Mini-Reviews in Organic Chemistry - Volume 12, Issue 6, 2015

Protein translational modification is an important regulating mechanism in the cellular activities and is also responsible for many human complex diseases, such as cancer. It is a foundation in the post-translational modification proteomics to accurately identify modified proteins or sites. A number of computational techniques have recently been developed to identify modified proteins or sites and greatly speed disclosure of hidden regulations in the cell. Here, we review in silico identifications for phosphorylation, acetylation, ubiquitination, methylation, glycosylation, S-nitrosylation, sumoylation, pupylation and palmitoylation. In addition, we summarize protein translational modification databases and discuss future direction of in silico identifications of modified proteins or sites.

Prediction of protein subcellular location is a meaningful task which attracts much attention in recent years. Particularly, the number of new protein sequences yielded by the high-throughput sequencing technology in the post genomic era has increased explosively. Facing such an avalanche of new protein sequences, it is both challenging and indispensable to develop an automated method for fast and accurately annotating the subcellular attributes of uncharacterized proteins. In fact, many efforts have been undertaken to predict the protein subcellular locations in silico during the last two decades. According to the recent studies, we found that there are different forms of PseAAC models for the feature representation of proteins. Based on evolutionary information and gene ontology, many researchers expanded them into different feature representation which is one of the key contents in this review. Another important content is classifier algorithms, and prediction algorithms of multiple sites are emerging. This review will discuss the key steps of protein subcellular location.

Protein folding, prediction of protein structure and functions are most important problems in bioinformatics. The protein fold process mainly reflects in the kinetic order of folding. Predicting the structural classes of low-homology protein is a difficult problem in the prediction of protein structure. In order to understanding the mechanism of programmed cell death, it is very necessary to obtain the information about subcellular locations and functions of apoptosis proteins. Predicting protein subnuclear localizations is a challenging problem which is harder than predicting protein subcellular locations. Predicting membrane protein types is related to the structure and function of proteins. In this review, we introduce some applications of nonlinear science methods and support vector machine methods to the above protein problems. The nonlinear science methods including the horizontal visibility network, kernel method, recurrence quantification analysis, global descriptor, Lempel-Ziv complexity, and Hilbert-Huang transform are used to extract features in these approaches.

Metabolic pathway analysis has become significant for evaluating intrinsic network characteristics in biochemical reaction network reconstruction. Current applications of metabolic pathway analysis involve identifying the enzyme for the desired production, identifying pathways of optimal production, determining non-redundant pathways for drug design, and genome comparisons by alignment of pathways for missing genes identification. With the expanded application of bioinformatics, more organized methods have been introduced to examine the overall metabolic networks and network reconstruction based on genomic data. There are several in silico approaches for analysing metabolic pathways, including elementary mode analysis and extreme pathway under pathway topology analysis, flux balance analysis and metabolic flux analysis under analysis of metabolic fluxes, and metabolic control analysis. In this paper, elementary mode analysis, flux balance analysis, metabolic flux analysis and metabolic control analysis are reviewed, together with their application in metabolic network reconstruction and biological production enhancement in biological organisms. Next, a comparison of strengths and weaknesses between each of the metabolic pathway analysis methods is presented in this paper.

The comparison of DNA sequences is the basic topic in computational biology and bioinformatics, helping in speculation about their previously ambiguous structure, function, and evolution relationship. In this article, we provide a novel DNA sequence comparison scheme by constructing feature vectors based on Markov chain and information entropy. A new measure, which is calculated as the entropy of K-string’s four one-step transition probabilities, is used to compose the feature vector to characterize DNA sequence. At the same time, we provide a novel concept to address the computation burden caused by the exponential growth of computation complexity when K grows in a traditional K-string model, which is named K-string list. The proposed scheme allows us to conduct similarity research and phylogenetic analysis on two real datasets, the first exon of 11 species’

A novel numerical characterization for graphical representations of DNA sequences is proposed, which consists of two steps: construction of a novel mathematical object that only takes the first row of the traditional mathematical object and then analysis of the similarities/dissimilarities using the distance between the objects of DNA sequences. Our novel mathematical object is only a row vector and calculation of the matrix invariant is avoided, it can significantly lessen the computation. We demonstrate that this novel method can extract enough information of DNA sequences by analyzing the probabilities of six cases. The contrast experiments show that our method has similar results to the traditional methods.