A deep learning model based on sparse auto-encoder for prioritizing cancer-related genes and drug target combinations.

Prioritization of cancer-related genes from gene expression profiles and proteomic data is vital to improve the targeted therapies research. Although computational approaches have been complementing high-throughput biological experiments on the understanding of human diseases, it still remains a big challenge to accurately discover cancer-related proteins/genes via automatic learning from large-scale protein/gene expression data and protein-protein interaction data. Most of the existing methods are based on network construction combined with gene expression profiles, which ignore the diversity between normal samples and disease cell lines. In this study, we introduced a deep learning model based on a sparse auto-encoder to learn the specific characteristics of protein interactions in cancer cell lines integrated with protein expression data. The model showed learning ability to identify cancer-related proteins/genes from the input of different protein expression profiles by extracting the characteristics of protein interaction information, which could also predict cancer-related protein combinations. Comparing with other reported methods including differential expression and network-based methods, our model got the highest area under the curve value (>0.8) in predicting cancer-related genes. Our study prioritized ~500 high-confidence cancer-related genes; among these genes, 211 already known cancer drug targets were found, which supported the accuracy of our method. The above results indicated that the proposed auto-encoder model could computationally prioritize candidate proteins/genes involved in cancer and improve the targeted therapies research.

Prediction of Protein-Protein Interactions by Evidence Combining Methods.

Most cellular functions involve proteins' features based on their physical interactions with other partner proteins. Sketching a map of protein-protein interactions (PPIs) is therefore an important inception step towards understanding the basics of cell functions. Several experimental techniques operating in vivo or in vitro have made significant contributions to screening a large number of protein interaction partners, especially high-throughput experimental methods. However, computational approaches for PPI predication supported by rapid accumulation of data generated from experimental techniques, 3D structure definitions, and genome sequencing have boosted the map sketching of PPIs. In this review, we shed light on in silico PPI prediction methods that integrate evidence from multiple sources, including evolutionary relationship, function annotation, sequence/structure features, network topology and text mining. These methods are developed for integration of multi-dimensional evidence, for designing the strategies to predict novel interactions, and for making the results consistent with the increase of prediction coverage and accuracy.

Construction and analysis of a genome-scale metabolic network for Bacillus licheniformis WX-02.

We constructed the genome-scale metabolic network of Bacillus licheniformis (B. licheniformis) WX-02 by combining genomic annotation, high-throughput phenotype microarray (PM) experiments and literature-based metabolic information. The accuracy of the metabolic network was assessed by an OmniLog PM experiment. The final metabolic model iWX1009 contains 1009 genes, 1141 metabolites and 1762 reactions, and the predicted metabolic phenotypes showed an agreement rate of 76.8% with experimental PM data. In addition, key metabolic features such as growth yield, utilization of different substrates and essential genes were identified by flux balance analysis. A total of 195 essential genes were predicted from LB medium, among which 149 were verified with the experimental essential gene set of B. subtilis 168. With the removal of 5 reactions from the network, pathways for poly-γ-glutamic acid (γ-PGA) synthesis were optimized and the γ-PGA yield reached 83.8 mmol/h. Furthermore, the important metabolites and pathways related to γ-PGA synthesis and bacterium growth were comprehensively analyzed. The present study provides valuable clues for exploring the metabolisms and metabolic regulation of γ-PGA synthesis in B. licheniformis WX-02.

Construction of a genome-scale metabolic network of the plant pathogen Pectobacterium carotovorum provides new strategies for bactericide discovery.

We reconstructed the first genome-scale metabolic network of the plant pathogen Pectobacterium carotovorum subsp. carotovorum PC1 based on its genomic sequence, annotation, and physiological data. Metabolic characteristics were analyzed using flux balance analysis (FBA), and the results were afterwards validated by phenotype microarray (PM) experiments. The reconstructed genome-scale metabolic model, iPC1209, contains 2235 reactions, 1113 metabolites and 1209 genes. We identified 19 potential bactericide targets through a comprehensive in silico gene-deletion study. Next, we performed virtual screening to identify candidate inhibitors for an important potential drug target, alkaline phosphatase, and experimentally verified that three lead compounds were able to inhibit both bacterial cell viability and the activity of alkaline phosphatase in vitro. This study illustrates a new strategy for the discovery of agricultural bactericides.

Prediction and functional analysis of the sweet orange protein-protein interaction network.

BACKGROUND: Sweet orange (Citrus sinensis) is one of the most important fruits world-wide. Because it is a woody plant with a long growth cycle, genetic studies of sweet orange are lagging behind those of other species. RESULTS: In this analysis, we employed ortholog identification and domain combination methods to predict the protein-protein interaction (PPI) network for sweet orange. The K-nearest neighbors (KNN) classification method was used to verify and filter the network. The final predicted PPI network, CitrusNet, contained 8,195 proteins with 124,491 interactions. The quality of CitrusNet was evaluated using gene ontology (GO) and Mapman annotations, which confirmed the reliability of the network. In addition, we calculated the expression difference of interacting genes (EDI) in CitrusNet using RNA-seq data from four sweet orange tissues, and also analyzed the EDI distribution and variation in different sub-networks. CONCLUSIONS: Gene expression in CitrusNet has significant modular features. Target of rapamycin (TOR) protein served as the central node of the hormone-signaling sub-network. All evidence supported the idea that TOR can integrate various hormone signals and affect plant growth. CitrusNet provides valuable resources for the study of biological functions in sweet orange.

Citrus sinensis annotation project (CAP): a comprehensive database for sweet orange genome.

Citrus is one of the most important and widely grown fruit crop with global production ranking firstly among all the fruit crops in the world. Sweet orange accounts for more than half of the Citrus production both in fresh fruit and processed juice. We have sequenced the draft genome of a double-haploid sweet orange (C. sinensis cv. Valencia), and constructed the Citrus sinensis annotation project (CAP) to store and visualize the sequenced genomic and transcriptome data. CAP provides GBrowse-based organization of sweet orange genomic data, which integrates ab initio gene prediction, EST, RNA-seq and RNA-paired end tag (RNA-PET) evidence-based gene annotation. Furthermore, we provide a user-friendly web interface to show the predicted protein-protein interactions (PPIs) and metabolic pathways in sweet orange. CAP provides comprehensive information beneficial to the researchers of sweet orange and other woody plants, which is freely available at http://citrus.hzau.edu.cn/.

Prediction and characterization of protein-protein interaction network in Xanthomonas oryzae pv. oryzae PXO99 A.

Xanthomonas oryzae pv. oryzae (Xoo), the causal agent of bacterial blight disease in rice, is one of the most serious plant pathogens worldwide. In the current analysis, we constructed a protein-protein interaction network of Xoo strain PXO99(A) with two computational approaches (interolog method and domain combination method), and verified by K-Nearest Neighbors classification method. The predicted PPI network of Xoo PXO99(A) contains 36,886 interactions among 1988 proteins. KNN verification and GO annotation confirm the reliability of the network. Detailed analysis of flagellar synthesis and chemotaxis system shows that σ factors (especially σ(28), σ(54)) in Xoo PXO99(A) are very important for flagellar synthesis and motility, and transcription factors RpoA, RpoB and RpoC are hubs to connect most σ factors. Furthermore, Xoo PXO99(A) may have both cAMP and c-di-GMP signal transduction system, and the latter is especially important for this plant pathogen. This study therefore provides valuable clues to explore the pathogenicity and metabolic regulation of Xoo PXO99(A).

The draft genome of sweet orange (Citrus sinensis).

Oranges are an important nutritional source for human health and have immense economic value. Here we present a comprehensive analysis of the draft genome of sweet orange (Citrus sinensis). The assembled sequence covers 87.3% of the estimated orange genome, which is relatively compact, as 20% is composed of repetitive elements. We predicted 29,445 protein-coding genes, half of which are in the heterozygous state. With additional sequencing of two more citrus species and comparative analyses of seven citrus genomes, we present evidence to suggest that sweet orange originated from a backcross hybrid between pummelo and mandarin. Focused analysis on genes involved in vitamin C metabolism showed that GalUR, encoding the rate-limiting enzyme of the galacturonate pathway, is significantly upregulated in orange fruit, and the recent expansion of this gene family may provide a genomic basis. This draft genome represents a valuable resource for understanding and improving many important citrus traits in the future.

DIGA--a database of improved gene annotation for phytopathogens.

BACKGROUND: Bacterial plant pathogens are very harmful to their host plants, which can cause devastating agricultural losses in the world. With the development of microbial genome sequencing, many strains of phytopathogens have been sequenced. However, some misannotations exist in these phytopathogen genomes. Our objective is to improve these annotations and store them in a central database DIGAP. DESCRIPTION: DIGAP includes the following improved information on phytopathogen genomes. (i) All the 'hypothetical proteins' were checked, and non-coding ORFs recognized by the Z curve method were removed. (ii) The translation initiation sites (TISs) of 20% approximately 25% of all the protein-coding genes have been corrected based on the NCBI RefSeq, ProTISA database and an ab initio program, GS-Finder. (iii) Potential functions of about 10% 'hypothetical proteins' have been predicted using sequence alignment tools. (iv) Two theoretical gene expression indices, the codon adaptation index (CAI) and the E(g) index, were calculated to predict the gene expression levels. (v) Potential agricultural bactericide targets and their homology-modeled 3D structures are provided in the database, which is of significance for agricultural antibiotic discovery. CONCLUSION: The results in DIGAP provide useful information for understanding the pathogenetic mechanisms of phytopathogens and for finding agricultural bactericides. DIGAP is freely available at http://ibi.hzau.edu.cn/digap/.