The applicability of recurrent neural networks for biological sequence analysis

Selection of machine learning techniques requires a certain sensitivity to the requirements of the problem. In particular, the problem can be made more tractable by deliberately using algorithms that are biased toward solutions of the requisite kind. In this paper, we argue that recurrent neural networks have a natural bias toward a problem domain of which biological sequence analysis tasks are a subset. We use experiments with synthetic data to illustrate this bias. We then demonstrate that this bias can be exploitable using a data set of protein sequences containing several classes of subcellular localization targeting peptides. The results show that, compared with feed forward, recurrent neural networks will generally perform better on sequence analysis tasks. Furthermore, as the patterns within the sequence become more ambiguous, the choice of specific recurrent architecture becomes more critical.

Sequence analysis of rabbit haemorrhagic disease virus (RHDV) in Australia: alterations after its release

Liver samples from rabbits killed by RHDV, collected from five States in Australia in 1996 and 1997 were analysed by RT-PCR. A 398 bp fragment of the capsid protein (VP60) gene was amplified by PCR and directly sequenced. The alignment of the nucleotide and amino acid sequences and their comparison with the original strain of the virus released in Australia indicated genetic changes after two years have been small with 98.2% to 100% identity. The constructed phylogenetic tree suggests slight differences in nucleotide substitutions in various States but there is no clear evidence of clustering of sequences according to their geographic origin. In practical terms, sequencing of viral RNA provides a means of testing the efficacy of further releases and subsequent spread of the virus if such a strategy is employed as a means of enhancing RHD as a biological control of the wild rabbit in Australia.

Sequence analysis and expression of the attachment and fusion proteins of canine distemper virus wild-type strain A75/17.

The biological properties of wild-type A75/17 and cell culture-adapted Onderstepoort canine distemper virus differ markedly. To learn more about the molecular basis for these differences, we have isolated and sequenced the protein-coding regions of the attachment and fusion proteins of wild-type canine distemper virus strain A75/17. In the attachment protein, a total of 57 amino acid differences were observed between the Onderstepoort strain and strain A75/17, and these were distributed evenly over the entire protein. Interestingly, the attachment protein of strain A75/17 contained an extension of three amino acids at the C terminus. Expression studies showed that the attachment protein of strain A75/17 had a higher apparent molecular mass than the attachment protein of the Onderstepoort strain, in both the presence and absence of tunicamycin. In the fusion protein, 60 amino acid differences were observed between the two strains, of which 44 were clustered in the much smaller F2 portion of the molecule. Significantly, the AUG that has been proposed as a translation initiation codon in the Onderstepoort strain is an AUA codon in strain A75/17. Detailed mutation analyses showed that both the first and second AUGs of strain A75/17 are the major translation initiation sites of the fusion protein. Similar analyses demonstrated that, also in the Onderstepoort strain, the first two AUGs are the translation initiation codons which contribute most to the generation of precursor molecules yielding the mature form of the fusion protein.

Comparative sequence analysis of the P-, M- and L-coding region of the measles virus CAM-70 live attenuated vaccine strain

Measles virus is a highly contagious agent which causes a major health problem in developing countries. The viral genomic RNA is single-stranded, nonsegmented and of negative polarity. Many live attenuated vaccines for measles virus have been developed using either the prototype Edmonston strain or other locally isolated measles strains. Despite the diverse geographic origins of the vaccine viruses and the different attenuation methods used, there was remarkable sequence similarity of H, F and N genes among all vaccine strains. CAM-70 is a Japanese measles attenuated vaccine strain widely used in Brazilian children and produced by Bio-Manguinhos since 1982. Previous studies have characterized this vaccine biologically and genomically. Nevertheless, only the F, H and N genes have been sequenced. In the present study we have sequenced the remaining P, M and L genes (approximately 1.6, 1.4 and 6.5 kb, respectively) to complete the genomic characterization of CAM-70 and to assess the extent of genetic relationship between CAM-70 and other current vaccines. These genes were amplified using long-range or standard RT-PCR techniques, and the cDNA was cloned and automatically sequenced using the dideoxy chain-termination method. The sequence analysis comparing previously sequenced genotype A strains with the CAM-70 Bio-Manguinhos strain showed a low divergence among them. However, the CAM-70 strains (CAM-70 Bio-Manguinhos and a recently sequenced CAM-70 submaster seed strain) were assigned to a specific group by phylogenetic analysis using the neighbor-joining method. Information about our product at the genomic level is important for monitoring vaccination campaigns and for future studies of measles virus attenuation.

Genotyping hepatitis C virus from hemodialysis patients in Central Brazil by line probe assay and sequence analysis

The present study examined the distribution of hepatitis C virus (HCV) genotypes and subtypes in a hemodialysis population in Goiás State, Central Brazil, and evaluated the efficiency of two genotyping methods: line probe assay (LiPA) based on the 5' noncoding region and nucleotide sequencing of the nonstructural 5B (NS5B) region of the genome. A total of 1095 sera were tested for HCV RNA by RT-nested PCR of the 5' noncoding region. The LiPA assay was able to genotype all 131 HCV RNA-positive samples. Genotypes 1 (92.4%) and 3 (7.6%) were found. Subtype 1a (65.7%) was the most prevalent, followed by subtypes 1b (26.7%) and 3a (7.6%). Direct nucleotide sequencing of 340 bp from the NS5B region was performed in 106 samples. The phylogenetic tree showed that 98 sequences (92.4%) were classified as genotype 1, subtypes 1a (72.6%) and 1b (19.8%), and 8 sequences (7.6%) as subtype 3a. The two genotyping methods gave concordant results within HCV genotypes and subtypes in 100 and 96.2% of cases, respectively. Only four samples presented discrepant results, with LiPA not distinguishing subtypes 1a and 1b. Therefore, HCV genotype 1 (subtype 1a) is predominant in hemodialysis patients in Central Brazil. By using sequence analysis of the NS5B region as a reference standard method for HCV genotyping, we found that LiPA was efficient at the genotype level, although some discrepant results were observed at the subtype level (sensitivity of 96.1% for subtype 1a and 95.2% for subtype 1b). Thus, analysis of the NS5B region permitted better discrimination between HCV subtypes, as required in epidemiological investigations.

Cloning, sequence analysis, and expression of cDNA coding for the major house dust mite allergen, Der f 1, in Escherichia coli

Our objective was to clone, express and characterize adult Dermatophagoides farinae group 1 (Der f 1) allergens to further produce recombinant allergens for future clinical applications in order to eliminate side reactions from crude extracts of mites. Based on GenBank data, we designed primers and amplified the cDNA fragment coding for Der f 1 by nested-PCR. After purification and recovery, the cDNA fragment was cloned into the pMD19-T vector. The fragment was then sequenced, subcloned into the plasmid pET28a(+), expressed in Escherichia coli BL21 and identified by Western blotting. The cDNA coding for Der f 1 was cloned, sequenced and expressed successfully. Sequence analysis showed the presence of an open reading frame containing 966 bp that encodes a protein of 321 amino acids. Interestingly, homology analysis showed that the Der p 1 shared more than 87% identity in amino acid sequence with Eur m 1 but only 80% with Der f 1. Furthermore, phylogenetic analyses suggested that D. pteronyssinus was evolutionarily closer to Euroglyphus maynei than to D. farinae, even though D. pteronyssinus and D. farinae belong to the same Dermatophagoides genus. A total of three cysteine peptidase active sites were found in the predicted amino acid sequence, including 127-138 (QGGCGSCWAFSG), 267-277 (NYHAVNIVGYG) and 284-303 (YWIVRNSWDTTWGDSGYGYF). Moreover, secondary structure analysis revealed that Der f 1 contained an a helix (33.96%), an extended strand (17.13%), a ß turn (5.61%), and a random coil (43.30%). A simple three-dimensional model of this protein was constructed using a Swiss-model server. The cDNA coding for Der f 1 was cloned, sequenced and expressed successfully. Alignment and phylogenetic analysis suggests that D. pteronyssinus is evolutionarily more similar to E. maynei than to D. farinae.

Sequence analysis of the 5′ third of glycoprotein C gene of South American bovine herpesviruses 1 and 5

Bovine herpesviruses 1 (BoHV-1) and 5 (BoHV-5) share high genetic and antigenic similarities, but exhibit marked differences in tissue tropism and neurovirulence. The amino-terminal region of glycoprotein C (gC), which is markedly different in each of the viruses, is involved in virus binding to cellular receptors and in interactions with the immune system. This study investigated the genetic and antigenic differences of the 5′ region of the gC (5′ gC) gene (amino-terminal) of South American BoHV-1 (n=19) and BoHV-5 (n=25) isolates. Sequence alignments of 374 nucleotides (104 amino acids) revealed mean similarity levels of 97.3 and 94.2% among BoHV-1 gC (gC1), respectively, 96.8 and 95.6% among BoHV-5 gC (gC5), and 62 and 53.3% between gC1 and gC5. Differences included the absence of 40 amino acid residues (27 encompassing predicted linear epitopes) scattered throughout 5′ gC1 compared to 5′ gC5. Virus neutralizing assays testing BoHV-1 and BoHV-5 antisera against each isolate revealed a high degree of cross-neutralization between the viruses, yet some isolates were neutralized at very low titers by heterologous sera, and a few BoHV-5 isolates reacted weakly with either sera. The virus neutralization differences observed within the same viral species, and more pronounced between BoHV-1 and BoHV-5, likely reflect sequence differences in neutralizing epitopes. These results demonstrate that the 5′ gC region is well conserved within each viral species but is divergent between BoHV-1 and BoHV-5, likely contributing to their biological and antigenic differences.

Gene cloning, sequence analysis, and expression of 2-methyl-3-hydroxypyridine-5-carboxylic acid oxygenase

The gene encoding 2-methyl-3-hydroxypyridine-5-carboxylic acid oxygenase (MHPCO; EC 1.14.12.4) was cloned by using an oligonucleotide probe corresponding to the N terminus of the enzyme to screen a DNA library of Pseudomonas sp. MA-1. The gene encodes for a protein of 379 amino acid residues corresponding to a molecular mass of 41.7 kDa, the same as that previously estimated for MHPCO. MHPCO was expressed in Escherichia coli and found to have the same properties as the native enzyme from Pseudomonas sp. MA-1. This study shows that MHPCO is a homotetrameric protein with one flavin adenine dinucleotide bound per subunit. Sequence comparison of the enzyme with other hydroxylases reveals regions that are conserved among aromatic flavoprotein hydroxylases.

Molecular dynamics of MHC genesis unraveled by sequence analysis of the 1,796,938-bp HLA class I region

The intensely studied MHC has become the paradigm for understanding the architectural evolution of vertebrate multigene families. The 4-Mb human MHC (also known as the HLA complex) encodes genes critically involved in the immune response, graft rejection, and disease susceptibility. Here we report the continuous 1,796,938-bp genomic sequence of the HLA class I region, linking genes between MICB and HLA-F. A total of 127 genes or potentially coding sequences were recognized within the analyzed sequence, establishing a high gene density of one per every 14.1 kb. The identification of 758 microsatellite provides tools for high-resolution mapping of HLA class I-associated disease genes. Most importantly, we establish that the repeated duplication and subsequent diversification of a minimal building block, MIC-HCGIX-3.8–1-P5-HCGIV-HLA class I-HCGII, engendered the present-day MHC. That the currently nonessential HLA-F and MICE genes have acted as progenitors to today’s immune-competent HLA-ABC and MICA/B genes provides experimental evidence for evolution by “birth and death,” which has general relevance to our understanding of the evolutionary forces driving vertebrate multigene families.

Genomic and cDNA sequence analysis of the cell matrix adhesion regulator gene

The cell matrix adhesion regulator (CMAR) gene has been suggested to be a signal transduction molecule influencing cell adhesion to collagen and, through this, possibly involved in tumor suppression. The originally reported CMAR cDNA was 464 bp long with a tyrosine phosphorylation site at the extreme 3′ end, which mutagenesis studies had shown to be central to the function of this gene. Since the discovery of a 4-bp insertion polymorphism within the originally reported coding region, further sequence information has been obtained. The cDNA has been extended 5′ by ≈2 kb revealing a 559-bp region showing strong homology to the proposed 5′ untranslated sequence of a murine protein kinase receptor family member, variant in kinase (vik). CMAR genomic sequencing has shown the presence of an intron, the intron/exon boundary lying within this region of homology. An RNA transcript for CMAR of ≈2.5 kb has also been identified. The data suggest complex mechanisms for control of expression of two closely associated genes, CMAR and the vik- associated sequence.

Molecular cloning and sequence analysis of the complestatin biosynthetic gene cluster

Streptomyces lavendulae produces complestatin, a cyclic peptide natural product that antagonizes pharmacologically relevant protein–protein interactions including formation of the C4b,2b complex in the complement cascade and gp120-CD4 binding in the HIV life cycle. Complestatin, a member of the vancomycin group of natural products, consists of an α-ketoacyl hexapeptide backbone modified by oxidative phenolic couplings and halogenations. The entire complestatin biosynthetic and regulatory gene cluster spanning ca. 50 kb was cloned and sequenced. It consisted of 16 ORFs, encoding proteins homologous to nonribosomal peptide synthetases, cytochrome P450-related oxidases, ferredoxins, nonheme halogenases, four enzymes involved in 4-hydroxyphenylglycine (Hpg) biosynthesis, transcriptional regulators, and ABC transporters. The nonribosomal peptide synthetase consisted of a priming module, six extending modules, and a terminal thioesterase; their arrangement and domain content was entirely consistent with functions required for the biosynthesis of a heptapeptide or α-ketoacyl hexapeptide backbone. Two oxidase genes were proposed to be responsible for the construction of the unique aryl-ether-aryl-aryl linkage on the linear heptapeptide intermediate. Hpg, 3,5-dichloro-Hpg, and 3,5-dichloro-hydroxybenzoylformate are unusual building blocks that repesent five of the seven requisite monomers in the complestatin peptide. Heterologous expression and biochemical analysis of 4-hydroxyphenylglycine transaminon confirmed its role as an aminotransferase responsible for formation of all three precursors. The close similarity but functional divergence between complestatin and chloroeremomycin biosynthetic genes also presents a unique opportunity for the construction of hybrid vancomycin-type antibiotics.

Evolution of chlorophyll and bacteriochlorophyll: the problem of invariant sites in sequence analysis.

Competing hypotheses seek to explain the evolution of oxygenic and anoxygenic processes of photosynthesis. Since chlorophyll is less reduced and precedes bacteriochlorophyll on the modern biosynthetic pathway, it has been proposed that chlorophyll preceded bacteriochlorophyll in its evolution. However, recent analyses of nucleotide sequences that encode chlorophyll and bacteriochlorophyll biosynthetic enzymes appear to provide support for an alternative hypothesis. This is that the evolution of bacteriochlorophyll occurred earlier than the evolution of chlorophyll. Here we demonstrate that the presence of invariant sites in sequence datasets leads to inconsistency in tree building (including maximum-likelihood methods). Homologous sequences with different biological functions often share invariant sites at the same nucleotide positions. However, different constraints can also result in additional invariant sites unique to the genes, which have specific and different biological functions. Consequently, the distribution of these sites can be uneven between the different types of homologous genes. The presence of invariant sites, shared by related biosynthetic genes as well as those unique to only some of these genes, has misled the recent evolutionary analysis of oxygenic and anoxygenic photosynthetic pigments. We evaluate an alternative scheme for the evolution of chlorophyll and bacteriochlorophyll.

Cloning and sequence analysis of pituitary prolactin cDNA from the northern brown bandicoot (Isoodon macrourus)

The nuclectide sequence for pituitary prolactin cDNA from the marsupial bandicoot (Isoodon macrourus) was determined by reverse transcription-polymerase chain reaction and 5'/3' rapid amplification of cDNA ends. The deduced amino acid sequence showed high sequence identity with brushtail possum prolactin (95%) and all of the expected structural features of a quadruped prolactin. A prolactin gene tree was constructed and rates of evolution calculated for bandicoot, possum, opossum and several mammalian and non-mammalian prolactins. Bootstrap analysis provided strong support for marsupials as a sister group with eutherian mammals and weak support for opossum and bandicoot as an independent grouping from the brushtail possum. The rates of molecular evolution for marsupial prolactins were comparable to the slow rate seen in the majority of quadruped prolactins that have been sequenced. (c) 2005 Elsevier Inc. All rights reserved.