International Journal of Bioinformatics Research

ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, 2013, pp.-285-288.

Available online at http://www.bioinfopublication.org/jouarchive.php?opt=&jouid=BIJ0000001

ANALYSIS AND PREDICTION OF MAJOR BLOOD PROTEINS BASED ON THEIR AMINO ACID
AND DIPEPTIDE COMPOSITION

MUTHUKRISHNAN S.1,2, PURI M.1,3* AND LEFEVRE C.3

1Protein Biotechnology Laboratory, Department of Biotechnology, Punjabi University, Patiala- 147002, Punjab, India.
2Instituteof Microbial Technology, Sector-39A, Chandigarh- 160036, India.
3Centre for Chemistry and Biotechnology, Deakin University, Victoria, Australia.

*Corresponding Author: Email- munish.puri@deakin.edu.au

Received: August 08, 2013; Accepted: September 18, 2013

Abstract- A method has been developed for predicting blood proteins using the SVM based machine learning approach. In this prediction
method a two-step strategy was deployed to predict blood proteins and their subclasses. We have developed models of blood proteins and
achieved the maximum accuracies of 90.57% and 91.39% with Matthews correlation coefficient (MCC) of 0.89 and 0.90 using single amino
acid and dipeptide composition respectively. Furthermore, the method is able to predict major subclasses of blood proteins; developed based
on amino acid (AC) and dipeptide composition (DC) with a maximum accuracy 90.38%, 92.83%, 87.41%, 92.52% and 85.27%, 89.07%,
94.82%, 86.31 for albumin, globulin, fibrinogen, and regulatory proteins respectively. All modules were trained, tested, and evaluated using the
five-fold cross-validation technique.
Keywords- Major Blood Proteins, Amino Acid Composition, Dipeptide Composition, SVM, five-fold cross-validation technique

Citation: Muthukrishnan S., Puri M. and Lefevre C. (2013) Analysis and Prediction of Major Blood Proteins Based on Their Amino Acid and
Dipeptide Composition. International Journal of Bioinformatics Research, ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, pp.-285-
288.

Copyright: Copyright©2013 Muthukrishnan S., et al. This is an open-access article distributed under the terms of the Creative Commons Attrib-
ution License, which permits unrestricted use, distribution and reproduction in any medium, provided the original author and source are credit-
ed.

Introduction teins are playing an important role in the regulation of a cellular

Determination of protein functions is one of the most challenging activity and many different functions in the immune system. Due to
problems in the genomic era. Enormous amounts of protein se- their great function of blood, a classification prediction system has
quences are available in the database as raw sequence data. Using been developed in order to facilitate better understanding of their
various computational approaches, these data’s are being pro- roles. The superior facility of classification system using machine
cessed into meaningful biological information’s. The support vector learning based approach rather than experimental techniques is
machine (SVM) based prediction system is fully automatic and relia- apparent. Currently, there is no classification of blood proteins avail-
ble. It has been used in many applications including sub-cellular able based on amino (AC) and dipeptide composition. Support vec-
localization, protein secondary structure prediction, and micro array tor machine (SVM) is one of the promising kernel based machine
learning for building effective model for predicting class labels of
data analysis of proteins [1-4]. However, no direct method is cur-
unknown protein data. Therefore, in this study, we have developed
rently available to predict blood proteins. Thus analysis and predic-
an integrative SVM based prediction with a two step approach to
tion studies of blood proteins are important for researchers.
predict the blood proteins and further classify them into different
Blood proteins found in blood plasma are also called serum pro- classes. The method presented is highly specific and sensitive to
teins. Major blood proteins are albumin, globulin, fibrinogen, and predict the blood proteins.
regulatory proteins [5,6]. Sixty percent of plasma proteins are made
up of albumins [7], which are major contributors to the osmotic pres- Results
sure of plasma and which assists in the transport of lipids and ster- To develop a prediction platform for blood proteins, amino acid
oid hormones. Globulins make up thirty five percent of plasma pro- composition dataset of all blood-proteins were made using five fold
teins and are used in the transport of ions, hormones, and lipids cross-validation technique. The SVM module was developed using
thus assisting in immune function [8]. Four percent is fibrinogen and blood-protein and non-blood protein training sets. Dataset was di-
it is essential in the clotting of blood when converted into insoluble vided into five equal sets randomly, four sets were used for training,
fibrin [9]. Regulatory proteins, which make up less than one percent and the remaining set was used for testing [10,11]. This process
of plasma proteins, are proteins such as enzymes, proenzymes and was repeated five times to test each protein at least once [12]. We
hormones. The main functions of the blood proteins are transporting labeled all blood-proteins as positive proteins and non blood-
lipids, hormones, vitamins and metal molecules. Thus, these pro- proteins were used as negative proteins. The results demonstrated

International Journal of Bioinformatics Research

ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, 2013

|| Bioinfo Publications || 285

 Analysis and Prediction of Major Blood Proteins Based on Their Amino Acid and Dipeptide Composition

that the method can differentiate blood-proteins from non blood- The dipeptide prediction accuracy was improved significantly over
proteins with great accuracy of 90.57% in 0.89 of MCC at a default single amino acid prediction. Therefore, the prediction accuracy can
cutoff score of 0. The best result was obtained using an optimal be increased using a wide range of information about a protein. The
RBF kernel with parameters g =3, C 375. The dipeptide composition sensitivity and specificity also has been calculated for blood pro-
method was also tested and achieved 91.39% accuracy with 0.90 of teins and subclasses, shown in [Table-1] & [Table-2].
the MCC. In average, all amino and dipeptide composition analysis
Analysis of Amino Acids
of blood-proteins were significantly different from non blood-proteins
[Table-1]. We have applied the same method for predicting the Determining the relative amino acid composition of a protein will
classification of major blood-proteins. Here we took one class of give a characteristic profile for protein. This amino acid analysis
blood proteins as positive and all other classes for negative exam- profile provides enough information to identify major blood-proteins.
ples. This was repeated to all other classes of blood proteins. We Here, we used the total number of amino acid divided by the total
prepared models of each blood protein classes based on their ami- number of amino acids in protein. The average amino acid compo-
no acid as well as dipeptide composition with different optimized sition of blood proteins has been calculated which shown in [Fig-1]
SVM kernels parameters. This indicates that each class of blood- and [Fig-2] with non blood proteins. In this analysis results shows
proteins can be discriminated from other classes of proteins based that Cys, Pro, Ser and Tyr are higher in blood proteins than the non
on their amino acid and dipeptide composition [13-16]. In amino blood proteins. In sub classes of blood proteins Leu residues are
acid composition we achieve maximum accuracy as shown in the higher in all classes, Cys, Asp, Gly and Asn are higher in fibrinogen
[Table-2], 90.38%, 92.83%, 87.41%, 92.52% with MCC of 0.88, proteins than other classes. Overall regulatory and globulin protein
0.85, 0.63, and 0.43 of albumin, globulin, fibrinogen, and regulatory is having a similar percentage in all residues.
proteins respectively. Regulatory and globulin proteins are showing
maximum accuracy (>92%) compared to other classes of blood
proteins. Since the simple amino acid composition provides only
information about frequency but not about local order of residues,
additional SVM module based on dipeptide composition were devel-
oped. While in amino acid composition SVM were provided with an
input vector of dimention 20 for amino acid composition, a vector of
400 dipeptide composition was used for the dipeptide module which
was optimized by g =10, C=400. The results are shown in [Table-2]
and we achieved the maximum accuracy of 85.27%, 89.07%,
94.82%, 86.31% to 0.87, 0.88, 0.87, and 0.70 of the MCC for albu-
min, globulin, fibrinogen, and regulatory proteins respectively. The Fig. 1- Average Amino acid composition chart of blood vs. non-
fibrinogen protein shows the highest accuracy 94.82% among sub- blood proteins
class proteins.
Table 1- Performance of various SVM modules of blood-protein
predictions developed using various types of compositions; amino
acids (AC) and dipeptides (DC).
Parameters
Methods ACC(%) SN(%) SP(%) MCC
γ C
AC 90.57 97.16 84.69 0.89 3 375
DC 91.4 96.77 86.6 0.9 10 400
AC- Amino acid composition, DC- dipeptide composition, ACC- accuracy,
SN- Sensitivity, SP- specificity, MCC- Matthews correlation coefficient,
C- tradeoff value, γ- gamma factor (a parameter in RBF kernel)

Fig. 2- Average amino acid percentage chart of subclasses blood

Table 2- Performance of various SVM modules of blood-protein
proteins.
classifications (albumin, globulin, fibrinogen and regulatory) predic-
tions developed using various types of compositions; amino acids Discussion
(AC) and dipeptides (DC). Blood proteins serve many different functions, such as circulatory
Proteins Methods ACC(%) SN(%) SP(%) MCC
Parameters transport molecules for (lipid hormones, vitamins, and metals), pro-
γ C tease inhibitors, and regulation of cellular activities, including the
AC 90.38 88.89 90.6 0.88 15 350 immune system. Based on their importance, we have decided to
Albumin
DC 85.27 94.1 84 0.87 2 375 develop a method using SVM for prediction of these proteins. Here,
AC 92.83 80.21 93.39 0.85 50 350
Globulin we have described amino and dipeptide based method which is
DC 89.07 89.58 89.05 0.88 5 200
helpful in differentiating various blood proteins. This method is use-
AC 87.41 97.94 49.4 0.63 100 75
Fibrinogen
DC 94.82 99.92 76.03 0.87 3 450
full to show whether a newly discovered protein sequence belongs
AC 92.53 27.5 94.88 0.43 15 450 to blood proteins and identify its subfamilies. This method is a high-
Regulatory ly accurate method and able to perform classification separation
DC 86.32 66.25 87.05 0.7 25 500
AC- Amino acid composition, DC- dipeptide composition, ACC- accuracy, properly. Finally, our results demonstrate that using the concept of
SN- Sensitivity, SP- specificity, MCC- Matthews correlation coefficient, amino and dipeptide using SVM is a successful method for predict-
C- tradeoff value, γ- gamma factor (a parameter in RBF kernel) ing these proteins.

International Journal of Bioinformatics Research

ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, 2013

|| Bioinfo Publications || 286

 Muthukrishnan S., Puri M. and Lefevre C.

Methods
(5)
Dataset
The final data set of blood proteins including subfamilies consist
717 (albumin 91, globulin 33, fibrinogen 564 and regulatory 29). As (6)
negative set 899 belonging to protease family were selected ran-
domly. These protein sequences were obtained from Uniprot and where, TP, TN, FP, FN’s are the number of true positives, true neg-
Expasy server. In this dataset “fragments”, “isoforms”, “potentials”, atives, false positives and false negatives respectively. TP and TN
“similarity”, or “probables” in comment field were removed, to avoid are number of correctly classified blood proteins and non blood
bias in the classifier. We have used 90% cutoff to generate non- proteins respectively. FP and FN are incorrectly classified as blood
redundant dataset of both blood and non-blood sequences. and non blood proteins.
Support Vector Machine (SVM) Prediction System
In the present study, a free downloadable package of SVM, The prediction of blood and blood related proteins are a multi-class
SVM_light has been used to classify major blood-protein sequenc- classification problem. To handle this multi-class situation, we have
es. This software enables the users to define a number of parame- to design a series of binary SVMs. For N class classification, N
ters as well as the choice of inbuilt kernel, such as a radial basis SVMs was constructed. The ith SVM will do training with all sam-
function (RBF) or a polynomial kernel (of given degree) [17]. In this ples of the ith subfamily being labeled as positive, and the samples
study, all parameters of a kernel were kept constant, except for the of all other subfamilies being labeled as negative. The SVMs trains
regulatory parameter C. The experimentation was conducted by in this way will reefer to as 1-v-r SVMs. In this classification ap-
using various types of kernels such as polynomial and radial base proach, each of the unknown proteins will achieve four scores. An
function. The SVMs required a fixed number of inputs for training, unknown protein will be classified into the subfamily that corre-
thus necessitating a strategy for encapsulating the global infor- sponds to the 1-v-r SVM with the highest output score.
mation about the proteins of variable length in a fixed length format.
Acknowledgements
The fixed length format was obtained from protein sequences of
variable length using amino acid and dipeptide composition. It has The authors are thankful to the Department of Biotechnology, Pun-
been successfully applied to numerous classification and pattern jabi University, Patiala and Institute of Microbial Technology, Chan-
recognition problems such as classification of microarray data, pro- digarh, India, for providing necessary facility to conduct research
tein secondary structure prediction and sub cellular localization [3,4, work. The authors acknowledge Deakin University, Australia for
18]. providing web server for this study.
Amino Acid Composition Conflict of Interest: None declared.
Amino acid composition is the fraction of each amino acid in a pro- References
tein. The fraction of all 20 natural amino acids was calculated by [1] Park K.J., Kanehisa M. (2003) Bioinformatics, 19, 1656-1663.
using [Eq-1] [19,20].
[2] Yu C.S., Chen Y.C., Lu C.H., Hwang J.K. (2006) Proteins, 64,
(1) 643-651.
[3] Chang D.T., Ou Y.Y., Hung H.G., Yang M.H., Chen C.Y.,
where i can be any amino acid. Oyang Y.J. (2008) BMC Res. Notes, 1(1), 51.
Dipeptide Composition [4] Watson M., Dukes J., Abu-Median A.B., King D.P., Britton P.,
Dipeptide composition is used to encapsulate the global information Detecti V. (2007) Genome Biol., 8(9), R190.
about each protein sequence, which gives a fixed pattern length of [5] Anderson N.L. and Anderson N.G. (1977) Proceeding of the
400 (20 X 20). The fraction of each dipeptide was calculated using National Academy of Sciences, 74, 5421-5425.
[Eq-2] [19]. [6] Adkins J.N., Varnum S.M., Auberry K.J., Moore R.J., Angell
N.H., Smith R.D., Springer D.L., Pounds J.G. (2002) Molecular
(2) and Cellular Proteomics, 1, 947-955.
where dep (i+1) is one out of 400 dipeptides. [7] Zunszain P.A., Ghuman J., Komatsu T., Tsuchida E., Curry S.
(2003) BMC Structural Biology, 3(1), 6 10.1186/1472-6807-3-6.
Evaluation of Performance
[8] Roux K.H. (1999) Int. Arch Allergy Immunol., 120, 85-99.
In this study, 5-fold cross-validation technique was adopted accord-
ing to which dataset was partitioned randomly into five equal sub- [9] Laurens N., Koolwijk P., de Maat M.P. (2006) J. Thromb. Hae-
sets. The training and testing were carried out five times, each time most., 4(5), 932-939.
using one subset for testing and remaining 4 subsets for training. [10] Bhasin M., Garg A., Raghava G.P.S. (2005) Bioinformatics, 21,
The performance of each classifier is measured in terms of accura- 2522-2524.
cy (ACC), sensitivity (SN), specificity (SP) and Matthews correlation [11] Kumar M., Gromiha M.M., Raghava G.P.S. (2007) BMC Bioin-
coefficient (MCC) by standard [Eq-3] to [Eq-6] [20]. formatics, 8, 463.
[12] Garg A., Bhasin M., Raghava G.P.S. (2005) J. Biol. Chem.,
(3)
280, 14427-14432.
[13] Yu C.S., Chen Y.C., Lu C.H., Hwang J.K. (2006) Proteins, 64,
(4) 643-651.

International Journal of Bioinformatics Research

ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, 2013

|| Bioinfo Publications || 287

 Analysis and Prediction of Major Blood Proteins Based on Their Amino Acid and Dipeptide Composition

[14] Cai C.Z., Han L.Y., Ji Z.L., Chen X., Chen Y.Z. (2003) Nucleic
Acids Research, 31, 3692-3697.
[15] Muthukrishnan S., Garg A., Raghava G.P.S. (2007) Genomics,
Proteomics & Bioinformatics, 5, 250-252.
[16] Lata S., Sharma B.K., Raghava G.P.S. (2007) BMC Bioinfor-
matics, 8, 263.
[17] Chou K.C., Shen H.B. (2007) Biochem Biophys Res. Commun.,
360, 339-345.
[18] Nair R., Rost B. (2003) Proteins, 53(4), 917-930.
[19] Bhasin M., Raghava G.P.S. (2004) J. Biol. Chem., 279, 23262-
23266.
[20] Kumar M., Verma R., Raghava G.P.S. (2006) J. Biol. Chem.,
281, 5357-5363.

International Journal of Bioinformatics Research

ISSN: 0975-3087 & E-ISSN: 0975-9115, Volume 5, Issue 1, 2013

|| Bioinfo Publications || 288