Journal of ISSN: 2377-4282JNMR

Nanomedicine Research
Review Article
Volume 3 Issue 6 - 2016
An Overview of the Protein Thermostability Prediction: Databases and Tools
Maryam Mahmoudi1, Seyyed Shahriar Arab1*, Javad Zahiri2 and Yasaman Parandian2
1Department of Biophysics, Tarbiat Modares University (TMU), Iran
2Bioinformatics and Computational Omics Lab (BioCOOL), Trabiat Modares University (TMU), Iran
Received: March 12, 2016 | Published: July 05, 2016
*Corresponding author: Seyyed Shahriar Arab, Department of Biophysics, School of Biological Sciences, Tarbiat Modares University (TMU), Tehran, Iran, Email:
Citation: Mahmoudi M, Arab AA, Zahiri J, Parandian Y (2016) An Overview of the Protein Thermostability Prediction: Databases and Tools. J Nanomed Res 3(6): 00072. DOI: 10.15406/jnmr.2016.03.00072

Abstract

Thermophilic proteins are characterized as high thermal stability proteins while mesophilic proteins are stable at lower temperatures. These types of proteins have numerous applications regarding protein engineering, drug design and industrial processes. Studies showed that thermal stability is strongly related to structural and sequential properties in thermophilic proteins. Some computational studies were being taken to identify the mentioned properties in heat resistant proteins. This paper reviews the studies of protein thermostability prediction and gives an introduction to the thermal stability related tools and databases.

Keywords: Rotein thermostability; Thermophilic proteins; Mesophilic proteins; Databases; computational methods; Bioinformatics

Introduction

Environmental temperature plays an important role in the cell life [1]. There are four classes of organism in relation to their optimal growth temperature namely hyperthermophile (>80â—¦C), thermophile (45-80â—¦C), mesophile (20-45â—¦C) and psychrophile (<20 â—¦C) [2]. Thermal stability is defined as the ability of material to resist changes in physical structure or chemical irreversibility, or spatial structure stability of polypeptide chains at high temperatures [3]. Studies showed that thermal stability of thermophilic proteins is related to a series of protein sequential and structural properties [4]. A small number of these mentioned properties are going to be introduced in this paper. Also, the amino acid compositions difference had been studied in mesophilic and thermophilic proteins [3,5-7]. For instance, Zhang and Gromiha research shows that Lys, Arg, Glu and Pro were higher and Ser, Met, Asp and Thr were lower in number of thermophilic than the of mesophilic proteins number [6,8] (Figure 1). Protein secondary structure stability like alpha-helix is considered as a necessary factor for thermal stability [6]. Studies suggested that thermal-stability is increased by certain characteristics in proteins. These characteristics are: increased number of hydrogen bonds [7], salt bridges, ion pairs [9], aromatic clusters [8], sidechain-sidechain interactions, electrostatic interactions of charged residues [9] and hydrophobic interactions [5].

Figure 1: Comparison of Amino acid composition in thermophilic and mesophilic proteins [6,8].

Protein’s Thermal Stability Prediction Methods

 Protein’s thermal stability can be predicted based on sequence or structure. Both mentioned methods and their corresponding advantages and limitations have been discussed here in further detail. Table 1 demonstrates an overview of the thermal stability prediction methods.

Sequence/Structure Feature

Algorithm

Reference

Amino acid sequence

Support vector machine

[10]

Primary structure

LogitBoost

[12]

Amino acid sequence and residues and dipeptide composition

Neural network

[8]

Primary, secondary and tertiary structure information

Decision tree

[11]

Amino acid distribution and dipeptide composation

Support vector machine

[13]

Amino acid composition-based similarity distance

KNN-ID

[1]

Dipeptide composation

Statistical Methods

[14]

Amino acid sequence

Genetic Algorithm

[15]

Thermodynamic parameters

Statistical Potentials

[16]

Table 1: An overview of protein thermostability prediction studies.

Sequence based prediction

This method utilizes sequence information of proteins; for instance, distribution of amino acid and di-peptide composition for discrimination of thermophilic and mesophilic proteins. Studies revealed the differences between amino acid and di-peptide composition in thermophilic and mesophilic proteins. For example, the frequency of Lys, Arg, Glu and Pro was higher in thermophilic than mesophilic protein [8,10] . These studies also show that the occurrences of EE, KK, RR, PP, KI, VV, VE, KE, and VK were higher in thermophilic proteins while QQ, AA, EQ, LL, NN, QT had lower occurrences [6]. In addition, the frequency of charged, hydrophobic and aromatic amino acids in thermophilic protein is higher than mesophilic ones [3]. Moreover, the correlation between protein amino acid composition and its biological function has been proven [1]. So, the protein sequence analysis provides valuable information to predict protein thermostability; particularly whenever the structural information of proteins is not available.

Structure based prediction

The studies of protein thermostability prediction are based on protein structures utilized protein secondary and tertiary information for discrimination of thermophilic and mesophilic proteins. Important features considered in this studies include amount of secondary structure, ion pairs, hydrogen bonds, disulfide bonds and accessible surface area [11]. Although the thermal stability is directly related to the protein structure stability [11]. Regarding the fact that structural and sequential features affect the thermal stability, applying the both mentioned features at the same time leads to a more accurate, precise prediction. The protein structural information may not be always available; This restrains structure based protein thermostability prediction.

  1. Protein’s thermal stability prediction procedure: Several machine learning methods have been applied to predict protein thermostability. Here, we briefly review these methods. Figure 2 provides an illustration of these methods. As illustrated in the figure, in order to predict the thermal stability of proteins, at first, a dataset of thermophil and mesophil proteins is collected from the related databases. Then, proteins are analyzed based on their sequential and structural characteristics. The goal in this stage is to select those features which are significantly important regarding protein thermostability prediction. It should be noted here that considering the structural and sequential features at the same time can produce more precise results. In the next stage, the dataset is going to be divided into the train and test datasets. The train dataset is then used for learning the machine learning algorithm while the test dataset is used to evaluate the model.
Figure 2: Thermal stability prediction procedure.

Prediction algorithms based on machine learning methods

The following section introduces a few machine learning algorithms. The selected algorithm is going to distinguish the thermophile from mesophile proteins.

  1. Support vector machines (SVMs): Support vector machines is an machine learning method for classification two classes of data and many kind of kernel functions can be used for classification in this algorithm [17].
  2. Artificial Neural Networks (ANN): The ANN concept is inspired by the neural structure of the brain. In this model of prediction, the system is supposed to learn from data - a large number of inputs and solve a wide variety of tasks. ANN software packages can be downloaded from Open NN (Available online: http://www.cimne.com/flood/download.asp)[15].
  3. Decision Tree: A decision tree is popular machine learning algorithm in bioinformatics and computational biology. It uses a tree-like graph or model of decisions and their possible consequences to classify input instances.

Performance Measures

Assessing a prediction tool is a critical task. Table 2 describes commonly used measures for performance prediction assessment: accuracy, sensitivity, specificity, strength, MCC, precision, F-measure and area under the ROC curve (AUC). These measures based on the following four basic parameters:

Expression

A brief description

Accuracy= TP+TN TP+TN+FP+FN MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVCI8FfYJH8YrFfeuY=Hhbbf9v8qqaqFr0xc9pk0xbb a9q8WqFfeaY=biLkVcLq=JHqpepeea0=as0Fb9pgeaYRXxe9vr0=vr 0=vqpWqaaeaabiGaciaacaqabeaadaqaaqaaaOqaaKqzGeaeaaaaaa aaa8qacaWGbbGaam4yaiaadogacaWG1bGaamOCaiaadggacaWGJbGa amyEaiabg2da9KqbaoaalaaakeaajugibiaabsfacaqGqbGaey4kaS Iaaeivaiaab6eaaOqaaKqzGeGaaeivaiaabcfacqGHRaWkcaqGubGa aeOtaiabgUcaRiaabAeacaqGqbGaey4kaSIaaeOraiaab6eaaaaaaa@4E25@

percent of correct prediction

Sensitivity= TP TP+FN MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVCI8FfYJH8YrFfeuY=Hhbbf9v8qqaqFr0xc9pk0xbb a9q8WqFfeaY=biLkVcLq=JHqpepeea0=as0Fb9pgeaYRXxe9vr0=vr 0=vqpWqaaeaabiGaciaacaqabeaadaqaaqaaaOqaaKqzGeaeaaaaaa aaa8qacaWGtbGaamyzaiaad6gacaWGZbGaamyAaiaadshacaWGPbGa amODaiaadMgacaWG0bGaamyEaiabg2da9Kqbaoaalaaakeaajugibi aadsfacaWGqbaakeaajugibiaadsfacaWGqbGaey4kaSIaamOraiaa d6eaaaaaaa@49AF@

percent of correctly predicted positive

Specificity= TN TN+FP MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVCI8FfYJH8YrFfeuY=Hhbbf9v8qqaqFr0xc9pk0xbb a9q8WqFfeaY=biLkVcLq=JHqpepeea0=as0Fb9pgeaYRXxe9vr0=vr 0=vqpWqaaeaabiGaciaacaqabeaadaqaaqaaaOqaaKqzGeaeaaaaaa aaa8qacaWGtbGaamiCaiaadwgacaWGJbGaamyAaiaadAgacaWGPbGa am4yaiaadMgacaWG0bGaamyEaiabg2da9Kqbaoaalaaakeaajugibi aabsfacaqGobaakeaajugibiaabsfacaqGobGaey4kaSIaaeOraiaa bcfaaaaaaa@4972@

percent of correctly predicted negative

Precision= TP TP+FP MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVCI8FfYJH8YrFfeuY=Hhbbf9v8qqaqFr0xc9pk0xbb a9q8WqFfeaY=biLkVcLq=JHqpepeea0=as0Fb9pgeaYRXxe9vr0=vr 0=vqpWqaaeaabiGaciaacaqabeaadaqaaqaaaOqaaKqzGeaeaaaaaa aaa8qacaWGqbGaamOCaiaadwgacaWGJbGaamyAaiaadohacaWGPbGa am4Baiaad6gacqGH9aqpjuaGdaWcaaGcbaqcLbsacaWGubGaamiuaa GcbaqcLbsacaWGubGaamiuaiabgUcaRiaadAeacaWGqbaaaaaa@47A7@

Positive Predictive Value

Fmeasure= 2 ×Presion ×Sensitivity Presion +Sensitivity MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVCI8FfYJH8YrFfeuY=Hhbbf9v8qqaqFr0xc9pk0xbb a9q8WqFfeaY=biLkVcLq=JHqpepeea0=as0Fb9pgeaYRXxe9vr0=vr 0=vqpWqaaeaabiGaciaacaqabeaadaqaaqaaaOqaaKqzGeaeaaaaaa aaa8qacaWGgbGaeyOeI0IaamyBaiaadwgacaWGHbGaam4Caiaadwha caWGYbGaamyzaiabg2da9KqbaoaalaaakeaajugibiaaikdacaqGGc Gaey41aqRaamiuaiaadkhacaWGLbGaam4CaiaadMgacaWGVbGaamOB aiaacckacqGHxdaTcaWGtbGaamyzaiaad6gacaWGZbGaamyAaiaads hacaWGPbGaamODaiaadMgacaWG0bGaamyEaaGcbaqcLbsacaWGqbGa amOCaiaadwgacaWGZbGaamyAaiaad+gacaWGUbGaaiiOaiabgUcaRi aadofacaWGLbGaamOBaiaadohacaWGPbGaamiDaiaadMgacaWG2bGa amyAaiaadshacaWG5baaaaaa@6CBF@

The harmonic mean of sensitivity
and specificity

Table 2: Commonly used measures for performance assessment in protein thermostability prediction.

  1. True positive (TP): The number of thermophile proteins, which have been correctly predicted as thermophile.
  2. True negative (TN): The number of mesophile proteins, which have been correctly predicted by the prediction method as mesophile.
  3. False positive (FP): The number of mesophile proteins, which have been incorrectly predicted as thermophile.
  4. False negative (FN): The number of thermophile proteins, which have been incorrectly predicted by the prediction method as mesophile.

Databases

To build a model capable of predicting the proteins thermal stability; at first, a dataset is created using the related databases. This dataset contains information about the structure and sequence of thermophilic and mesophilic proteins. Table 3 describes a few databases that have been used in studies of protein’s thermal stability prediction. According to Table 3, PGT and ProTherm DBs are specifically used to predict the thermal stability. PDB database is used to extract structural information while Uniport gives the sequential information of thermophilic and mesophilic proteins.

Data bases

Note

Ref. Num

General Databases

UniProt

The Universal Protein Resource (UniProt) provides a stable, comprehensive, freely accessible, central resource on protein sequences and functional annotation. This DB is used to extract the sequential information of thermophilic and mesophilic proteins.
Availability: http://www.uniprot.org.

[18]

PDB

The Protein Data Bank contains information of the 3D structures of large biological molecules, including proteins and nucleic acids. This DB is used to extract structural information of thermophilic and mesophilic proteins.
Availability: http://www.rcsb.org.

[19]

Specific Databases

Pro Therm

ProTherm is a thermodynamic database that contains experimentally determined thermodynamic parameters of protein stability. This DB is specifically used to predict the thermal stability.
Availability: http://gibk26.bse.kyutech.ac.jp/jouhou/Protherm/protherm.htm

[20]

PGT

PGT contains Prokaryotic Growth Temperature database (PGTdb). This DB is specifically used to predict the thermal stability.
Availability: http://pgtdb.csie.ncu.edu.tw

[2]

Table 3: List of databases in protein thermostability prediction.

Conclusion

Due to the recent pervasive use of thermostable proteins and enzymes in industry, protein engineering and other theoretical/experimental studies play a significant role in identification of protein thermal stability. Regarding the high expense rate of laboratory procedures, the employment of theoretical methods for predicting the thermal stability with high accuracy could be so helpful. So far, most computational thermophilic and mesophilic protein identification studies have been solely based on the protein sequence. Regarding the fact that both structural and sequential features affect the thermal stability, applying the both mentioned features at the same time leads to a more accurate, precise prediction.

References

  1. Zuo YC, Chen W, Fan GL, Li QJ (2013) A similarity distance of diversity measure for discriminating mesophilic and thermophilic proteins. Amino acids 44(2): 573-580.
  2. Huang SL, Wu LC, Liang HK, Pan KT, Horng JT, et al. (2004) PGTdb: a database providing growth temperatures of prokaryotes. Bioinformatics 20(2): 276-278.
  3. Zhou XX, Wang YB, Pan YJ, Li WF (2008) Differences in amino acids composition and coupling patterns between mesophilic and thermophilic proteins. Amino acids 34(1): 25-33.
  4. Vieille C, Zeikus GJ (2001) Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol and Mol Biol Rev 65(1): 1-43.
  5. Zhang G, Fang B (2006) Discrimination of thermophilic and mesophilic proteins via pattern recognition methods. Process Biochemistry 41(3): 552-556.
  6. Zhang G, Fang B (2006) Application of amino acid distribution along the sequence for discriminating mesophilic and thermophilic proteins. Process biochemistry 41(8): 1792-1798.
  7. Jahandideh S, Abdolmaleki P, Jahandideh M, Barzegari AE (2007) Sequence and structural parameters enhancing adaptation of proteins to low temperatures. J Theor Biol 246(1): 159-166.
  8. Gromiha MM, Suresh MX (2008) Discrimination of mesophilic and thermophilic proteins using machine learning algorithms. Proteins 70(4): 1274-1279.
  9. Kumar S, Tsai CJ, Nussinov R (2001) Thermodynamic differences among homologous thermophilic and mesophilic proteins. Biochem 40(47): 14152-14165.
  10. Zhang G, Fang B (2006) Support vector machine for discrimination of thermophilic and mesophilic proteins based on amino acid composition. Prot pept lett 13(10): 965-970.
  11. Wu LC, Lee JX, Haung HD, Liu BJ, Horng JT (2009) An expert system to predict protein thermostability using decision tree. ESWA 36(5): 9007-9014.
  12. Zhang G, Fang B (2007) LogitBoost classifier for discriminating thermophilic and mesophilic proteins. J biotechnol 127(3): 417-424.
  13. Lin H, Chen W (2011) Prediction of thermophilic proteins using feature selection technique. J Microbiol Methods 84(1): 67-70.
  14. Ku T, Lu P, Chan C, Wang T, Lai S, et al. (2009) Predicting melting temperature directly from protein sequences. Comput biol chem 33(6): 445-450.
  15. Wang L, Li C (2014) Optimal subset selection of primary sequence features using the genetic algorithm for thermophilic proteins identification. Biotechnol lett 36(10): 1963-1969.
  16. Pucci F, Dhanani M, Dehouck Y, Rooman M (2014) Protein thermostability prediction within homologous families using temperature-dependent statistical potentials. PloS One 9(3): 91659.
  17. Si J, Zhao R, Wu R (2015) An Overview of the Prediction of Protein DNA-Binding Sites. Int J Mol Sci 16(3): 5194-5215.
  18. Consortium U (2007) The universal protein resource (UniProt). Nucleic Acids Res 35: D193-D197.
  19. Berman HM, Westbrook J, Feng Z, Gilliland Z, Bhat TN, et al. (2000) The protein data bank. Nucleic Acids Res 28(1): 235-242.
  20. Kumar MD, Bava KA, Gromiha MM, Prabakaran P, Kitajima K, et al. (2006) ProTherm and ProNIT: thermodynamic databases for proteins and protein–nucleic acid interactions. Nucleic Acids Res 34(1): 204-206.
© 2014-2016 MedCrave Group, All rights reserved. No part of this content may be reproduced or transmitted in any form or by any means as per the standard guidelines of fair use.
Creative Commons License Open Access by MedCrave Group is licensed under a Creative Commons Attribution 4.0 International License.
Based on a work at http://medcraveonline.com
Best viewed in Mozilla Firefox | Google Chrome | Above IE 7.0 version | Opera |Privacy Policy