Published papers

A new mathematical index was developed to identify and characterize non-coding RNA (ncRNA) genes encoded within the Escherichia coli (E. coli) genome. It was designated the GMMI (Gapped Markov Model Index) and used to evaluate sequence patterns located at the separate positions of consensus sequences, codon biases and/or possible RNA structures on the basis of the Markov model. The GMMI was able to separate a set of known mRNA sequences from a mixture of ncRNAs including tRNAs and rRNAs. Consequently, the GMMI was employed to predict novel ncRNA candidates. At the beginning, possible transcription units were extracted from the E. coli genome using consensus sequences for the sigma70 promoter and the rho-independent terminator. Then, these units were evaluated by using the GMMI. This identified 133 candidate ncRNAs, which contain 29 previously annotated small RNA genes and 46 possible antisense ncRNAs. Furthermore 12 transcripts (including five antisense RNAs) were confirmed according to the expression analysis. These data suggests that the expression of small antisense RNAs might be more common than previously thought in the E. coli genome. © 2006 Elsevier B.V. All rights reserved.

Protein-protein interactions play key roles in protein function and the structural organization of a cell. A thorough description of these interactions should facilitate elucidation of cellular activities, targeted-drug design, and whole cell engineering. A large-scale comprehensive pull-down assay was performed using a His-tagged Escherichia coli ORF clone library. Of 4339 bait proteins tested, partners were found for 2667, including 779 of unknown function. Proteins copurifying with hexahistidine-tagged baits on a Ni 2+-NTA column were identified by MALDI-TOF MS (matrix-assisted laser desorption ionization time of flight mass spectrometry). An extended analysis of these interacting networks by bioinformatics and experimentation should provide new insights and novel strategies for E. coli systems biology. ©2006 by Cold Spring Harbor Laboratory Press.

Background: The modeling of dynamic systems requires estimating kinetic parameters from experimentally measured time-courses. Conventional global optimization methods used for parameter estimation, e.g. genetic algorithms (GA), consume enormous computational time because they require iterative numerical integrations for differential equations. When the target model is stiff, the computational time for reaching a solution increases further. Results: In an attempt to solve this problem, we explored a learning technique that uses radial basis function networks (RBFN) to achieve a parameter estimation for biochemical models. RBFN reduce the number of numerical integrations by replacing derivatives with slopes derived from the distribution of searching points. To introduce a slight search bias, we implemented additional data selection using a GA that searches data-sparse areas at low computational cost. In addition, we adopted logarithmic transformation that smoothes the fitness surface to obtain a solution simply. We conducted numerical experiments to validate our methods and compared the results with those obtained by GA. We found that the calculation time decreased by more than 50% and the convergence rate increased from 60% to 90%. Conclusion: In this work, our RBFN technique was effective for parameter optimization of stiff biochemical models. © 2006 Matsubara et al; licensee BioMed Central Ltd.

Background: Successful realization of a "systems biology" approach to analyzing cells is a grand challenge for our understanding of life. However, current modeling approaches to cell simulation are labor-intensive, manual affairs, and therefore constitute a major bottleneck in the evolution of computational cell biology. Results: We developed the Genome-based Modeling (GEM) System for the purpose of automatically prototyping simulation models of cell-wide metabolic pathways from genome sequences and other public biological information. Models generated by the GEM System include an entire Escherichia coli metabolism model comprising 968 reactions of 1195 metabolites, achieving 100% coverage when compared with the KEGG database, 92.38% with the EcoCyc database, and 95.06% with iJR904 genome-scale model. Conclusion: The GEM System prototypes qualitative models to reduce the labor-intensive tasks required for systems biology research. Models of over 90 bacterial genomes are available at our web site. © 2006 Arakawa et al; licensee BioMed Central Ltd.

Background: The ribosome is a central player in the translation system, which in mammals consists of four RNA species and 79 ribosomal proteins (RPs). The control mechanisms of gene expression and the functions of RPs are believed to be identical. Most RP genes have common promoters and were therefore assumed to have a unified gene expression control mechanism. Results: We systematically analyzed the homogeneity and heterogeneity of RP genes on the basis of their expression profiles, promoter structures, encoded amino acid compositions, and codon compositions. The results revealed that (1) most RP genes are coordinately expressed at the mRNA level, with higher signals in the spleen, lymph node dissection (LND), and fetal brain. However, 17 genes, including the P protein genes (RPLP0, RPLP1, RPLP2), are expressed in a tissue-specific manner. (2) Most promoters have GC boxes and possible binding sites for nuclear respiratory factor 2, Yin and Yang 1, and/or activator protein 1. However, they do not have canonical TATA boxes. (3) Analysis of the amino acid composition of the encoded proteins indicated a high lysine and arginine content. (4) The major RP genes exhibit a characteristic synonymous codon composition with high rates of G or C in the third-codon position and a high content of AAG, CAG, ATC, GAG, CAC, and CTG. Conclusion: Eleven of the RP genes are still identified as being unique and did not exhibit at least some of the above characteristics, indicating that they may have unknown functions not present in other RP genes. Furthermore, we found sequences conserved between human and mouse genes around the transcription start sites and in the intronic regions. This study suggests certain overall trends and characteristic features of human RP genes. © 2006 Ishii et al; licensee BioMed Central Ltd.

Translational initiation signals, such as Shine-Dalgarno (SD) sequences in bacteria and Kozak consensus sequences in vertebrates, direct ribosomes to initiate protein synthesis from mRNAs. Investigating sequence characteristics of these signals is important, particularly to infer translational initiation mechanisms. Although various statistical analyses of translational initiation signals have been done, few have focused on base correlations that assess base dependencies in the signal sequences. We used relative entropy and mutual information to analyze base conservation and correlation, respectively, in the 5′ UTRs of various species. In eukaryotes, we found peaks of relative entropy at - 3 from the translational start site but no peak of mutual information at that position, indicating that the base at that position (known as the core base of the Kozak sequence) is well conserved but not correlated with neighboring bases and thus functions as a single base. We observed unexpected peaks of mutual information between positions - 2 and - 1 in most eukaryotes. Surprisingly these base correlation also occurred in some bacteria and archaea, although there were no base preferences at neither position. Various dinucleotide patterns existed at these positions, and the correlation between bases at - 2 and - 1 may be relevant to the context of translational initiation. Because dinucleotide patterns of correlated pairs of nucleotides at - 2 and - 1 were not unique within respective organisms, the correlation could not be found when analyzing single-nucleotide conservation. Therefore, mutual information allowed us to discover signals that were not found by simply analyzing base conservation. © 2006 Elsevier B.V. All rights reserved.

Physicochemcial properties of amino acids are important factors in determining protein structure and function. Most approaches make use of averaged properties over entire domains or even proteins to analyze their structure or function. This level of coarseness tends to hide the richness of the variability in the different properties across functional domains. This paper studies the conservation of physicochemical properties in a functionally similar family of proteins using a novel wavelet-based technique known as multiresolution analysis. Such an analysis can help uncover characteristics that can otherwise remain hidden. We have studied the protein kinase family of sequences and our findings are as follows: (a) a number of different properties are conserved over the functional catalytic domain irrespective of the sequence identities; (b) conservation of properties can be observed at different frequency levels and they agree well with the known structural/ functional properties of the subdomains for the protein kinase family; (c) structural differences between the different kinase family members are reflected in the waveforms; and (d) functionally important mutations show distortions in the waveforms of conserved properties. The potential usefulness of the above findings in identifying functionally similar sequences in the twilight and midnight zones is demonstrated through a simple prediction model for the protein kinase family which achieved a recall of 93.7% and a precision of 96.75% in cross-validation tests. © Imperial College Press.

We developed a computational method to predict the retention times of peptides in HPLC using artificial neural networks (ANN). We performed stepwise multiple linear regressions and selected for ANN input amino acids that significantly affected the LC retention time. Unlike conventional linear models, the trained ANN accurately predicted the retention time of peptides containing up to 50 amino acid residues. In 834 peptides, there was a strong correlation (R2= 0.928) between measured and predicted retention times. We demonstrated the utility of our method by the prediction of the retention time of 121 273 peptides resulting from LysC-digestion of the Escherichia coli proteome. Our approach is useful for the proteome-wide characterization of peptides and the identification of unknown peptide peaks obtained in proteome analysis. © 2006 American Chemical Society.

The genome of an extremely thermophilic bacterium, Thermus thermophilus HB8, contains a single ORF (open reading frame) encoding an RNase-HII-like sequence. Despite the presence of significant amino acid sequence identities with RNase (ribonuclease) HII enzymes, the ORF TTHA0198 could not suppress the temperature-sensitive growth defect of an RNase-H-deficient Escherichia coli mutant and the purified recombinant protein could not cleave an RNA strand of an RNA/DNA heteroduplex, suggesting that the TTHA0198 exhibited no RNase H activity both in vivo and in vitro. When oligomeric RNA-DNA/DNAs were used as a mimic substrate for Okazaki fragments, however, the protein cleaved them only at the 5′ side of the last ribonucleotide at the RNA-DNA junction. In fact, the TTHA0198 protein prefers the RNA-DNA junction to the RNA/DNA hybrid. We have referred to this activity as JRNase (junction RNase) activity, which recognizes an RNA-DNA junction of the RNA-DNA/DNA heteroduplex and cleaves it leaving a mono-ribonucleotide at the 5′ terminus of the RNA-DNA junction. E. coli and Deinococcus radiodurans RNases HII also cleaved the RNA-DNA/DNA substrates at the same site with a different metal-ion preference from that for RNase H activity, implying that the enzymes have JRNase activity as well as RNase H activity. The specialization in the JRNase activity of the RNase HII orthologue from T. thermophilus HB8 (Tth-JRNase) suggests that the JRNase activity of RNase HII enzymes might be independent of the RNase H activity. © The Authors.

Background: The architectural organization of protein structures has been the focus of intense research since it can hopefully lead to an understanding of how proteins fold. In earlier works we had attempted to identify the inherent structural organization in proteins through a study of protein topology. We obtained a modular partitioning of protein structures with the modules correlating well with experimental evidence of early folding units or "foldons". Residues that connect different modules were shown to be those that were protected during the transition phase of folding. Methodology/Principal Findings: In this work, we follow the topological path of ubiquitin through molecular dynamics unfolding simulations. We observed that the use of recurrence quantification analysis (RQA) could lead to the identification of the transition state during unfolding. Additionally, our earlier contention that the modules uncovered through our graph partitioning approach correlated well with early folding units was vindicated through our simulations. Moreover, residues identified from native structure as connector hubs and which had been shown to be those that were protected during the transition phase of folding were indeed more stable (less flexible) well beyond the transition state. Further analysis of the topological pathway suggests that the all pairs shortest path in a protein is minimized during folding. Conclusions: We observed that treating a protein native structure as a network by having amino acid residues as nodes and the non-covalent interactions among them as links allows for the rationalization of many aspects of the folding process. The possibility to derive this information directly from 3D structure opens the way to the prediction of important residues in proteins, while the confirmation of the minimization of APSP for folding allows for the establishment of a potentially useful proxy for kinetic optimality in the validation of sequence-structure predictions. © 2008 Krishnan et al.

Increasing numbers of sense - antisense transcripts (SATs), which are transcribed from the same chromosomal location but in opposite directions, have been identified in various eukaryotic species, but the biological meanings of most SATs remain unclear. To improve understanding of natural sense - antisense transcription, we performed comparative expression profiling of SATs conserved among humans and mice. Using custom oligo-arrays loaded with probes that represented SATs with both protein-coding and non-protein-coding transcripts, we showed that 33% of the 291 conserved SATs displayed identical expression patterns in the two species. Among these SATs, expressional balance inversion of sense - antisense genes was mostly observed in testis at a tissue-specific manner. Northern analyses of the individual conserved SAT loci revealed that: (i) a smeary hybridization pattern was present in mice, but not in humans, and (2) small RNAs (about 60 to 80 nt) were detected from the exon-overlapping regions of SAT loci. In addition, further analyses showed marked alteration of sense - antisense expression balance throughout spermatogenesis in testis. These results suggest that conserved SAT loci are rich in potential regulatory roles that will help us understand this new class of transcripts underlying the mammalian genome. © The Author 2008. Published by Oxford University Press. All rights reserved.

The discovery of separate 5′ and 3′ halves of transfer RNA (tRNA) molecules - so called split tRNA - in the archaeal parasite Nanoarchaeum equitans made us wonder whether ancestral tRNA was encoded on 1 or 2 genes. We performed a comprehensive phylogenetic analysis of tRNAs in 45 archaal species to explore the relationship between the three types of tRNAs (nonintronic, intronic and split). We classified 1953 mature tRNA sequences into 22 clusters. All split tRNAs have shown phylogenetic relationships with other tRNAs possessing the same anticodon. We also mimicked split tRNA by artificially separating the tRNA sequences of 7 primitive archaeal species at the anticodon and analyzed the sequence similarity and diversity of the 5′ and 3′ tRNA halves. Network analysis revealed specific characteristics of and topological differences between the 5′ and 3′ tRNA halves: the 5′ half sequences were categorized into 6 distinct groups with a sequence similarity of > 80%, while the 3′ half sequences were categorized into 9 groups with a higher sequence similarity of > 88%, suggesting different evolutionary backgrounds of the 2 halves. Furthermore, the combinations of 5′ and 3′ halves coresponded with the variation of amino acids in the codon table. We found not only universally conserved combinations of 5′-3′ tRNA halves in tRNS Imet, tRNA Thr, tRNAIle, tRNAGly, tRNAGin, tRNAGlu, tRNAAsp, tRNALys, tRNA Arg and tRNA but also phylum-specific combinations in tRNAPro, tRNAAla, and tRNATrp. Our results support the idea that tRNA emerged through the combination of separate genes and explain the sequence diversity that arose during archaeal tRNA evolution. © 2008 Fujishima et al.

The network paradigm is based on the derivation of emerging properties of studied systems by their representation as oriented graphs: any system is traced back to a set of nodes (its constituent elements) linked by edges (arcs) correspondent to the relations existing between the nodes. This allows for a straightforward quantitative formalization of systems by means of the computation of mathematical descriptors of such graphs (graph theory). The network paradigm is particularly useful when it is clear which elements of the modelled system must play the role of nodes and arcs respectively, and when topological constraints have a major role with respect to kinetic ones. In this review we demonstrate how nodes and arcs of protein topology are characterized at different levels of definition: 1. Recurrence matrix of hydrophobicity patterns along the sequence 2. Contact matrix of alpha carbons of 3D structures 3. Correlation matrix of motions of different portion of the molecule in molecular dynamics. These three conditions represent different but potentially correlated reticular systems that can be profitably analysed by means of network analysis tools. © 2008 Bentham Science Publishers Ltd.

Retention times in HPLC yield valuable information for the identification of various analytes and the prediction of peptide retention is useful for the identification of peptides/proteins in LC-MS-based proteomics. Informatics methods such as artificial neural networks and support vector machines capable of solving nonlinear problems made possible the accurate modeling of quantitative structure-retention relationships of peptides (including large polymers) up to 5 kDa to which classical linear models cannot be applied, as well as the proteome-wide prediction of peptide retention. Proteome-wide retention prediction and accurate mass-information facilitate the identification of peptides in complex proteomic samples. In this review, we address recent developments in solid informatics methods and their application to peptide-retention properties in 'bottom-up' shotgun proteomics. We also describe future prospects for the standardization and application of retention times. © 2008 Wiley-VCH Verlag GmbH & Co. KGaA.

The genome of Bacillus subtilis 168 was modified to yield a genome vector for the cloning of DNA several Mb in size. Unlike contemporary plasmid-based vectors, this 4.2 Mb genome vector requires specific in vivo handling protocols because of its large size. Inversion mutagenesis, a method to modify local genome structure without gain or loss of genes, was applied intensively to the B. subtilis genome; this technique made possible both exchange and translocation of designated regions of the genome. This method not only reshuffles the genome of B. subtilis, but can provide insight into the biologic principles underlying genome plasticity. © 2007 The Japanese Biochemical Society.

Gene Regulatory Networks (GRNs) have become a major focus of interest in recent years. Although much work has been done in elucidating the transcriptional regulatory network, the underlying mechanisms that have possibly influenced the evolution of these GRNs are still debatable. We have developed a framework to analyze the effect of objective functions, input types and starting populations on the evolution of GRNs with a specific emphasis on the robustness of evolved GRNs. We observed that robustness evolves along with the networks as an emergent property even in the absence of specific selective pressure towards more robust systems. In addition, robustness was independent of the selective pressure, input types or the initial starting populations. We also observed the existence of multiple genotypes giving rise to the same phenotype in accordance with the theoretical view that natural selection operates on phenotypes thereby accommodating variation in the genotype by fixing those changes that are phenotype-neutral. This study gives a proof-of-concept of the fact that robustness is an emergent property of GRNs as well as of the degeneracy of the network topology/function relationship analogous to the sequence/structure problem in proteins. © 2007 Elsevier Ltd. All rights reserved.

Metabolic changes in response to histidine starvation were observed in histidine-auxotrophic Escherichia coli using a capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS)-based metabolomics technique. Prior to the analysis, we prepared an E. coli metabolome list of 727 metabolites reported in the literature. An improved method for metabolite extraction was developed, which resulted in higher extraction efficiency in phosphate-rich metabolites, e.g., ATP and GTP. Based on the results, 375 charged, hydrophilic intermediates in primary metabolisms were analysed simultaneously, providing quantitative data of 198 metabolites. We confirmed that the intracellular levels of intermediates in histidine biosynthesis are rapidly accumulated in response to a drop in histidine level under histidine-starved conditions. Simultaneously, disciplined responses were observed in the glycolysis, tricarboxylic acid cycle, and amino acid and nucleotide biosynthesis pathways as regulated by amino acid starvation. © The Royal Society of Chemistry.

We have developed an on-line automated system for phosphoproteome analysis using titania-based phosphopeptide enrichment followed by nanoLC-MS/MS. Titania beads were prepared by calcination of commercial chromatographic titania beads at 800°C to convert the crystalline structure. The obtained rutile-form titania exhibited higher selectivity in phosphopeptide enrichment than commercial titania, even in the absence of a competitive chelating reagent for non-phosphopeptides. For phosphoproteome analysis of human cervical cancer HeLa cells, tryptic digests of the cell extracts were directly injected into this on-line system, and 696 non-redundant phosphopeptides with 671 unambiguously determined phosphorylation sites, derived from 512 phosphoproteins, were successfully identified. This is the first successful application of an on-line automated phosphoproteome analysis system to complex biological samples. 2008 © The Japan Society for Analytical Chemistry.