A computational approach to identify genes for functional RNAs in genomic sequences

Currently there is no successful computational approach for identification of genes encoding novel functional RNAs (fRNAs) in genomic sequences. We have developed a machine learning approach using neural networks and support vector machines to extract common features among known RNAs for prediction of new RNA genes in the unannotated regions of prokaryotic and archaeal genomes. The Escherichia coli genome was used for development, but we have applied this method to several other bacterial and archaeal genomes. Networks based on nucleotide composition were 80–90% accurate in jackknife testing experiments for bacteria and 90–99% for hyperthermophilic archaea. We also achieved a significant improvement in accuracy by combining these predictions with those obtained using a second set of parameters consisting of known RNA sequence motifs and the calculated free energy of folding. Several known fRNAs not included in the training datasets were identified as well as several hundred predicted novel RNAs. These studies indicate that there are many unidentified RNAs in simple genomes that can be predicted computationally as a precursor to experimental study. Public access to our RNA gene predictions and an interface for user predictions is available via the web.     

Gene identification in novel eukaryotic genomes by self-training algorithm

Finding new protein-coding genes is one of the most important goals of eukaryotic genome sequencing projects. However, genomic organization of novel eukaryotic genomes is diverse and ab initio gene finding tools tuned up for previously studied species are rarely suitable for efficacious gene hunting in DNA sequences of a new genome. Gene identification methods based on cDNA and expressed sequence tag (EST) mapping to genomic DNA or those using alignments to closely related genomes rely either on existence of abundant cDNA and EST data and/or availability on reference genomes. Conventional statistical ab initio methods require large training sets of validated genes for estimating gene model parameters. In practice, neither one of these types of data may be available in sufficient amount until rather late stages of the novel genome sequencing. Nevertheless, we have shown that gene finding in eukaryotic genomes could be carried out in parallel with statistical models estimation directly from yet anonymous genomic DNA. The suggested method of parallelization of gene prediction with the model parameters estimation follows the path of the iterative Viterbi training. Rounds of genomic sequence labeling into coding and non-coding regions are followed by the rounds of model parameters estimation. Several dynamically changing restrictions on the possible range of model parameters are added to filter out fluctuations in the initial steps of the algorithm that could redirect the iteration process away from the biologically relevant point in parameter space. Tests on well-studied eukaryotic genomes have shown that the new method performs comparably or better than conventional methods where the supervised model training precedes the gene prediction step. Several novel genomes have been analyzed and biologically interesting findings are discussed. Thus, a self-training algorithm that had been assumed feasible only for prokaryotic genomes has now been developed for ab initio eukaryotic gene identification.     

Finding prokaryotic genes by the 'frame-by-frame' algorithm: targeting gene starts and overlapping genes

MOTIVATION: Tightly packed prokaryotic genes frequently overlap with each other. This feature, rarely seen in eukaryotic DNA, makes detection of translation initiation sites and, therefore, exact predictions of prokaryotic genes notoriously difficult. Improving the accuracy of precise gene prediction in prokaryotic genomic DNA remains an important open problem. RESULTS: A software program implementing a new algorithm utilizing a uniform Hidden Markov Model for prokaryotic gene prediction was developed. The algorithm analyzes a given DNA sequence in each of six possible global reading frames independently. Twelve complete prokaryotic genomes were analyzed using the new tool. The accuracy of gene finding, predicting locations of protein-coding ORFs, as well as the accuracy of precise gene prediction, and detecting the whole gene including translation initiation codon were assessed by comparison with existing annotation. It was shown that in terms of gene finding, the program performs at least as well as the previously developed tools, such as GeneMark and GLIMMER. In terms of precise gene prediction the new program was shown to be more accurate, by several percentage points, than earlier developed tools, such as GeneMark.hmm, ECOPARSE and ORPHEUS. The results of testing the program indicated the possibility of systematic bias in start codon annotation in several early sequenced prokaryotic genomes. AVAILABILITY: The new gene-finding program can be accessed through the Web site: http:@dixie.biology.gatech.edu/GeneMark/fbf.cgi CONTACT: mark@amber.gatech.edu.     

原核基因翻译起始位点预测的新方法

翻译起始位点（TIS,即基因5’端）的精确定位是原核生物基因预测的一个关键问题,而基因组GC含量和翻译起始机制的多样性是影响当前TIS预测水平的重要因素．结合基因组结构的复杂信息（包括GC含量、TIS邻近序列及上游调控信号、序列编码潜能、操纵子结构等）,发展刻画翻译起始机制的数学统计模型,据此设计TIS预测的新算法MED．StartPlus．并将MED．StartPlus与同类方法RBSfinder、GS．Finder、MED-Start、TiCo和Hon-yaku等进行系统地比较和评价．测试针对两种数据集进行：当前14个已知的TIS被确认的基因数据集,以及300个物种中功能已知的基因数据集．测试结果表明,MED-StartPlus的预测精度在总体上超过同类方法．尤其是对高GC含量基因组以及具有复杂翻译起始机制的基因组,MED-StartPlus具有明显的优势．     

Exact mapping of prokaryotic gene starts

It is known that while the programs used to find genes in prokaryotic genomes reliably map protein-coding regions, they often fail in the exact determination of gene starts. This problem is further aggravated by sequencing errors, most notably insertions and deletions leading to frame-shifts. Therefore, the exact mapping of gene starts and identification of frame-shifts are important problems of the computer-assisted functional analysis of newly sequenced genomes. Here we review methods of gene recognition and describe a new algorithm for correction of gene starts and identification of frame-shifts in prokaryotic genomes. The algorithm is based on the comparison of nucleotide and protein sequences of homologous genes from related organisms, using the assumption that the rate of evolutionary changes in protein-coding regions is lower than that in non-coding regions. A dynamic programming algorithm is used to align protein sequences obtained by formal translation of genomic nucleotide sequences. The possibility of frame-shifts is taken into account. The algorithm was tested on several groups of related organisms: gamma-proteobacteria, the Bacillus/Clostridium group, and three Pyrococcus genomes. The testing demonstrated that, dependent or a genome, 1-10 per cent of genes have incorrect starts or contain frame-shifts. The algorithm is implemented in the program package Orthologator-GeneCorrector.     

Genome-wide de novo prediction of cis-regulatory binding sites in prokaryotes

Although cis-regulatory binding sites (CRBSs) are at least as important as the coding sequences in a genome, our general understanding of them in most sequenced genomes is very limited due to the lack of efficient and accurate experimental and computational methods for their characterization, which has largely hindered our understanding of many important biological processes. In this article, we describe a novel algorithm for genome-wide de novo prediction of CRBSs with high accuracy. We designed our algorithm to circumvent three identified difficulties for CRBS prediction using comparative genomics principles based on a new method for the selection of reference genomes, a new metric for measuring the similarity of CRBSs, and a new graph clustering procedure. When operon structures are correctly predicted, our algorithm can predict 81% of known individual binding sites belonging to 94% of known cis-regulatory motifs in the Escherichia coli K12 genome, while achieving high prediction specificity. Our algorithm has also achieved similar prediction accuracy in the Bacillus subtilis genome, suggesting that it is very robust, and thus can be applied to any other sequenced prokaryotic genome. When compared with the prior state-of-the-art algorithms, our algorithm outperforms them in both prediction sensitivity and specificity.     

FrameD: A flexible program for quality check and gene prediction in prokaryotic genomes and noisy matured eukaryotic sequences

We describe FrameD, a program that predicts coding regions in prokaryotic and matured eukaryotic sequences. Initially targeted at gene prediction in bacterial GC rich genomes, the gene model used in FrameD also allows to predict genes in the presence of frameshifts and partially undetermined sequences which makes it also very suitable for gene prediction and frameshift correction in unfinished sequences such as EST and EST cluster sequences. Like recent eukaryotic gene prediction programs, FrameD also includes the ability to take into account protein similarity information both in its prediction and its graphical output. Its performances are evaluated on different bacterial genomes. The web site (http://genopole.toulouse.inra.fr/bioinfo/FrameD/FD) allows direct prediction, sequence correction and translation and the ability to learn new models for new organisms.     

ZCURVE: a new system for recognizing protein-coding genes in bacterial and archaeal genomes

A new system, ZCURVE 1.0, for finding protein- coding genes in bacterial and archaeal genomes has been proposed. The current algorithm, which is based on the Z curve representation of the DNA sequences, lays stress on the global statistical features of protein-coding genes by taking the frequencies of bases at three codon positions into account. In ZCURVE 1.0, since only 33 parameters are used to characterize the coding sequences, it gives better consideration to both typical and atypical cases, whereas in Markov-model-based methods, e.g. Glimmer 2.02, thousands of parameters are trained, which may result in less adaptability. To compare the performance of the new system with that of Glimmer 2.02, both systems were run, respectively, for 18 genomes not annotated by the Glimmer system. Comparisons were also performed for predicting some function-known genes by both systems. Consequently, the average accuracy of both systems is well matched; however, ZCURVE 1.0 has more accurate gene start prediction, lower additional prediction rate and higher accuracy for the prediction of horizontally transferred genes. It is shown that the joint applications of both systems greatly improve gene-finding results. For a typical genome, e.g. Escherichia coli, the system ZCURVE 1.0 takes approximately 2 min on a Pentium III 866 PC without any human intervention. The system ZCURVE 1.0 is freely available at: http://tubic. tju.edu.cn/Zcurve_B/.     

Comparison of Methods of Detection of Exceptional Sequences in Prokaryotic Genomes

Many proteins need recognition of specific DNA sequences for functioning. The number of recognition sites and their distribution along the DNA might be of biological importance. For example, the number of restriction sites is often reduced in prokaryotic and phage genomes to decrease the probability of DNA cleavage by restriction endonucleases. We call a sequence an exceptional one if its frequency in a genome significantly differs from one predicted by some mathematical model. An exceptional sequence could be either under- or over-represented, depending on its frequency in comparison with the predicted one. Exceptional sequences could be considered biologically meaningful, for example, as targets of DNA-binding proteins or as parts of abundant repetitive elements. Several methods to predict frequency of a short sequence in a genome, based on actual frequencies of certain its subsequences, are used. The most popular are methods based on Markov chain models. But any rigorous comparison of the methods has not previously been performed. We compared three methods for the prediction of short sequence frequencies: the maximum-order Markov chain model-based method, the method that uses geometric mean of extended Markovian estimates, and the method that utilizes frequencies of all subsequences including discontiguous ones. We applied them to restriction sites in complete genomes of 2500 prokaryotic species and demonstrated that the results depend greatly on the method used: lists of 5% of the most under-represented sites differed by up to 50%. The method designed by Burge and coauthors in 1992, which utilizes all subsequences of the sequence, showed a higher precision than the other two methods both on prokaryotic genomes and randomly generated sequences after computational imitation of selective pressure. We propose this method as the first choice for detection of exceptional sequences in prokaryotic genomes.     

Probabilistic methods of identifying genes in prokaryotic genomes: connections to the HMM theory

In this paper, we review developments in probabilistic methods of gene recognition in prokaryotic genomes with the emphasis on connections to the general theory of hidden Markov models (HMM). We show that the Bayesian method implemented in GeneMark, a frequently used gene-finding tool, can be augmented and reintroduced as a rigorous forward-backward (FB) algorithm for local posterior decoding described in the HMM theory. Another earlier developed method, prokaryotic GeneMark.hmm, uses a modification of the Viterbi algorithm for HMM with duration to identify the most likely global path through hidden functional states given the DNA sequence. GeneMark and GeneMark.hmm programs are worth using in concert for analysing prokaryotic DNA sequences that arguably do not follow any exact mathematical model. The new extension of GeneMark using the FB algorithm was implemented in the software program GeneMark.fba. Given the DNA sequence, this program determines an a posteriori probability for each nucleotide to belong to coding or non-coding region. Also, for any open reading frame (ORF), it assigns a score defined as a probabilistic measure of all paths through hidden states that traverse the ORF as a coding region. The prediction accuracy of GeneMark.fba determined in our tests was compared favourably to the accuracy of the initial (standard) GeneMark program. Comparison to the prokaryotic GeneMark.hmm has also demonstrated a certain, yet species-specific, degree of improvement in raw gene detection, ie detection of correct reading frame (and stop codon). The accuracy of exact gene prediction, which is concerned about precise prediction of gene start (which in a prokaryotic genome unambiguously defines the reading frame and stop codon, thus, the whole protein product), still remains more accurate in GeneMarkS, which uses more elaborate HMM to specifically address this task.     

Identification of functional open reading frames in chloroplast genomes

We have used a rapid computer dot-matrix comparison method to identify all DNA regions which have been evolutionarily conserved between the completely sequenced chloroplast genomes of tobacco and a liverwort. Analysis of these regions reveals 74 homologous open reading frames (ORFs) which have been conserved as to length and amino acid sequence; these ORFs also have an excess of nucleotide substitutions at silent sites of codons. Since the nonfunctional parts of these genomes have become saturated with mutations and show no sequence similarity whatsoever, the homologous ORFs are almost certainly functional. A further four pairs of ORFs show homology limited to only a short part of their putative gene products. Amino acid sequence identities range between 50 and 99%; some chloroplast proteins are seen to be among the most slowly evolving of all known proteins. A search of the nucleotide and amino acid sequence databanks has revealed several previously unidentified genes in chloroplast sequences from other species, but no new homologies to prokaryotic genes.     

Starts of bacterial genes: estimating the reliability of computer predictions.

Exact mapping of gene starts is an important problem in the computer-assisted functional analysis of newly sequenced prokaryotic genomes. We describe an algorithm for finding ribosomal binding sites without a learning sample. This algorithm is particularly useful for analysis of genomes with little or no experimentally mapped genes. There is a clear correlation between the ribosomal binding site (RBS) properties of a given genome and the potential gene start prediction accuracy. This correlation is of considerable predictive power and may be useful for estimating the expected success of future genome analysis efforts. We also demonstrate that the RBS properties depend on the phylogenetic position of a genome.     

SCGPred： A Score-based Method for Gene Structure Prediction by Combining Multiple Sources of Evidence

Predicting protein-coding genes still remains a significant challenge. Although a variety of computational programs that use commonly machine learning methods have emerged, the accuracy of predictions remains a low level when implementing in large genomic sequences. Moreover, computational gene finding in newly se- quenced genomes is especially a difficult task due to the absence of a training set of abundant validated genes. Here we present a new gene-finding program, SCGPred, to improve the accuracy of prediction by combining multiple sources of evidence. SCGPred can perform both supervised method in previously well-studied genomes and unsupervised one in novel genomes. By testing with datasets composed of large DNA sequences from human and a novel genome of Ustilago maydi, SCGPred gains a significant improvement in comparison to the popular ab initio gene predictors. We also demonstrate that SCGPred can significantly improve prediction in novel genomes by combining several foreign gene finders with similarity alignments, which is superior to other unsupervised methods. Therefore, SCGPred can serve as an alternative gene-finding tool for newly sequenced eukaryotic genomes. The program is freely available at http：//bio.scu.edu.cn/SCGPred/.     

A compression-based approach for coding sequences identification. I. Application to prokaryotic genomes.

Most of the gene prediction algorithms for prokaryotes are based on Hidden Markov Models or similar machine-learning approaches, which imply the optimization of a high number of parameters. The present paper presents a novel method for the classification of coding and non-coding regions in prokaryotic genomes, based on a suitably defined compression index of a DNA sequence. The main features of this new method are the non-parametric logic and the costruction of a dictionary of words extracted from the sequences. These dictionaries can be very useful to perform further analyses on the genomic sequences themselves. The proposed approach has been applied on some prokaryotic complete genomes, obtaining optimal scores of correctly recognized coding and non-coding regions. Several false-positive and false-negative cases have been investigated in detail, which have revealed that this approach can fail in the presence of highly structured coding regions (e.g., genes coding for modular proteins) or quasi-random non-coding regions (e.g., regions hosting non-functional fragments of copies of functional genes; regions hosting promoters or other protein-binding sequences). We perform an overall comparison with other gene-finder software, since at this step we are not interested in building another gene-finder system, but only in exploring the possibility of the suggested approach.     

Prokaryotic gene finding based on physicochemical characteristics of codons calculated from molecular dynamics simulations

An ab initio model for gene prediction in prokaryotic genomes is proposed based on physicochemical characteristics of codons calculated from molecular dynamics (MD) simulations. The model requires a specification of three calculated quantities for each codon: the double-helical trinucleotide base pairing energy, the base pair stacking energy, and an index of the propensity of a codon for protein-nucleic acid interactions. The base pairing and stacking energies for each codon are obtained from recently reported MD simulations on all unique tetranucleotide steps, and the third parameter is assigned based on the conjugate rule previously proposed to account for the wobble hypothesis with respect to degeneracies in the genetic code. The third interaction propensity parameter values correlate well with ab initio MD calculated solvation energies and flexibility of codon sequences as well as codon usage in genes and amino acid composition frequencies in ∼175,000 protein sequences in the Swissprot database. Assignment of these three parameters for each codon enables the calculation of the magnitude and orientation of a cumulative three-dimensional vector for a DNA sequence of any length in each of the six genomic reading frames. Analysis of 372 genomes comprising ∼350,000 genes shows that the orientations of the gene and nongene vectors are well differentiated and make a clear distinction feasible between genic and nongenic sequences at a level equivalent to or better than currently available knowledge-based models trained on the basis of empirical data, presenting a strong support for the possibility of a unique and useful physicochemical characterization of DNA sequences from codons to genomes.     

A comparative genomic method for computational identification of prokaryotic translation initiation sites

The ever growing number of completely sequenced prokaryotic genomes facilitates cross-species comparisons by genomic annotation algorithms. This paper introduces a new probabilistic framework for comparative genomic analysis and demonstrates its utility in the context of improving the accuracy of prokaryotic gene start site detection. Our frame work employs a product hidden Markov model (PROD-HMM) with state architecture to model the species-specific trinucleotide frequency patterns in sequences immediately upstream and downstream of a translation start site and to detect the contrasting non-synonymous (amino acid changing) and synonymous (silent) substitution rates that differentiate prokaryotic coding from intergenic regions. Depending on the intricacy of the features modeled by the hidden state architecture, intergenic, regulatory, promoter and coding regions can be delimited by this method. The new system is evaluated using a preliminary set of orthologous Pyrococcus gene pairs, for which it demonstrates an improved accuracy of detection. Its robustness is confirmed by analysis with cross-validation of an experimentally verified set of Escherichia coli K-12 and Salmonella thyphimurium LT2 orthologs. The novel architecture has a number of attractive features that distinguish it from previous comparative models such as pair-HMMs.     

Genomic GC content prediction in prokaryotes from a sample of genes

GC level is a key feature in prokaryotic genomes. Widely employed in evolutionary studies, new insights appear however limited because of the relatively low number of characterized genomes. Since public databases mainly comprise several hundreds of prokaryotes with a low number of sequences per genome, a reliable prediction method based on available sequences may be useful for studies that need a trustworthy estimation of whole genomic GC. As the analysis of completely sequenced genomes shows a great variability in distributional shapes, it is of interest to compare different estimators. Our analysis shows that the mean of GC values of a random sample of genes is a reasonable estimator, based on simplicity of the calculation and overall performance. However, usually sequences come from a process that cannot be considered as random sampling. When we analyzed two introduced sources of bias (gene length and protein functional categories) we were able to detect an additional bias in the estimation for some cases, although the precision was not affected. We conclude that the mean genic GC level of a sample of 10 genes is a reliable estimator of genomic GC content, showing comparable accuracy with many widely employed experimental methods.     

Automatic prediction of non-coding RNA genes in prokaryotes based on compositional statistics

Although non-coding RNA (ncRNA) genes do not encode proteins, they play vital roles in cells by producing functionally important RNAs. In this paper, we present a novel method for predicting ncRNA genes based on compositional features extracted directly from gene sequences. Our method consists of two Support Vector Machine (SVM) models--Codon model which uses codon usage features derived from ncRNA genes and protein-coding genes and Kmer model which utilizes features of nucleotide and dinucleotide frequency extracted respectively from ncRNA genes and randomly chosen genome sequences. The 10-fold cross-validation accuracy for the two models is found to be 92% and 91%, respectively. Thus, we could make an automatic prediction of ncRNA genes in one genome without manual filtration of protein-coding genes. After applying our method in Sulfolobus solfataricus genome, 25 prediction results have been generated according to 25 cut-off pairs. We have also applied the approach in E. coli and found our results comparable to those of previous studies. In general, our method enables automatic identification of ncRNA genes in newly sequenced prokaryotic genomes.