Skip to main content

Predicting the phenotypic effects of non-synonymous single nucleotide polymorphisms based on support vector machines



Human genetic variations primarily result from single nucleotide polymorphisms (SNPs) that occur approximately every 1000 bases in the overall human population. The non-synonymous SNPs (nsSNPs) that lead to amino acid changes in the protein product may account for nearly half of the known genetic variations linked to inherited human diseases. One of the key problems of medical genetics today is to identify nsSNPs that underlie disease-related phenotypes in humans. As such, the development of computational tools that can identify such nsSNPs would enhance our understanding of genetic diseases and help predict the disease.


We propose a method, named Parepro (P redicting the a mino acid re placement pro bability), to identify nsSNPs having either deleterious or neutral effects on the resulting protein function. Two independent datasets, HumVar and NewHumVar, taken from the PhD-SNP server, were applied to train the model and test the robustness of Parepro. Using a 20-fold cross validation test on the HumVar dataset, Parepro achieved a Matthews correlation coefficient (MCC) of 50% and an overall accuracy (Q2) of 76%, both of which were higher than those predicted by the methods, such as PolyPhen, SIFT, and HydridMeth. Further analysis on an additional dataset (NewHumVar) using Parepro yielded similar results.


The performance of Parepro indicates that it is a powerful tool for predicting the effect of nsSNPs on protein function and would be useful for large-scale analysis of genomic nsSNP data.


Almost 90% of human genetic variations result from single nucleotide polymorphisms (SNPs) [1]. Among SNPs resulting in amino acid changes, non-synonymous SNPs (nsSNPs) are an important source of individual variation and can result in inherited diseases and drug sensitivity [24]. Therefore, the identification of nsSNPs that affect protein function and relate to disease will be a challenge in the coming years [3, 58].

A variety of methods have been developed to identify whether an nsSNP is detrimental to protein function in vitro. Most of these methods utilize evolutionary data [3, 817], protein structure information [2, 18, 19], or both [2, 7, 2022]. Ng and Henikoff [8, 16, 23] developed the software SIFT (Sorting Intolerant from Tolerant) to predict the effect of nsSNPs on protein function; SIFT is based on sequence conservation and scores from position-specific scoring matrices. Some studies [2426] have used phylogenetics to identify functionally critical residues within a protein. The MAPP (M ultivariate A nalysis of P rotein P olymorphism) [18] software exploits the physicochemical variation between wild-type amino acid residues and newly introduced residues to identify nsSNPs that impair protein function. The method Align-GVGD [9] uses both genetic biochemical variation and genetic distance between the wild-type residue and newly introduced residue to predict the effects of an nsSNP. Some methods [2, 2022] take advantage of three-dimensional structural information to analyze the impact of amino acid changes on protein function. Wang and Moult [4] found that the vast majority of nsSNPs that are related to diseases affect protein stability rather than function. Specific factors that determine stability of a protein were then used to predict the effects of nsSNPs. Chen et al. [27] used solvent accessibility of residues to predict deleterious mutations.

Support vector machine (SVM) has gained popularity over other machine learning methods for interpreting biological data [2835] because of their ability to very effectively handle noise and large datasets/input spaces [36, 37]. Then, some methods [2, 7, 10, 21] have been designed based on the SVM [38] to predict the effect of nsSNPs. Capriotti et al. [10] developed a method that depends only on the evolutionary information around the nsSNP. Peng Yue and John Moult [2] also proposed a method that uses the conservation and type of residues observed at a base change position within a protein family. Karchin et al. [7] and Bao et al. [21] introduced two methods based on structural and evolutionary information. The structural information mainly concerns areas in the protein that are buried, as well as the fraction polar secondary structure, solvent accessibility, z-score and buried charge. The evolutionary information mainly uses Hidden Markov model PHC score, Hidden Markov model relative entropy, SIFT score and the biochemical difference between the wild-type residue and newly introduced residue.

Here, we propose a method that predicts nsSNPs based on the SVM [38]. This method, named Parepro (P redicting the a mino acid re placement pro bability) uses evolutionary information surrounding an nsSNP. In addition, properties from the AAindex [39, 40] and from evolutionary information are combined to determine the dissimilarity between the wild-type and newly introduced residues. Parepro predicted the total number of nsSNPs with higher accuracy than other methods and was not dependent on structural information. In this study, two independent datasets, HumVar and NewHumVar, taken from the PhD-SNP server [10], were applied to train the model and test the robustness of Parepro, respectively.


The nsSNP prediction performance of Parepro

Figure 1 presents a flowchart illustrating the procedure of Parepro. Homologous sequences of a protein containing the target nsSNPs were selected from the Swiss-Prot database, aligned, and weighted. Position-specific amino acid probabilities (PSAP) of the amino acids surrounding mutation position were then estimated. Next, three attribute sets, namely residue differences (RD), mutation position information (MI), and information on the environment around the mutation position (IE) were constructed and combined. In this study, the attribute set IE was calculated from the six residues on either side of the mutation, because this was the smallest number of residues that produced accurate results. To evaluate the performance of different attribute sets, a 20-fold cross-validation test on the HumVar dataset was carried out. All variants in the HumVar dataset could be predicted by using different attribute sets.

Figure 1

Brief flow chart illustrating the prediction procedure of Parepro. First, the position-specific amino acid probabilities (PSAP) of the target sequence are calculated. Second, three attribute sets are constructed using the PSAP information in combination with the RD, MI, and IE properties of the amino acids. Finally, the complex vector of Parepro is integrated and used to predict the effect of an nsSNP.

Table 1 shows the performance of the three attribute sets when applied individually or in various combinations. The prediction performance of attribute set IE was the poorest among the three. By comparison, the performance of the other attribute sets (RD or MI) was high. Nevertheless, association of the attribute set RD or MI with IE improved performance such that the overall accuracy (Q2) and Matthews correlation coefficient (MCC) were approximately 75%, respectively. The highest prediction accuracy was obtained, however, after these three attribute sets were combined into a new vector, Parepro, suggesting that the three attribute sets reinforce each other in the analysis.

Table 1 The prediction performance of the Parepro attribute sets when applied alone or in combination

Effect of the number of homologous sequences on Parepro performance

To examine how the number of homologous sequences influenced the performance of Parepro, the HumVar dataset was split into seven sub datasets (i.e., F1, F2, F3, F4, F5, F6, F7) according to the number of homologous sequences, as summarized in Table 2. Then 20-fold cross-validation test was carried out on every sub datasets. Importantly, caution was taken to ensure that every test protein that contained the corresponding nsSNP was not included in the training set. As shown in Figure 2, the overall accuracy and MCC on sub dataset F1 were only about 70% and 36%, respectively. This result indicated that the prediction on the two classes (disease-related mutations and neutral polymorphisms) using sub dataset F1 was imbalance and only the major class obtained the better score. However, Parepro obtained the highest accuracy on sub dataset F3, which the overall accuracy (Q2) was 77% and the MCC was 54%. Therefore, these results indicated that the efficacy of Parepro for predicting amino acid variants depends on the number of homologous sequences.

Figure 2

The overall accuracy (Q2) and Matthews' correlation coefficient (MCC) of Parepro when testing the subsets from F1 to F7. The x-axis denotes the different test subsets from F1 to F7, and the y-axis denotes the overall accuracy (Q2) or Matthews correlation coefficient (MCC).

Table 2 Range of the number of homologous sequences

Reliability index of Parepro for nsSNP prediction

When machine learning approaches are selected to predict the effects of nsSNPs on protein function, it is important to know the reliability of the predicted result [10, 41]. In this study, a Reliability Index (RI) was also assigned to a predicted nsSNP based on the output of support vector machines that LIBSVM was used in this work. Consider that an output of LIBSVM with parameter of "-b 1" for a nsSNP is O; the RI value is thus computed as: RI = INTEGER(20 × abs(O - 0.5))+1. The RI assignment yields information about the certainty of the classification decision and thus can be used as an indicator of prediction certainty for a particular variant. Figure 3 shows the expected prediction accuracy and the proportion of the sequences with a given RI value. For example, about 66% of all nsSNPs had an RI ≥ 6, and of these nsSNPs about 88% were correctly predicted. These results are based on the HumVar dataset.

Figure 3

Average prediction accuracy calculated cumulatively with RI above a given value. For example, about 66% of all nsSNPs have RI ≥ 6, and of these nsNSPs about 88% are corrctly predicted. The result is based on the NumVar dataset.

Comparison of Parepro with other methods

We compared Parepro with other predictors, HybridMeth [10], PolyPhen [3] and SIFT [8, 16, 23]. HybridMeth uses the profile and sequence information surrounding a mutation. PolyPhen [3] is based on a decision tree and takes into account several pieces of information derived by structural parameters, functional annotations, and evolutionary information. SIFT [8, 16, 23] mainly uses information from homologous sequences.

As shown in Table 3, Parepro obtained the highest scores with respect to sensitivity, specificity, overall accuracy (Q2) and Matthews correlation coefficient (MCC) (the definition of these parameters could be find in method section) among the four methods. Because there was an obvious disparity in the number of disease-related mutations and the neutral polymorphisms in the dataset, MCC combined both the sensitivity and the specificity of the predictor and should be selected as the main score among the six scores in the evaluation [20, 21, 41, 42]. The MCC for Parepro was higher by 6%, 17% and 4% compared with the MCC obtained with PolyPhen, SIFT and HydridMeth, respectively. Furthermore, Parepro could predict all mutations in the HumVar dataset. By contrast, PolyPhen and SIFT could only predict approximately 93% of the amino acid mutations, because these programs require more specific functional or evolutionary information. These results indicate that Parepro is a powerful tool for predicting the effect of mutations.

Table 3 Comparison of performance between Parepro and other methods using the HumVar dataset

Predicted efficacy of Parepro on the NewHumVar dataset

To test the robustness of Parepro and compare it with other methods available on the web, the dataset NewHumVar was selected, which includes only new variants submitted to the Swiss-Prot database. Variants that were the same as in the HumVar dataset were removed. As shown in Table 4, all amino acid mutations in the NewHumVar dataset were predicted by Parepro. The MCC for Parepro was significantly higher than the MCCs calculated by HybridMeth, PolyPhen, and SIFT. These results indicate that Parepro outperformed these other prediction methods.

Table 4 Comparison of performance parameters of Parepro with other methods using the NewHumVar dataset


Predicting phenotypes resulting from nsSNPs is an important aspect of post-genome biology. The present study helps advance the analysis of genetic variation and may therefore lead to a better understanding of the resulting phenotypic variations among individuals with an aim toward drug design and development [2, 7, 20, 25]. Two tests using different datasets indicated that Parepro outperformed several widely used methods.

Unlike the other methods that use the machine learning method [10, 12, 2022, 43, 44], Parepro was constructed from three attribute sets RD, MI, and IE, all of which incorporate evolutionary information. In general, if the RD between the newly introduced amino acid and the residue in the mutation position has a high value, the substitution would be considered to have a high probability of being deleterious [16, 18, 25]. At the same time, attribute sets MI and IE were used to characterize the condition at the mutation position and around the mutation position, respectively. For example, when residues surrounding a mutation were found to be conserved, the region was related to either function or structure [10, 27], and thus the mutation would be deleterious. This information reinforced the characterization provided by RD. Moreover, the results indicated that these three attribute sets complemented one another to yield a higher overall accuracy (Q2) and Matthews correlation coefficient(MCC).

The attribute vector of Parepro did not contain structural features. Thus, it is possible that some of the information directly derived from the protein structure [19] was ignored by Parepro. However, the lack of structural information was likely overcome by the inclusion of 50 discrete amino acid properties in the RD attribute set, thereby enhancing the efficacy of the sequence-based Parepro program.


We present an SVM-based prediction method, Parepro, which predicts the effect of nsSNPs on protein function. Comprehensive comparisons of the prediction performance on two datasets showed that Parepro, which utilizes information from the amino acids surrounding the mutation position and from the residue difference between the newly introduced amino acid and the average residue in the mutation position, outperformed several other widely used prediction methods. Moreover, Parepro was able to predict all mutations within two distinct test sets. Therefore, we anticipate that Parepro will be a useful tool for large-scale analysis of nsSNPs in genomic databases.


The prediction procedure of Parepro (Figure 1) begins by calculating the position-specific amino acid probabilities (PSAP) of a target protein that contains a corresponding nsSNP. Next, three attribute sets were constructed using PSAP and the properties of amino acids from AAindex [39, 40]; these three sets were then used to describe residue differences (RD) and mutation position information (MI) and to yield information on the environment around the mutation positions (IE). Finally, a complex vector that consisted of 94 attributes was used to predict the effects of the nsSNPs. The attribute sets RD, MI and IE comprised 50, 23, 21 attributes, respectively.

The mutation datasets

We used two datasets, HumVar and NewHumVar, taken from the PhD-SNP server [10]. The dataset HumVar consisted of 21,185 different SNPs (12,944 were disease-related, and 8,241 were neutral polymorphisms) obtained from 3,587 protein sequences in the Swiss-Prot database (Release 48). The NewHumVar dataset was comprised of SNPs obtained from the Swiss-Prot database (Release 50) after eliminating any variants also present in the HumVar dataset. Therefore, the dataset NewHumVar consisted of 935 single amino acid mutations (149 were disease-related variants, and 786 were neutral mutations) from 469 different proteins.

Computing position-specific amino acid probabilities (PSAPs)

A Dirichlet mixture method [4547] was adopted to estimate the PSAPs, which was then used to construct the vector of Parepro and was calculated as follows:

(1) PSI-BLAST [48] with parameter -e 0.001 was run for three iterations to collect sequences similar to the target protein that contained the corresponding nsSNP from the Swiss-Prot database (Release 50.0) [49]. The identified sequences were aligned by ClustalX [50, 51] with default parameters. The position-based sequence weight method [52] was used to derive the weight w i of the i th sequence in the alignment. If no homologous sequence was selected, the weight w i of the target sequence was designated as 1.0.

(2) An alignment column was summarized by its weighted composition into a vector c. The element of vector c was calculated as follows:

c m = i = 1 N w i × δ i m ( m = 1 , 2 21 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4yam2aaSbaaSqaaiabd2gaTbqabaGccqGH9aqpdaaeWbqaaiabdEha3naaBaaaleaacqWGPbqAaeqaaOGaey41aqlcciGae8hTdq2aaSbaaSqaaiabdMgaPjabd2gaTbqabaaabaGaemyAaKMaeyypa0JaeGymaedabaGaemOta4eaniabggHiLdGccqGGOaakcqWGTbqBcqGH9aqpcqaIXaqmcqGGSaalcqaIYaGmcqWIVlctcqaIYaGmcqaIXaqmcqGGPaqkaaa@4B28@

where N is the total number of aligned sequences, w i is the weight of the i th sequence, the value of m from 1 to 20 represents any one of 20 amino acids, and a value of 21 represents a gap. If the symbol type of the i th sequence at the column is an amino acid a m (m = 1, 220) or gap (m = 21), the value of δ im is 1.0; otherwise it is 0.

(3) A new vector u, which incorporated the gap information into the 20 amino acids, was constructed as follows:

u m = c m + c21 × h m (m = 1, 220)

where the vector h is the frequency of occurrence of any one of the 20 amino acids [53].

(4) The Dirichlet mixture method [4547] was adopted to estimate the PSAPs. The posterior probability of amino acid m at a position, p ^ m MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGafmiCaaNbaKaadaWgaaWcbaGaemyBa0gabeaaaaa@2EDB@ , was calculated from a 20-component Dirichlet mixture[47]:

p ^ m = X m k = 1 20 X k ( m = 1 , 2 20 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGafmiCaaNbaKaadaWgaaWcbaGaemyBa0gabeaakiabg2da9KqbaoaalaaabaGaemiwaG1aaSbaaeaacqWGTbqBaeqaaaqaamaaqahabaGaemiwaG1aaSbaaeaacqWGRbWAaeqaaaqaaiabdUgaRjabg2da9iabigdaXaqaaiabikdaYiabicdaWaGaeyyeIuoaaaGcdaqadaqaaiabd2gaTjabg2da9iabigdaXiabcYcaSiabikdaYiabl+UimjabikdaYiabicdaWaGaayjkaiaawMcaaaaa@4838@
X m = j = 1 l q j B ( α j + n ) B ( α j ) × α j , m + n m | a j | + | n | MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemiwaG1aaSbaaSqaaiabd2gaTbqabaGccqGH9aqpdaaeWbqaaiabdghaXnaaBaaaleaacqWGQbGAaeqaaKqbaoaalaaabaGaemOqaiKaeiikaGccciGaf8xSdeMbaSaadaWgaaqaaiabdQgaQbqabaGaey4kaSIafmOBa4MbaSaacqGGPaqkaeaacqWGcbGqcqGGOaakcuWFXoqygaWcamaaBaaabaGaemOAaOgabeaacqGGPaqkaaGccqGHxdaTjuaGdaWcaaqaaiab=f7aHnaaBaaabaGaemOAaOMaeiilaWIaemyBa0gabeaacqGHRaWkcqWGUbGBdaWgaaqaaiabd2gaTbqabaaabaWaaqWaaeaacuWGHbqygaWcamaaBaaabaGaemOAaOgabeaaaiaawEa7caGLiWoacqGHRaWkdaabdaqaaiqbd6gaUzaalaaacaGLhWUaayjcSdaaaaWcbaGaemOAaOMaeyypa0JaeGymaedabaGaemiBaWganiabggHiLdaaaa@6015@

where q j is the mixture coefficient of each component, B is the Beta function, α j MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaacciGaf8xSdeMbaSaadaWgaaWcbaGaemOAaOgabeaaaaa@2F14@ = (αj 1...αj 20) is the parameter for each component j of the Dirichlet mixture, and l is the number of components. The vector n was calculated by the equation, n m = u m × N(m = 1, 220), where N is the total number of homologous sequences and u m is calculated from equation (2).

Inputs and Encoding Schemes of Parepro

The Parepro vector was comprised of three attribute sets, which were used to describe the RD, the MI, and the IE.

The first attribute set, RD, was designed to depict the property differences between the newly introduced amino acid and the average residue in the mutation position, which was composed of 50 elements and was constructed as follows:

(1) The 544 amino acid properties were downloaded from AAindex [39, 40], as shown in Additional file 1. Then the value of each property t km (k = 1,,544, m = 1, 220) was standardized as follows:

t k m ' = t k m μ k s k MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemiDaq3aa0baaSqaaiabdUgaRjabd2gaTbqaaiabcEcaNaaakiabg2da9KqbaoaalaaabaGaemiDaq3aaSbaaeaacqWGRbWAcqWGTbqBaeqaaiabgkHiTGGaciab=X7aTnaaBaaabaGaem4AaSgabeaaaeaacqWGZbWCdaWgaaqaaiabdUgaRbqabaaaaaaa@3E72@

where μ k and s k 2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4Cam3aa0baaSqaaiabdUgaRbqaaiabikdaYaaaaaa@2FC0@ are the mean and variance of the property k, respectively, and were calculated as follows: μ k = 1 20 m = 1 20 t k m MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaacciGae8hVd02aaSbaaSqaaiabdUgaRbqabaGccqGH9aqpjuaGdaWcaaqaaiabigdaXaqaaiabikdaYiabicdaWaaakmaaqahabaGaemiDaq3aaSbaaSqaaiabdUgaRjabd2gaTbqabaaabaGaemyBa0Maeyypa0JaeGymaedabaGaeGOmaiJaeGimaadaniabggHiLdaaaa@3F72@ and s k 2 = 1 19 m = 1 20 ( t k m μ k ) 2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4Cam3aa0baaSqaaiabdUgaRbqaaiabikdaYaaakiabg2da9KqbaoaalaaabaGaeGymaedabaGaeGymaeJaeGyoaKdaaOWaaabCaeaacqGGOaakcqWG0baDdaWgaaWcbaGaem4AaSMaemyBa0gabeaakiabgkHiTGGaciab=X7aTnaaBaaaleaacqWGRbWAaeqaaOGaeiykaKYaaWbaaSqabeaacqaIYaGmaaaabaGaemyBa0Maeyypa0JaeGymaedabaGaeGOmaiJaeGimaadaniabggHiLdaaaa@4741@ .

(2) The position-dependent properties d k were given by

d k m = p m × t k m ' MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemizaq2aaSbaaSqaaiabdUgaRjabd2gaTbqabaGccqGH9aqpcqWGWbaCdaWgaaWcbaGaemyBa0gabeaakiabgEna0kabdsha0naaDaaaleaacqWGRbWAcqWGTbqBaeaacqGGNaWjaaaaaa@3BBF@

where p m is the PSAP at a mutated position calculated from equation (3).

(3) With respect to property k, the distance r k between the weighted property d kn of a newly introduced amino acid n and the mean of d k was

r k = d k n μ k ' s k ' MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemOCai3aaSbaaSqaaiabdUgaRbqabaGccqGH9aqpjuaGdaWcaaqaaiabdsgaKnaaBaaabaGaem4AaSMaemOBa4gabeaacqGHsisliiGacqWF8oqBdaqhaaqaaiabdUgaRbqaaiabcEcaNaaaaeaacqWGZbWCdaqhaaqaaiabdUgaRbqaaiabcEcaNaaaaaaaaa@3DC4@

where μ k ' MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaacciGae8hVd02aa0baaSqaaiabdUgaRbqaaiabcEcaNaaaaaa@2FF2@ and s k ' 2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4Cam3aa0baaSqaaiabdUgaRbqaaiabcEcaNiabikdaYaaaaaa@3096@ are the mean and variance of d k , respectively, and were calculated as follows: μ k ' = 1 20 m = 1 20 d k m MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaacciGae8hVd02aa0baaSqaaiabdUgaRbqaaiabcEcaNaaakiabg2da9KqbaoaalaaabaGaeGymaedabaGaeGOmaiJaeGimaadaaOWaaabCaeaacqWGKbazdaWgaaWcbaGaem4AaSMaemyBa0gabeaaaeaacqWGTbqBcqGH9aqpcqaIXaqmaeaacqaIYaGmcqaIWaama0GaeyyeIuoaaaa@4029@ , s k ' 2 = 1 19 m = 1 20 ( d k m μ k ' ) 2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaem4Cam3aa0baaSqaaiabdUgaRbqaaiabcEcaNiabikdaYaaakiabg2da9KqbaoaalaaabaGaeGymaedabaGaeGymaeJaeGyoaKdaaOWaaabCaeaacqGGOaakcqWGKbazdaWgaaWcbaGaem4AaSMaemyBa0gabeaakiabgkHiTGGaciab=X7aTnaaDaaaleaacqWGRbWAaeaacqGGNaWjaaGccqGGPaqkdaahaaWcbeqaaiabikdaYaaaaeaacqWGTbqBcqGH9aqpcqaIXaqmaeaacqaIYaGmcqaIWaama0GaeyyeIuoaaaa@48CE@ .

(4) A new vector r was then constructed using the 544 elements from Additional file 1. The software weka3.4 [54] was used to simplify the vector r, in which the evaluator CfsSubsetEvalwas selected. The redundant and low-contribution elements in vector r were removed. After these modifications, 50 elements remained and were included in the RD attribute set.

The second attribute set, MI, was used to define the status of a mutation position and consisted of 23 values. The first 20 elements were the PSAP values of the 20 amino acids in the mutation position calculated from equation (3). The 21st and 22nd elements were the PSAP values of the wild-type residue and the newly introduced residue, respectively. The 23rd value was the entropy (E) [55, 56] of amino acids in the mutation position and was calculated as follows:

E = 1 ln 20 m = 1 20 p m ln p m MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaGaemyrauKaeyypa0JaeyOeI0scfa4aaSaaaeaacqaIXaqmaeaacyGGSbaBcqGGUbGBcqaIYaGmcqaIWaamaaGcdaaeWbqaaiabdchaWnaaBaaaleaacqWGTbqBaeqaaOGagiiBaWMaeiOBa4MaemiCaa3aaSbaaSqaaiabd2gaTbqabaaabaGaemyBa0Maeyypa0JaeGymaedabaGaeGOmaiJaeGimaadaniabggHiLdaaaa@4595@

where 20 is the number of amino acids, and p m is the PSAP value at the mutation position calculated from equation (3).

The third attribute set, IE, encoded the information surrounding the mutation position and consisted of 21 elements. The first 20 elements represented the PSAP values of the 20 amino acids and were calculated from equation (3), and the last element represented entropy and was calculated from equation (8). Residues in the immediate vicinity of the mutation carried more significance with respect to the mutation. Therefore, a significance coefficient was assigned to each residue in proximity to the mutation. The element of IE was then calculated as follows:

I E a = 1 f + 1 × m = f f f + 1 a b s ( m ) f + 1 y ( i + m ) a ( a = 1 , 2 21 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=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@61D0@

where i is the mutation position, f is the number of residues located to the left or right of the mutation position, and a represents one element of IE from 1 to 21. If the value of a is between 1 and 20, y(i+m)ais p a in the position of i + m calculated from equation (3). However, if the value of a is 21, y(i+m)ais the entropy Ei+mcalculated from equation (8). Furthermore, if the mutation is located at the N-terminal position (i + m > l) or at the C-terminal position, then y(i+m)ais y la or y la , respectively, where l is the number of residues in the protein.

Support vector machine

The SVM is a classifier seeking an optimal hyperplane to separate two classes of samples. SVM uses kernel functions to map original data to a feature space of higher dimensions and locates an optimal separating hyperplane. For SVM implementation, we used LIBSVM [57] with a Radial Basis Function (RBF kernel function) K(x i , x j ) = exp(-G||x i - x j ||2). The parameter was selected with the LIBSVM parameter selection tool.

Scoring the performance

The proteins in the dataset were randomly divided into 20 subsets. For each individual test, the mutations in one of the 20 sub-datasets were used as the test set and the others in the 19 subsets were combined to form a training set. The procedure was repeated 20 times so that each sample was used exactly once for testing and training. We defined disease-associated nsSNPs as positive and neutral nsSNPs as negative. In this work, we adopted sensitivity, specificity, overall accuracy(Q2) and Matthews correlation coefficient(MCC) to score the performance of the corresponding method:

S e n c i t i v i t y = T P T P + F N , S p e c i f i c i t y = T N T N + F P , Q 2 = T P + T N T P + F P + T N + F N M C C = T P × T N F P × F N ( T N + F N ) × ( T N + F P ) × ( T P + F N ) × ( T P + F P ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=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@AA30@

where TP is the number of true positives, TN is the number of true negatives, FP is the number of false positives, and FN is the number of false negatives. Because there was an obvious disparity in the number of positive samples and negative samples in the dataset, MCC combined both the sensitivity and the specificity of the predictor and should be selected as the main score among the six scores in the evaluation [20, 21, 41, 42].

Availability and requirements

Project name: Parepro

Project home page:

Operating systems: Windows

Programming language: Perl

License: GNU General Public License. This license allows the source code to be redistributed and/or modified under the terms of the GNU General Public License as published by the Free Software Foundation. The source code for the application is available at no charge.

Any restrictions to use by non-academics: None


  1. 1.

    Collins FS, Brooks LD, Chakravarti A: A DNA polymorphism discovery resource for research on human genetic variation. Genome Res. 1998, 8 (12): 1229-1231.

    CAS  PubMed  Google Scholar 

  2. 2.

    Yue P, Moult J: Identification and analysis of deleterious human SNPs. J Mol Biol. 2006, 356 (5): 1263-1274. 10.1016/j.jmb.2005.12.025.

    CAS  Article  PubMed  Google Scholar 

  3. 3.

    Ramensky V, Bork P, Sunyaev S: Human non-synonymous SNPs: server and survey. Nucleic Acids Res. 2002, 30 (17): 3894-3900. 10.1093/nar/gkf493.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  4. 4.

    Wang Z, Moult J: SNPs, protein structure, and disease. Hum Mutat. 2001, 17 (4): 263-270. 10.1002/humu.22.

    Article  PubMed  Google Scholar 

  5. 5.

    Cooper DN, Ball EV, Krawczak M: The human gene mutation database. Nucleic Acids Res. 1998, 26 (1): 285-287. 10.1093/nar/26.1.285.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  6. 6.

    Stenson PD, Ball EV, Mort M, Phillips AD, Shiel JA, Thomas NS, Abeysinghe S, Krawczak M, Cooper DN: Human Gene Mutation Database (HGMD): 2003 update. Hum Mutat. 2003, 21 (6): 577-581. 10.1002/humu.10212.

    CAS  Article  PubMed  Google Scholar 

  7. 7.

    Karchin R, Diekhans M, Kelly L, Thomas DJ, Pieper U, Eswar N, Haussler D, Sali A: LS-SNP: large-scale annotation of coding non-synonymous SNPs based on multiple information sources. Bioinformatics. 2005, 21 (12): 2814-2820. 10.1093/bioinformatics/bti442.

    CAS  Article  PubMed  Google Scholar 

  8. 8.

    Ng PC, Henikoff S: Accounting for human polymorphisms predicted to affect protein function. Genome Res. 2002, 12 (3): 436-446. 10.1101/gr.212802.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  9. 9.

    Mathe E, Olivier M, Kato S, Ishioka C, Hainaut P, Tavtigian SV: Computational approaches for predicting the biological effect of p53 missense mutations: a comparison of three sequence analysis based methods. Nucleic Acids Res. 2006, 34 (5): 1317-1325. 10.1093/nar/gkj518.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  10. 10.

    Capriotti E, Calabrese R, Casadio R: Predicting the insurgence of human genetic diseases associated to single point protein mutations with support vector machines and evolutionary information. Bioinformatics. 2006, 22 (22): 2729-2734. 10.1093/bioinformatics/btl423.

    CAS  Article  PubMed  Google Scholar 

  11. 11.

    Ferrer-Costa C, Gelpi JL, Zamakola L, Parraga I, de la Cruz X, Orozco M: PMUT: a web-based tool for the annotation of pathological mutations on proteins. Bioinformatics. 2005, 21 (14): 3176-3178. 10.1093/bioinformatics/bti486.

    CAS  Article  PubMed  Google Scholar 

  12. 12.

    Capriotti E, Fariselli P, Calabrese R, Casadio R: Predicting protein stability changes from sequences using support vector machines. Bioinformatics. 2005, 21 (Suppl 2): ii54-58. 10.1093/bioinformatics/bti1109.

    CAS  Article  PubMed  Google Scholar 

  13. 13.

    Brunham LR, Singaraja RR, Pape TD, Kejariwal A, Thomas PD, Hayden MR: Accurate prediction of the functional significance of single nucleotide polymorphisms and mutations in the ABCA1 gene. PLoS Genet. 2005, 1 (6): e83-10.1371/journal.pgen.0010083.

    PubMed Central  Article  PubMed  Google Scholar 

  14. 14.

    Tchernitchko D, Goossens M, Wajcman H: In silico prediction of the deleterious effect of a mutation: proceed with caution in clinical genetics. Clin Chem. 2004, 50 (11): 1974-1978. 10.1373/clinchem.2004.036053.

    CAS  Article  PubMed  Google Scholar 

  15. 15.

    Thomas PD, Campbell MJ, Kejariwal A, Mi H, Karlak B, Daverman R, Diemer K, Muruganujan A, Narechania A: PANTHER: a library of protein families and subfamilies indexed by function. Genome Res. 2003, 13 (9): 2129-2141. 10.1101/gr.772403.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  16. 16.

    Ng PC, Henikoff S: SIFT: Predicting amino acid changes that affect protein function. Nucleic Acids Res. 2003, 31 (13): 3812-3814. 10.1093/nar/gkg509.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  17. 17.

    Fleming MA, Potter JD, Ramirez CJ, Ostrander GK, Ostrander EA: Understanding missense mutations in the BRCA1 gene: an evolutionary approach. Proc Natl Acad Sci USA. 2003, 100 (3): 1151-1156. 10.1073/pnas.0237285100.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  18. 18.

    Stone EA, Sidow A: Physicochemical constraint violation by missense substitutions mediates impairment of protein function and disease severity. Genome Res. 2005, 15 (7): 978-986. 10.1101/gr.3804205.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  19. 19.

    Saunders CT, Baker D: Evaluation of structural and evolutionary contributions to deleterious mutation prediction. J Mol Biol. 2002, 322 (4): 891-901. 10.1016/S0022-2836(02)00813-6.

    CAS  Article  PubMed  Google Scholar 

  20. 20.

    Dobson RJ, Munroe PB, Caulfield MJ, Saqi MA: Predicting deleterious nsSNPs: an analysis of sequence and structural attributes. BMC Bioinformatics. 2006, 7: 217-10.1186/1471-2105-7-217.

    PubMed Central  Article  PubMed  Google Scholar 

  21. 21.

    Bao L, Cui Y: Prediction of the phenotypic effects of non-synonymous single nucleotide polymorphisms using structural and evolutionary information. Bioinformatics. 2005, 21 (10): 2185-2190. 10.1093/bioinformatics/bti365.

    CAS  Article  PubMed  Google Scholar 

  22. 22.

    Krishnan VG, Westhead DR: A comparative study of machine-learning methods to predict the effects of single nucleotide polymorphisms on protein function. Bioinformatics. 2003, 19 (17): 2199-2209. 10.1093/bioinformatics/btg297.

    CAS  Article  PubMed  Google Scholar 

  23. 23.

    Ng PC, Henikoff S: Predicting deleterious amino acid substitutions. Genome Res. 2001, 11 (5): 863-874. 10.1101/gr.176601.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  24. 24.

    Armon A, Graur D, Ben-Tal N: ConSurf: an algorithmic tool for the identification of functional regions in proteins by surface mapping of phylogenetic information. J Mol Biol. 2001, 307 (1): 447-463. 10.1006/jmbi.2000.4474.

    CAS  Article  PubMed  Google Scholar 

  25. 25.

    Landau M, Mayrose I, Rosenberg Y, Glaser F, Martz E, Pupko T, Ben-Tal N: ConSurf 2005: the projection of evolutionary conservation scores of residues on protein structures. Nucleic Acids Res. 2005, 33 (Web Server): W299-302. 10.1093/nar/gki370.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  26. 26.

    Pupko T, Bell RE, Mayrose I, Glaser F, Ben-Tal N: Rate4Site: an algorithmic tool for the identification of functional regions in proteins by surface mapping of evolutionary determinants within their homologues. Bioinformatics. 2002, 18 (Suppl 1): S71-77.

    Article  PubMed  Google Scholar 

  27. 27.

    Chen H, Zhou HX: Prediction of solvent accessibility and sites of deleterious mutations from protein sequence. Nucleic Acids Res. 2005, 33 (10): 3193-3199. 10.1093/nar/gki633.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  28. 28.

    Natt NK, Kaur H, Raghava GP: Prediction of transmembrane regions of beta-barrel proteins using ANN- and SVM-based methods. Proteins. 2004, 56 (1): 11-18. 10.1002/prot.20092.

    CAS  Article  PubMed  Google Scholar 

  29. 29.

    Bhasin M, Raghava GP: ESLpred: SVM-based method for subcellular localization of eukaryotic proteins using dipeptide composition and PSI-BLAST. Nucleic Acids Res. 2004, 32 (Web Server): W414-419. 10.1093/nar/gkh350.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  30. 30.

    Byvatov E, Schneider G: Support vector machine applications in bioinformatics. Appl Bioinformatics. 2003, 2 (2): 67-77.

    PubMed  Google Scholar 

  31. 31.

    Ding CH, Dubchak I: Multi-class protein fold recognition using support vector machines and neural networks. Bioinformatics. 2001, 17 (4): 349-358. 10.1093/bioinformatics/17.4.349.

    CAS  Article  PubMed  Google Scholar 

  32. 32.

    Zien A, Ratsch G, Mika S, Scholkopf B, Lengauer T, Muller KR: Engineering support vector machine kernels that recognize translation initiation sites. Bioinformatics. 2000, 16 (9): 799-807. 10.1093/bioinformatics/16.9.799.

    CAS  Article  PubMed  Google Scholar 

  33. 33.

    Jaakkola T, Diekhans M, Haussler D: A discriminative framework for detecting remote protein homologies. J Comput Biol. 2000, 7 (1–2): 95-114. 10.1089/10665270050081405.

    CAS  Article  PubMed  Google Scholar 

  34. 34.

    Furey TS, Cristianini N, Duffy N, Bednarski DW, Schummer M, Haussler D: Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics. 2000, 16 (10): 906-914. 10.1093/bioinformatics/16.10.906.

    CAS  Article  PubMed  Google Scholar 

  35. 35.

    Brown MP, Grundy WN, Lin D, Cristianini N, Sugnet CW, Furey TS, Ares M, Haussler D: Knowledge-based analysis of microarray gene expression data by using support vector machines. Proc Natl Acad Sci USA. 2000, 97 (1): 262-267. 10.1073/pnas.97.1.262.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  36. 36.

    Idicula-Thomas S, Kulkarni AJ, Kulkarni BD, Jayaraman VK, Balaji PV: A support vector machine-based method for predicting the propensity of a protein to be soluble or to form inclusion body on overexpression in Escherichia coli. Bioinformatics. 2006, 22 (3): 278-284. 10.1093/bioinformatics/bti810.

    CAS  Article  PubMed  Google Scholar 

  37. 37.

    Zavaljevski N, Stevens FJ, Reifman J: Support vector machines with selective kernel scaling for protein classification and identification of key amino acid positions. Bioinformatics. 2002, 18 (5): 689-696. 10.1093/bioinformatics/18.5.689.

    CAS  Article  PubMed  Google Scholar 

  38. 38.

    N C: Support Vector Machines and other kernel-based learning methods. 2000, Cambridge University Press

    Google Scholar 

  39. 39.

    Kawashima S, Ogata H, Kanehisa M: AAindex: Amino Acid Index Database. Nucleic Acids Res. 1999, 27 (1): 368-369. 10.1093/nar/27.1.368.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  40. 40.

    Kawashima S, Kanehisa M: AAindex: amino acid index database. Nucleic Acids Res. 2000, 28 (1): 374-10.1093/nar/28.1.374.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  41. 41.

    Baldi P, Brunak S, Chauvin Y, Andersen CA, Nielsen H: Assessing the accuracy of prediction algorithms for classification: an overview. Bioinformatics. 2000, 16 (5): 412-424. 10.1093/bioinformatics/16.5.412.

    CAS  Article  PubMed  Google Scholar 

  42. 42.

    Matthews BW: Comparison of the predicted and observed secondary structure of T4 phage lysozyme. Biochim Biophys Acta. 1975, 405 (2): 442-451.

    CAS  Article  PubMed  Google Scholar 

  43. 43.

    Cheng J, Randall A, Baldi P: Prediction of protein stability changes for single-site mutations using support vector machines. Proteins. 2006, 62 (4): 1125-1132. 10.1002/prot.20810.

    CAS  Article  PubMed  Google Scholar 

  44. 44.

    Capriotti E, Fariselli P, Casadio R: A neural-network-based method for predicting protein stability changes upon single point mutations. Bioinformatics. 2004, 20 (Suppl 1): i63-68. 10.1093/bioinformatics/bth928.

    CAS  Article  PubMed  Google Scholar 

  45. 45.

    Brown M, Hughey R, Krogh A, Mian IS, Sjolander K, Haussler D: Using Dirichlet mixture priors to derive hidden Markov models for protein families. Proc Int Conf Intell Syst Mol Biol. 1993, 1: 47-55.

    CAS  PubMed  Google Scholar 

  46. 46.

    Lau AY, Chasman DI: Functional classification of proteins and protein variants. Proc Natl Acad Sci USA. 2004, 101 (17): 6576-6581. 10.1073/pnas.0305043101.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  47. 47.

    Sjolander K, Karplus K, Brown M, Hughey R, Krogh A, Mian IS, Haussler D: Dirichlet mixtures: a method for improved detection of weak but significant protein sequence homology. Comput Appl Biosci. 1996, 12 (4): 327-345.

    CAS  PubMed  Google Scholar 

  48. 48.

    Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997, 25 (17): 3389-3402. 10.1093/nar/25.17.3389.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  49. 49.

    Boeckmann B, Bairoch A, Apweiler R, Blatter MC, Estreicher A, Gasteiger E, Martin MJ, Michoud K, O'Donovan C, Phan I, Pilbout S, Schneider M: The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003. Nucleic Acids Res. 2003, 31 (1): 365-370. 10.1093/nar/gkg095.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  50. 50.

    Thompson JD, Higgins DG, Gibson TJ, CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 1994, 22 (22): 4673-4680. 10.1093/nar/22.22.4673.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  51. 51.

    Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG: The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 1997, 25 (24): 4876-4882. 10.1093/nar/25.24.4876.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  52. 52.

    Henikoff S, Henikoff JG: Position-based sequence weights. J Mol Biol. 1994, 243 (4): 574-578. 10.1016/0022-2836(94)90032-9.

    CAS  Article  PubMed  Google Scholar 

  53. 53.

    Jones DT, Taylor WR, Thornton JM: The rapid generation of mutation data matrices from protein sequences. Comput Appl Biosci. 1992, 8 (3): 275-282.

    CAS  PubMed  Google Scholar 

  54. 54.

    Frank E, Hall M, Trigg L, Holmes G, Witten IH: Data mining in bioinformatics using Weka. Bioinformatics. 2004, 20 (15): 2479-2481. 10.1093/bioinformatics/bth261.

    CAS  Article  PubMed  Google Scholar 

  55. 55.

    Sander C, Schneider R: Database of homology-derived protein structures and the structural meaning of sequence alignment. Proteins. 1991, 9 (1): 56-68. 10.1002/prot.340090107.

    CAS  Article  PubMed  Google Scholar 

  56. 56.

    Valdar WS: Scoring residue conservation. Proteins. 2002, 48 (2): 227-241. 10.1002/prot.10146.

    CAS  Article  PubMed  Google Scholar 

  57. 57.

    LIBSVM. []

Download references


The authors thank Dr. R. Casadio for providing the datasets, HumVar and NewHumVar. This work was supported by the National Natural Science Foundation of China (Grant no.30470031).

Author information



Corresponding author

Correspondence to Ningfeng Wu.

Additional information

Authors' contributions

Jian Tian wrote the code of Parepro. Ningfeng Wu, Juhua Zhang and Yunliu Fan supervised the work. Jian Tian, Ningfeng Wu, Xuexia Guo and Jun Guo were involved in the preparation of the manuscript. Jian Tian, Ningfeng Wu, Xuexia Guo, Jun Guo, Juhua Zhang and Yunliu Fan read and approved the manuscript.

Electronic supplementary material

Authors’ original submitted files for images

Below are the links to the authors’ original submitted files for images.

Authors’ original file for figure 1

Authors’ original file for figure 2

Authors’ original file for figure 3

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Tian, J., Wu, N., Guo, X. et al. Predicting the phenotypic effects of non-synonymous single nucleotide polymorphisms based on support vector machines. BMC Bioinformatics 8, 450 (2007).

Download citation


  • Support Vector Machine
  • Reliability Index
  • Mutation Position
  • Matthews Correlation Coefficient
  • Amino Acid Mutation