Classifying Multifunctional Enzymes by Incorporating Three Different Models into Chou’s General Pseudo Amino Acid Composition

Abstract

With the avalanche of the newly found protein sequences in the post-genomic epoch, there is an increasing trend for annotating a number of newly discovered enzyme sequences. Among the various proteins, enzyme was considered as the one of the largest kind of proteins. It takes part in most of the biochemical reactions and plays a key role in metabolic pathways. Multifunctional enzyme is enzyme that plays multiple physiological roles. Given a multifunctional enzyme sequence, how can we identify its class? Especially, how can we deal with the multi-classes problem since an enzyme may simultaneously belong to two or more functional classes? To address these problems, which are obviously very important both to basic research and drug development, a multi-label classifier was developed via three different prediction models with multi-label K-nearest algorithm. Experimental results obtained on a stringent benchmark dataset of enzymes by jackknife cross-validation test show that the predicting results were exciting, indicating that the current method could be an effective and promising high throughput method in the enzyme research. We hope it could play an important complementary role to the existing predictors in identifying the classes of enzymes.

DEEPre: sequence-based enzyme EC number prediction by deep learning

“…Although this category of methods is widely used in practice, they are unable to make a prediction when encountering a sequence without significant homologies in the current databases. Thirdly, extracting features from the sequence and classifying the enzyme using machine learning algorithms is the most extensively studied direction ( Cai et al , 2003 , 2004, 2005 ; Cai and Chou, 2005 ; Chou, 2005 ; Chou and Elrod, 2003 ; De Ferrari et al , 2012 ; Huang et al , 2007 ; Kumar and Choudhary, 2012 ; Lee et al , 2008 ; Li et al , 2016 ; Lu et al , 2007 ; Nasibov and Kandemir-Cavas, 2009 ; Qiu et al , 2009 , 2010 ; Sharif et al , 2015 ; Shen and Chou, 2007 ; Volpato et al , 2013 ; Wang et al , 2010 , 2011 ; Zhou et al , 2007 ; Zou and Xiao, 2016 ). Although this direction has already been studied for over 15 years with a number of softwares and servers available, few of them combine the procedure of feature extraction and classification optimization together.…”

MF-EFP: Predicting Multi-Functional Enzymes Function Using Improved Hybrid Multi-Label Classifier

“…How can we use its sequence information to predict which function(s) the enzyme P belongs to? In recent, many methods for predicting various protein attributes were based on the split Amino acid composition (SAAC) discrete model [18], [20] because the SAAC avoids completely losing the sequenceorder information. Afridi and Lee have developed SAACbased method and genetic ensemble classifier to predict mitochondrial achieved reasonable accuracy [21].…”

“…With regard to multifunctional enzyme prediction, Zou et al [7] proposed two feature models to make predictions and obtained 99.54% and 98.73% accuracy by using 20-D and 188-D features, respectively; however, dataset redundancy was not mentioned in the paper. Subsequently, Zou and Xiao [18] and Che et al [10] predict multifunctional enzyme in the case of taking redundancy into account. Zou used three feature extraction algorithms to compare results, the best one is SAAC with 90.57% accuracy.…”

mlDEEPre: Multi-Functional Enzyme Function Prediction With Hierarchical Multi-Label Deep Learning

“…Despite the success of those methods in predicting mono-functional enzyme function with very high accuracy, seldom have people worked on the prediction of multi-functional enzyme function, which actually constitutes a relatively large part of all the enzymes. Until now, to our knowledge, only five methods (De Ferrari et al, 2012; Zou et al, 2013; Che et al, 2016; Zou and Xiao, 2016; Amidi et al, 2017) are able to address that specific type of enzymes. Among them, De Ferrari et al (2012) use InterPro signatures as the features and multi-label k-nearest neighbor (KNN) as the algorithm.…”

“…Che et al (2016) also utilized features extracted from PSSM, combined with the multi-label KNN algorithm. Zou and Xiao (2016) deployed three variants of the famous feature, Pseudo Amino Acid Composition (PseAAC), and the multi-label KNN algorithm. Amidi et al (2017) used the predicted structure information, combined with the sequence information, as the feature, and multi-label KNN and multi-label support vector machine (SVM) as the classifier.…”