeolicus and E. coli with aeolicus and E. coli.
the results of Menconi (2005) where they computed the complex-
Genome K H1 entropyl2 entropyl1 entropyl
ity K of a genome and the probabilistic entropy H1 (for more
details on the notions of K and H1 see Menconi, 2005). What is of A. aeolicus 1.883 1.976 0.478 0.499 0.499
interest is that in the case of entropyl1 and entropyl results are E. Coli 1.893 1.987 0.489 0.499 0.499
practically identical, while entropyl2 results in a small but
identifiable difference between the two genomes in the same
sense like K and H1.
l1 ðA,CÞ
(3) clarityl1 ðAÞ ¼ :
3. Clarity and fuzzy polynucleotide space cðCÞ

2l1 ðA,CÞ
(4) clarityl1 ðAÞ ¼ :
n
Definition 2. Let d be a metric in I , X ¼{x1,y, xn} a set and n
A¼(a1,y,an), where ai A ½0,1, a fuzzy set of X. The clarity of A, with cðCÞcðA4Ac Þ
(5) clarityl1 ðAÞ ¼ :
respect to the metric d, denoted by clarityd(A) is defined to be the cðCÞ
number
2lðA,CÞ
1entropyd ðAÞ, (6) clarityl ðAÞ ¼ pffiffiffi :
n
that is (7) clarityl ðAÞ ¼ lðA,CÞ:

clarityd ðAÞ ¼ 1entropyd ðAÞ:

Proof. Follows by Theorems 1 and 2 and by fact that clarityd(A)¼
Example 9. Let X¼{x1,x2} be a set, A¼(0.4,0.8) a fuzzy set of X and
1 entropyd(A). &
consider the metric space (I2, l1). Then, we have

clarityl1 ðAÞ ¼ 1entropyl1 ðAÞ ¼ 10:6 ¼ 0:4: Remark. According the Definition 2 and Theorem 3 we have a
geometrical interpretation of clarity as illustrated in Fig. 1. The
Also, using the definition of clarity given in Sadegh-Zadeh
clarity of a fuzzy set A is the Euclidean distance a ¼l2(A,C) divided
(2000) we have
by the distance b ¼ l2 ðC,|Þ, that is
3 4 a
clarðAÞ ¼ 1entðAÞ ¼ 1 ¼ ¼ 0:5714: clarityl1 ðA,BÞ ¼ :
7 7 b
We observe that
clarityl1 ðAÞ ¼ 0:4 aclarðAÞ ¼ 0:5714: Example 10.

Theorem 3. Let X ¼{x1,y,xn} and A ¼(a1,y,an) a fuzzy set of X.

Then, the following statements are true: (1) Let X¼{x1,x2} be a set and A¼(0,1), B ¼(1,0) two fuzzy sets of
X. Then
(1) entropyd ðAÞ,clarityd ðAÞ A ½0,1.
entropyd ðAÞ ¼ entropyd ðBÞ ¼ 0
(2) entropyd(A)¼ 1  clarityd(A). and

Table 5 clarityd ðAÞ ¼ clarityd ðBÞ ¼ 1,

(a) The number of nucleotides at the three base sites of a codon in the codon where d ¼l1 or d ¼l2. Also,
sequence of Mycoplasma Pneumoniae and (b) Fractions of nucleotides at the three
base sites of a codon in the coding sequences of Mycoplasma Pneumoniae. entropyl ðAÞ ¼ entropyl ðBÞ ¼ 13
T C A G
and

(a) clarityl ðAÞ ¼ clarityl ðBÞ ¼ 113 ¼ 23:

First base 48995 42525 78622 70293
Second base 73438 46554 86585 33858 (2) Let X¼{x1,y,xn} be a set and A a fuzzy set of X such that A¼C.
Third base 77233 54942 62523 45737
Then entropyd(A)¼ 1 and clarityd(A)¼0.
(b) (3) We consider the following polynucleotide sequences:
First base 0.2038 0.1769 0.327 0.2923 s1 ¼UACUGU (tyrosine/cysteine)
Second base 0.3054 0.1936 0.3601 0.1408
Third base 0.3212 0.2285 0.26 0.1902
s2 ¼CACUGU (histidine/cysteine)
s3 ¼CUCUGU (leucine/cysteine)
 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105 103

Table 8 sequence of the polynucleotide. However, results in both cases

Calculated clarity values for sequences s1, s2 and s3 using the metrics l, l1 and l2. show, as expected, that entropy is related to the complexity of
description of the sequence. The value of entropy/clarity is
Metric s1 s2 s3
representative of the complexity of description of the polynucleo-
l2 1 1 1 tide. Similar entropy means similar degree of complexity. We also
l1 1 1 1 apply the definition of probabilistic entropy in the case of selected
l 0.8 0.8 0.8 polynucleotides and we observe that the entropy based on
metrics presents similar behavior with that of probabilistic
nature.
Table 9 Further studies are in progress in order to investigate in more
Calculated clarity values for sequences s1, s2 and s3 using the metrics l, l1 and l2. detail the properties of these notions and their biological implica-
tions since it seems that the use of FPS space can lead to a
Metric s1 s2 s3
reduction of the necessary information and the use of appropriate
2
l 0.816497 0.707107 0.707107 metrics can be used to differentiate the degree of their complexity.
l1 0.66667 0.5 0.5
l 0.67385 0.5 0.5

Acknowledgements

The research of J.J. Nieto has been partially supported by

Table 10 Ministerio de Educacion y Ciencia and FEDER, project MTM2007-
Computed clarity for sequences M. tuberculosis, E. coli, A. aeolicus and 61724, and by Xunta de Galicia and FEDER, project PGIDIT05P-
M. pneumoniae.
XIC20702PN.
Metric M. tuberculosis E. coli A. aeolicus M. pneumoniae

l2 0.524124 0.511186 0.521231 0.57945 Appendix A

l1 0.499967 0.500033 0.500083 0.500033
l 0.499967 0.500033 0.500083 0.50
"( P12 ) #
2 X
12
i ¼ 1 ða½i0:5Þ
ln½1 :¼ NMaximize 1 pffiffiffi , a½i ¼ ¼ 3 ,Array½a,f12g
3 i¼1

We have the following representations in the I24 space: Out½1 ¼ f0:5,fa½1-0:25,a½2-0:25,a½3-0:25,a½4-0:25,a½5

-0:25,a½6-0:25,a½7-0:25,a½8-0:25,a½9
s1 ¼(1, 0, 0, 0, 0, 1, 0, 0, 0, 0,1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0) -0:25,a½10-0:25,a½11-0:25,a½12-0:25gg
s2 ¼(0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0)
s3 ¼(0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0)

Using the distances l2, l1 and l for the clarity of s1 ,s2 ,s3 A I24 we
obtain the results for entropy values that are summarized in Table 8.
The results are consistent with the fact that in the case of the
I12  k space (in this case of I24) all above sequences are known
with precision. However, if we use their representation in the I12
dimensional FPS results the results present some differences.
Using the distances l2, l1 and l for the clarity of s1 ,s2 ,s3 A I12 we
have the results that appear in Table 9:
Again, results indicate the degree of complexity necessary for
the description of the polynucleotides.
(4) We consider the following fuzzy set of frequencies of the
genomes of M. tuberculosis, E. coli, A. aeolicus. and M.
pneumoniae. Using the distances l2, l1 and l for the clarity
the corresponding results are summarized in Table 10.
4. Conclusions
We present results concerning the notions of fuzzy entropy
and clarity of a polynucleotide. We propose a new definition of
entropy and we consider several applications using different
metrics in the case of I12  k space, where k is the number of
codons of the polynucleotide. We also examine the behavior of
these notions when those polynucleotides are projected in the I12
Fuzzy Polynucleotide Space. We observe that in both cases we
have a different interpretation of the obtained results for entropy.
While in the former case low entropy means that we are close to a
corner of the 12  k space in the latter case it means that we have
repetition of the same triplet or part of a triplet all along the
References
Aguero-Chapin, G., Gonzalez-Diaz, H., Molina, R., Varona-Santos, J., Uriarte, E.,
Gonzalez-Diaz, Y., 2006. Novel 2D maps and coupling numbers for protein
sequences. The first, QSAR study of polygalacturonases; isolation and
prediction of a novel sequence from, Psidium guajava. FEBS, Letter 580 (3),
723–730.
Aguero-Chapin, G., Gonzalez-Diaz, H., Riva, G.D., Rodriguez, E., Sanchez-Rodriguez, A.,
Podda, G., Vazquez-Padron, R.I., 2008. MMM-QSAR, recognition of ribonucleases
without alignment: comparison with an HMM model and isolation
from Schizosaccharomyces pombe, prediction, and experimental assay of
a new sequence. Journal of Chemical Information and Modelling 48 (2),
434–448.
Bardossy, A., Duckstein, L., 1995. Fuzzy Rule-Based Modeling with Applications to
Geophysical, Biological and Engineering Systems. CRC Press, Boca Raton.
Bezdek, J.C., 1981. Pattern Recognition with Fuzzy Objective Function Algorithms.
Plenum Press, New York.
Chen, C., Zhou, X., Tian, Y., Zou, X., Cai, P., 2006a. Predicting protein structural class
with pseudo-amino acid composition and support vector machine fusion
network. Annals Biochemistry 357, 116–121.
Chen, C., Tian, Y.X., Zou, X.Y., Cai, P.X., Mo, J.Y., 2006b. Using pseudo-amino acid
composition and support vector machine to predict protein structural class.
Journal of Theoretical Biology 243, 444–448.
Chen, Y.L., Li, Q.Z., 2007a. Prediction of apoptosis protein subcellular location using
improved hybrid approach and pseudo amino acid composition. Journal of
Theoretical Biology 248, 377–381.
Chen, Y.L., Li, Q.Z., 2007b. Prediction of the subcellular location of apoptosis
proteins. Journal of Theoretical Biology 245, 775–783.
Chou, K.C., 1995. A novel approach to predicting protein structural classes in a
(20-1)-D amino acid composition space. Proteins: Structure, Function &
Genetics 21, 319–344.
Chou, K.C., 2000b. Review: prediction of protein structural classes and subcellular
locations. Current Protein and Peptide Science 1, 171–208.
Chou, K.C., 2001. Prediction of protein cellular attributes using pseudo amino acid
composition. Proteins: Structure, Function, and Genetics (Erratum: ibid., 2001,
Vol.44, 60) 43, 246–255.
Chou, K.C., 2000a. Prediction of protein subcellular locations by incorporating
quasi-sequence-order effect. Biochemical & Biophysical Research Communica-
tions 278, 477–483.
104 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105
Chou, K.C., 2005a. Prediction of G-protein-coupled receptor classes. Journal of
Proteome Research 4, 1413–1418.
Chou, K.-C., 2009. Pseudo amino acid composition and its applications in
bioinformatics, proteomics and system biology. Current Proteomics 6, 262–274.
Chou, K.C., Cai, Y.D., 2003. Predicting protein quaternary structure by pseudo
amino acid composition. Proteins: Structure, Function, and Genetics 53,
282–289.
Chou, K.C., Cai, Y.D., 2004. Predicting enzyme family class in a hybridization space.
Protein Science 13, 2857–2863.
Chou, K.C., Cai, Y.D., 2005. Prediction of membrane protein types by incorporating
amphipathic effects. Journal of Chemical Information and Modeling 45,
407–413.
Chou, K.C., Cai, Y.D., 2006. Predicting protein–protein interactions from sequences
in a hybridization space. Journal of Proteome Research 5, 316–322.
Chou, K.C., Cai, Y.D., Zhong, W.Z., 2006. Predicting networking couples for
metabolic pathways of Arabidopsis. EXCLI Journal 5, 55–65.
Chou, K.C., Elrod, D.W., 2002. Bioinformatical analysis of G-protein-coupled
receptors. Journal of Proteome Research 1, 429–433.
Chou, K.C., Elrod, D.W., 1999a. Protein subcellular location prediction. Protein
Engineering 12, 107–118.
Chou, K.C., Elrod, D.W., 1999b. Prediction of membrane protein types and
subcellular locations. Proteins: Structure, Function, and Genetics 34,
137–153.
Chou, K.C., Elrod, D.W., 2003. Prediction of enzyme family classes. Journal of
Proteome Research 2, 183–190.
Chou, K.C., Shen, H.B., 2007a. Euk-mPLoc: a fusion classifier for large-scale
eukaryotic protein subcellular location prediction by incorporating multiple
sites. Journal of Proteome Research 6, 1728–1734.
Chou, K.C., Shen, H.B., 2007b. Review: recent progresses in protein subcellular
location prediction. Analytical Biochemistry 370, 1–16.
Chou, K.C., Shen, H.B., 2007c. Large-scale plant protein subcellular location
prediction. Journal of Cellular Biochemistry 100, 665–678.
Chou, K.C., Shen, H.B., 2007d. MemType-2L: a web server for predicting membrane
proteins and their types by incorporating evolution information through Pse-
PSSM. Biochemistry and Biophysics Research Communication 360, 339–345.
Chou, K.C., Shen, H.B., 2007e. Signal-CF: a subsite-coupled and window-fusing
approach for predicting signal peptides. Biochemistry and Biophysics Research
Communication 357, 633–640.
Chou, K.C., Shen, H.B., 2008. Cell-PLoc: a package of web-servers for predicting
subcellular localization of proteins in various organisms. Nature Protocols 3,
153–162.
Chou, K.C., 2005b. Using amphiphilic pseudo amino acid composition to predict
enzyme subfamily classes. Bioinformatics 21, 10–19.
Chechetkin, V.R., 2003. Block structure and stability of the genetic code. Journal of
Theoretical Biology 222, 177–188.
DasGupta, B., Jiang, T., Kannan, S., Sweedyk, E., 1998. On the complexity and
approximation of syntenic distance. Discrete Applied Mathematics 88, 59–82.
Ding, Y.S., Zhang, T.L., Chou, K.C., 2007. Prediction of protein structure classes with
pseudo amino acid composition and fuzzy support vector machine network.
Protein & Peptide Letters 14, 811–815.
Du, P., Li, Y., 2006. Prediction of protein submitochondria locations by hybridizing
pseudo-amino acid composition with various physicochemical features of
segmented sequence. BMC Bioinformatics 7, 518.
Dress, A., Lokot, T., 2003. A simple proof of the triangle inequality for the NTV
metric. Applied Mathematics Letters 16, 809–813.
Dress, A., Lokot, T., Pustyl’nikov, L.D., 2004. A new scale-invariant geometry of L1
space. Applied Mathematics Letters 17, 815–820.
De Luca, A., Termini, S., 1972. A definition of a nonprobabilistic entropy in the
setting of fuzzy sets theory. Information and Control 20, 301–312.
Engelking, R., 1977. General Topology, Warszawa.
Fan, J.-L., Ma, Y.-L., 2002. Some new fuzzy entropy formulas. Fuzzy Sets and
Systems 128, 277–284.
Foster, M., Heath, A., Afzal, M., 1999. Application of distance geometry to 3D
visualization of sequence relation-ships. Bionformatics 15, 89–90.
Feng, Z.P., 2002. An overview on predicting the subcellular location of a protein. In
Silico Biology 2, 291–303.
Freeland, S.J., Hurst, L.D., 1998. The genetic code is one in a million. Journal of
Molecular Evolution 47, 238–248.
Gusev, V.D., Nemytikova, L.A., Chuzhanova, N.A., 1999. On the complexity
measures of genetic sequences. Bioinformatics 15, 994–999.
Georgiou, D.N., Karakasidis, T.E., Nieto, J.J., Torres, A., 2009. Use of fuzzy clustering
technique and matrices to classify amino acids and its impact to Chou’s pseudo
amino acid composition. Journal of Theoretical Biology 257, 17–26.
Gonzalez-Diaz, H., Aguero-Chapin, G., Varona, J., Molina, R., Delogu, G., Santana, L.,
Uriarte, E., Podda, G., 2007b. 2D-RNA-coupling numbers: a new computational
chemistry approach to link secondary structure topology with biological
function. Journal of Computational Chemistry 28 (6), 1049–1056.
Gonzalez-Diaz, H., Gonzalez-Diaz, Y., Santana, L., Ubeira, F.M., Uriarte, E., 2008.
Proteomics, networks and connectivity indices. Proteomics 8 (4), 750–778.
Gonzalez-Diaz, H., Perez-Castillo, Y., Podda, G., 2007a. Uriarte E. Computational
chemistry comparison of stable/nonstable protein mutants classification
models based on 3D and topological indices. Journal of Computational
Chemistry 28 (12), 1990–1995.
Gonzalez-Diaz, H., Vilar, S., Santana, L., Uriarte, E., 2007c. Medicinal chemistry and
bioinformatics-current trends in drugs discovery with networks topological
indices. Current Topics in Medicinal Chemistry 7 (10), 1015–1029.
Guo, Y.Z., Li, M., Lu, M., Wen, Z., Wang, K., Li, G., Wu, J., 2006. Classifying G protein-
coupled receptors and nuclear receptors based on protein power spectrum
from fast Fourier transform. Amino Acids 30, 397–402.
Gao, Y., Shao, S.H., Xiao, X., Ding, Y.S., Huang, Y.S., Huang, Z.D., Chou, K.C., 2005.
Using pseudo amino acid composition to predict protein subcellular location:
approached with Lyapunov index, Bessel function, and Chebyshev filter. Amino
Acids 28, 373–376.
Gu, F., Chen, H., 2009. Evaluating long-term relationship of protein sequence by
use of D-interval conditional probability and its impact on protein structural
class prediction. Protein and Peptide Letters 16 (10), 1267–1276.
Hegalson, C.M., Jobe, T.H., 1998. The fuzzy cube and causal efficacy: representation
of concomitant mechanisms in stroke. Neural Networks 11, 549–555.
Hashimoto, H., 1983. Szpilrajn’s theorem on fuzzy orderings. Fuzzy Sets and
Systems 10, 101–108.
Homaeian, L., Kurgan, L.A., Cios, K.J., Ruan, J., Chen, K., 2007. Prediction of protein
secondary structure content for the twilight zone sequences. Proteins 69 (3),
486–498.
Jiang, T., Lin, G., Ma, B., Zhang, K., 2002. A general edit distance between RNA
structures. Journal of Computational Biology 9, 371–388.
Jamshidi, N., Edwards, J.S., Fahland, T., Church, G.M., Palsson, B.O., 2001. Dynamic
simulation of the human red blood cell matabolic network. Bioinformatics 17,
286–287.
Karakasidis, T.E., Georgiou, D.N., 2004. Partitioning elements of the periodic table
via fuzzy clustering technique. Soft Computing (Springer-Verlag) 8, 231–236.
Kawashima, S., Kanehisa, M., 2000. AAindex: amino acid index database. Nucleic
Acids Research 28, 374.
Kawashima, S., Ogata, H., Kanehisa, M., 1999. AAindex: amino acid index database.
Nucleic Acids Research 27, 368–369.
Kedarisetti, K., Kurgan, L., Dick, S., 2006. Classifier ensembles for protein structural
class prediction with varying homology. Biochemical and Biophysical Research
Communications 348 (3), 981–988.
Klir, G.J., Yuan, B., 1995. Fuzzy Sets and Fuzzy Logic (Theory and Applications).
Prentice Hall PRT, New Jersey.
Kosko, B., 1992. Neural Networks and Fuzzy Systems. Prentice-Hall, Englewood
Cliffs, NJ.
Kurgan, L.A., Stach, W., Ruan, J., 2007. Novel scales based on hydrophobicity
indices for secondary protein structure. Journal of Theoretical Biology 248,
354–366.
Kurgan, L., Chen, K., 2007. Prediction of protein structural class for the twilight
zone sequences. Biochemical and Biophysical Research Communications
357 (2), 453–460.
Lin, C.T., 1997. Adaptive subsethood for radial basis fuzzy systems. In: Kosko, B.
(Ed.), Fuzzy Engineering. Prentice-Hall, Upper Saddle River, NJ, pp. 429–464.
Liabres, M., Rossello, F., 2004. A new family of metrics for biopolymer contact
structures. Computational Biology and Chemistry 28, 21–37.
Liben-Nowell, D., 2001. On the structure of syntenic distance. Journal of
Computational Biology 8, 53–67.
Li, M., Badger, J.H., Chen, X., Kwong, S., Kearney, P., Zhang, H., 2001. An
information-based sequence distance and its application to whole mitochon-
drian phylogeny. Bioinformatics 17, 149–154.
Lin, Z., Pan, X., 2001. Accurate prediction of protein secondary structural content.
Journal of Protein Chemistry 20, 217–220.
Lin, H., Li, Q.Z., 2007a. Using pseudo amino acid composition to predict protein
structural class: approached by incorporating 400 dipeptide components.
Journal of Computational Chemistry 28, 1463–1466.
Lin, H., Li, Q.Z., 2007b. Predicting conotoxin superfamily and family by using
pseudo amino acid composition and modified Mahalanobis discriminant.
Biochemistry and Biophysics Research Communication 354, 548–551.
Lin, W.-Z., Xiao, X., Chou, K.-C., 2009. GPCR-GIA: a web-server for identifying
G-protein coupled receptors and their families with grey incidence analysis.
Protein Engineering, Design and Selection 22, 699–705.
Liu, H., Wang, M., Chou, K.C., 2005. Low-frequency Fourier spectrum for predicting
membrane protein types. Biochemistry and Biophysics Research Communica-
tion 336, 737–739.
Menconi, G., 2005. Sublinear growth of information in DNA sequences. Bulletin of
Mathematical Biology 67, 737–759.
Morgenstern, B., 2002. A simple and space-efficient fragment-chaining algorithm
for alignment of DNA and protein sequences. Applied Mathematics Letter 15
(1), 11–16.
Moulton, V., Zuker, M., Steel, M., Pointon, R., Penny, D., 2000. Metrics on RNA
secontary structures. Journal of Computational Biology 7 (1/2), 277–292.
Mondal, S., Bhavna, R., Mohan Babu, R., Ramakumar, S., 2006. Pseudo amino acid
composition and multi-class support vector machines approach for conotoxin
superfamily classification. Journal of Theoretical Biology 243, 252–260.
Mocz, G., 1995. Fuzzy cluster analysis of simple physicochemical properties of
amino acids for recognizing secondary structure in proteins. Protein Science 4,
1178–1187.
Mundra, P., Kumar, M., Kumar, K.K., Jayaraman, V.K., Kulkarni, B.D., 2007. Using
pseudo amino acid composition to predict protein subnuclear localization:
approached with PSSM. Pattern Recognition Letters 28, 1610–1615.
Nieto, J.J., Torres, A., Vazquez-Trasande, M.M., 2003. A metric space to study
differences between polynucleotides. Applied Mathematics Letter 16, 1289–1294.
Nieto, J.J., Torres, A., 2003. Midpoints for fuzzy sets and their application in
medicine. Artificial Intelligence in Medicine 17, 81–101.
Nieto, J.J., Torres, A., Georgiou, D.N., Karakasidis, T.E., 2006. Fuzzy polynucleotide
spaces and metrics. Bulletin of Mathematical Biology 68, 703–725.
 D.N. Georgiou et al. / Journal of Theoretical Biology 267 (2010) 95–105
Paun, Gh., Rozenberg, G., Saloma, A., 1998. DNA Computing: New Computing
Paradigms. Springer, Berlin.
Percus, J., 2002. Mathematics of Genome Analysis. Cambridge University Press,
Cambridge.
Perez-Montoto, L.G., Santana, L., Gonzalez-Diaz, H., 2009. Scoring function for
DNA-drug docking of anticancer and antiparasitic compounds based on
spectral moments of 2D lattice graphs for molecular dynamics trajectories.
European Journal of Medicinal Chemistry 44 (11), 4461–4469.
Schneider, G., Wrede, P., 1994. The rational design of amino acid sequences by
artificial neural networks and simulated molecular evolution: de novo design
of an idealized leader peptidase cleavage site. Biophysical Journal 66, 335–344.
Shen, H.B., Chou, K.C., 2005a. Using optimized evidence-theoretic K-nearest neighbor
classifier and pseudo amino acid composition to predict membrane protein types.
Biochemical and Biophysical Research Communications 334, 288–292.
Shen, H.B., Chou, K.C., 2006. Ensemble classifier for protein fold pattern
recognition. Bioinformatics 22, 1717–1722.
Shen, H.B., Chou, K.C., 2007a. Hum-mPLoc: an ensemble classifier for large-scale
human protein subcellular location prediction by incorporating samples with
multiple sites. Biochemical and Biophysical Research Communications 355,
1006–1011.
Shen, H.B., Chou, K.C., 2007b. EzyPred: a top–down approach for predicting
enzyme functional classes and subclasses. Biochemical and Biophysical
Research Communications 364, 53–59.
Shen, H.B., Chou, K.C., 2007c. Signal-3L: a 3-layer approach for predicting signal
peptide. Biochemical and Biophysical Research Communications 363,
297–303.
Shen, H.B., Yang, J., Chou, K.C., 2006. Fuzzy KNN for predicting membrane protein
types from pseudo amino acid composition. Journal of Theoretical Biology 240,
9–13.
Shen, H.-B., Chou, K.-C., 2009a. Gpos-mploc: a top–down approach to improve the
quality of predicting subcellular localization of gram-positive bacterial
proteins. Protein and Peptide Letters 16, 1478–1484.
Shen, H.-B., Chou, K.-C., 2009b. A top–down approach to enhance the power of
predicting human protein subcellular localization: Hum-mPLoc 2.0. Analytical
Biochemistry 394, 269–274.
Shen, H.B., Chou, K.C., 2005b. Predicting protein subnuclear location with
optimized evidence-theoretic K-nearest classifier and pseudo amino acid
composition. Biochemical and Biophysical Research Communications 337,
752–756.
Sadegh-Zadeh, K., 2000. Fuzzy genomes. Artificial Intelligence in Medicine 18, 1–28.
Sadovsky, M.G., 2003. The method to compare nucleotide sequences based on the
minimum entropy principle. Bulletin of Mathematical Biology 65, 309–322.
Shannon, C.E., 1948. A mathematical theory of communication. The Bell Systems
Technical Journal 27, 379–423.
Sadegh-Zadeh, K., 1999. Fundamentals of clinical methodology: 3. Nosology.
Artificial Intelligence in Medicine 17, 87–108.
Samaras, P., Kungolos, A., Karakasidis, T., Georgiou, D., Perakis, K., 2001. Statistical
evaluation of PCDD/F emission data during solid waste combustion by fuzzy
clustering techniques. Journal of Environmental Science and Health, Marcel
Dekker, Inc. (part A) 36, 153–161.
Stephen, Y.L., Freeland, J., 2008. A quantitative investigation of the chemical space
surrounding amino acid alphabet formation. Journal of Theoretical Biology
250, 349–361.
Tang, B., 2000. Evaluation of some DNA cloning strategies. Computers and
Mathematics with Application 39 (11), 43–48.
Torres, A., Nieto, J.J., 2003. The fuzzy polynucleotide space: basic properties.
Bioinformatics 19 (5), 587–592.
Terano, T., Asai, K., Sugeno, M., 1992. Fuzzy Systems Theory and its Applications.
Academic Press, Harcount Brace Jovanovich Publishers, San Diego, California.
105
Torres A., Nieto J.J., 2006, Fuzzy logic in medicine and bioinformatics. Journal of
Biomedicine and Biotechnology, Article ID 91908.
Trinquier, G., Sanejouand, Y.-H.I., 1998. Which effective property of amino acids is
best preserved by the genetic code? Protein Engineering 11 153–169.
Vilar, S., Gonzalez-Diaz, H., Santana, L., Uriarte, E., 2009. A network-QSAR model
for prediction of genetic-component biomarkers in human colorectal cancer.
Journal of Theoretical Biology 261 (3), 449–458.
Wolfenden, R., 2007. Experimental measures of amino acid hydrophobicity and the
topology of transmembrane and globular proteins. Journal of Cell Biology 177
i10–i10.
Wang, M., Yang, J., Liu, G.P., Xu, Z.J., Chou, K.C., 2004. Weighted-support vector
machines for predicting membrane protein types based on pseudo amino acid
composition. Protein Engineering, Design, and Selection 17, 509–516.
Wang, S.Q., Yang, J., Chou, K.C., 2006. Using stacked generalization to predict
membrane protein types based on pseudo amino acid composition. Journal of
Theoretical Biology 242, 941–946.
Xiao, X., Shao, S.H., Huang, Z.D., Chou, K.C., 2006a. Using pseudo amino acid
composition to predict protein structural classes: approached with complexity
measure factor. Journal of Computational Chemistry 27, 478–482.
Xiao, X., Shao, S., Ding, Y., Huang, Z., Chen, X., Chou, K.C., 2005b. Using cellular
automata to generate image representation for biological sequences. Amino
Acids 28, 29–35.
Xiao, X., Shao, S., Ding, Y., Huang, Z., Chen, X., Chou, K.C., 2005c. An application of
gene comparative image for predicting the effect on replication ratio by HBV
virus gene missense mutation. Journal of Theoretical Biology 235, 555–565.
Xiao, X., Shao, S., Ding, Y., Huang, Z., Huang, Y., Chou, K.C., 2005a. Using complexity
measure factor to predict protein subcellular location. Amino Acids 28, 57–61.
Xiao, X., Shao, S.H., Ding, Y.S., Huang, Z.D., Chou, K.C., 2006b. Using cellular
automata images and pseudo amino acid composition to predict protein sub-
cellular location. Amino Acids 30, 49–54.
Xiao, X., Lin, W.Z., Chou, K.C., 2008a. Using grey dynamic modeling and pseudo
amino acid composition to predict protein structural classes. Journal of
Computational Chemistry 29, 2018–2024.
Xiao, X., Wang, P., Chou, K.C., 2008b. Predicting protein structural classes with
pseudo amino acid composition: an approach using geometric moments of
cellular automaton image. Journal of Theoretical Biology 254, 691–696.
Xiao, X., Wang, P., Chou, K.C., 2009a. Predicting protein quaternary structural
attribute by hybridizing functional domain composition and pseudo amino
acid composition. Journal of Applied Crystallography 42, 169–173.
Xiao, X., Wang, P., Chou, K.C., 2009b. GPCR-CA: a cellular automaton image
approach for predicting G-protein-coupled receptor functional classes. Journal
of Computational Chemistry 30, 1414–1423.
Xiao, X., Lin, W.Z., 2009. Application of protein grey incidence degree measure to
predict protein quaternary structural types. Amino Acids 37, 741–749.
Xiao, X., Wang, P., Chou, K.C., 2010. Quat-2L: a web-server for predicting protein
quaternary structural attributes. Molecular Diversity, DOI:10.1007/
s11030-010-9227-8.
Zhang, T.L., Ding, Y.S., Chou, K.C., 2008. Prediction protein structural classes with
pseudo-amino acid composition: approximate entropy and hydrophobicity
pattern. Journal of Theoretical Biology 250, 186–193.
Zhang, Z.D., Sun, Z.R., Zhang, C.T., 2001. A new approach to predict the helix/strand
content of globular proteins. Journal of Theoretical Biology 208, 65–78.
Zhou, X.B., Chen, C., Li, Z.C., Zou, X.Y., 2007. Using Chou’s amphiphilic pseudo-
amino acid composition and support vector machine for prediction of enzyme
subfamily classes. Journal of Theoretical Biology 248, 546–551.
Zimmermann, H.J., 1991. Fuzzy Theory and its Applications. Kluwer Academic
Publishers, New York.
Zaus, M., 1999. Crisp and Soft Computing with Hypercubical Calculus.
Physica-Verlag, Heideberg.