Identification of apolipoprotein using feature selection technique

Abstract

Apolipoprotein is a kind of protein which can transport the lipids through the lymphatic and circulatory systems. The abnormal expression level of apolipoprotein always causes angiocardiopathy. Thus, correct recognition of apolipoprotein from proteomic data is very crucial to the comprehension of cardiovascular system and drug design. This study is to develop a computational model to predict apolipoproteins. In the model, the apolipoproteins and non-apolipoproteins were collected to form benchmark dataset. On the basis of the dataset, we extracted the g-gap dipeptide composition information from residue sequences to formulate protein samples. To exclude redundant information or noise, the analysis of various (ANOVA)-based feature selection technique was proposed to find out the best feature subset. The support vector machine (SVM) was selected as discrimination algorithm. Results show that 96.2% of sensitivity and 99.3% of specificity were achieved in five-fold cross-validation. These findings open new perspectives to improve apolipoproteins prediction by considering the specific dipeptides. We expect that these findings will help to improve drug development in anti-angiocardiopathy disease.

A computational method for prediction of xylanase enzymes activity in strains of Bacillus subtilis based on pseudo amino acid composition features

“…The concept of PseAAC was proposed by Chou [ 25 ]. Since then, the concept of PseAAC has penetrated into almost all the fields of computational proteomics [ 26 – 30 , 58 ]. Encouraged by the successes of introducing the PseAAC approach into computational proteomics, a novel feature vector, called ‘pseudo K-tuple nucleotide composition’(PseKNC) [ 31 , 32 ], was developed to represent DNA sequence samples to improve the quality of predicting the elements [ 33 – 37 , 39 , 40 , 57 , 66 ].…”

“…The main reason for choosing PseAAC feature vectors as representative of xylanase enzymes in activity prediction task was the fact that PseAAC features have been vastly used in computational biology for prediction of different properties of proteins and nucleic acid sequences since 2001 [ 24 – 58 ]. Some of its recent applications are related to RNA and DNA sequence analysis fields.…”

predSucc-Site: Lysine Succinylation Sites Prediction in Proteins by using Support Vector Machine and Resolving Data Imbalance Issue

“…Since the information about the succinylation sites is mostly desired than non-succinylation sites [5,8] from the biological point of view, predSucc-Site will be more preferable than such type of problematic predictors [8]. Moreover, it can be noted here that, the programs such as BLAST and FASTA have been widely used in genomic and proteomic analysis or prediction based on similarity search [36]. But, unfortunately these programs are helpless while facing sequences having low similarities, especially, when more and more orphan genes are discovered.…”

“…Therefore, in the case of orphan genes or low-similar proteins, it is urgent to develop a statistical predictive model. From this biological viewpoint, this study is very meaningful and important [36]. However, in order to make a more consistent comparison of our predSucc-Site with SucPred, pSuc-Lys and iSuc-PseOpt in the case of specificity metric, two thresholds of specificity 95% and 97% have been taken into consideration and values of other metrics of predSucc-Site at those level of specificities have been calculated too.…”

Recent Advances in Conotoxin Classification by Using Machine Learning Methods

“…In order to select optimal features from the 400D dipeptide compositions, Wang et al classified the ion channel-targeted conotoxins with the analysis of variance (ANOVA) method [ 45 ]. The variance-based analysis is used to calculate the ratio of the variance among groups and the variance within the group for each attribute [ 72 , 73 ]. It has a good foundation of statistics and can test the feature difference between groups intuitively.…”