Accurate Gene Prediction Research Articles

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.

Improvement in the Accuracy of Gene Prediction in Human cDNA Sequences

8 30-38%, 38-42%, 42-46%, 46-50%, 50-54%, 54-58%, 58-62%, 62-70% Ten matrices 10 30-34%, 34-38%, 38-42%, 42-46%, 46-50%, 50-54%, 54-58%, 58-62%, 62-66%, 66-70%

Leveraging the mouse genome for gene prediction in human: from whole-genome shotgun reads to a global synteny map.

The availability of draft sequences for both the mouse and human genomes makes it possible, for the first time, to annotate whole mammalian genomes using comparative methods. TWINSCAN is a gene-prediction system that combines the methods of single-genome predictors like GENSCAN with information derived from genome comparison, thereby improving accuracy. Because TWINSCAN uses genomic sequence only, it is less biased toward highly and/or ubiquitously expressed genes than GENEWISE, GENOMESCAN, and other methods based on evidence derived from transcripts. We show that TWINSCAN improves gene prediction in human using intermediate products from various stages of the sequencing and analysis of the mouse genome, from low-redundancy, whole-genome shotgun reads to the draft assembly and the synteny map. TWINSCAN improves on the prior state of the art even when alignments from only 1X coverage of the mouse genome are available. Gene prediction accuracy improves steadily from 1X through 3X, more slowly from 3X to 4X, and relatively little thereafter. The assembly and the synteny map greatly speed the computations, however. Our human annotation using the mouse assembly is conservative, predicting only 25,622 genes, and appears to be one of the best de novo annotations of the human genome to date.

Re-annotation of genome microbial CoDing-Sequences: finding new genes and inaccurately annotated genes

BackgroundAnalysis of any newly sequenced bacterial genome starts with the identification of protein-coding genes. Despite the accumulation of multiple complete genome sequences, which provide useful comparisons with close relatives among other organisms during the annotation process, accurate gene prediction remains quite difficult. A major reason for this situation is that genes are tightly packed in prokaryotes, resulting in frequent overlap. Thus, detection of translation initiation sites and/or selection of the correct coding regions remain difficult unless appropriate biological knowledge (about the structure of a gene) is imbedded in the approach.ResultsWe have developed a new program that automatically identifies biologically significant candidate genes in a bacterial genome. Twenty-six complete prokaryotic genomes were analyzed using this tool, and the accuracy of gene finding was assessed by comparison with existing annotations. This analysis revealed that, despite the enormous effort of genome program annotators, a small but not negligible number of genes annotated within the framework of sequencing projects are likely to be partially inaccurate or plainly wrong. Moreover, the analysis of several putative new genes shows that, as expected, many short genes have escaped annotation. In most cases, these new genes revealed frameshifts that could be either artifacts or genuine frameshifts. Some entirely unexpected new genes have also been identified. This allowed us to get a more complete picture of prokaryotic genomes. The results of this procedure are progressively integrated into the SWISS-PROT reference databank.ConclusionsThe results described in the present study show that our procedure is very satisfactory in terms of gene finding accuracy. Except in few cases, discrepancies between our results and annotations provided by individual authors can be accounted for by the nature of each annotation process or by specific characteristics of some genomes. This stresses that close cooperation between scientists, regular update and curation of the findings in databases are clearly required to reduce the level of errors in genome annotation (and hence in reducing the unfortunate spreading of errors through centralized data libraries).

A probabilistic method for identifying start codons in bacterial genomes.

As the pace of genome sequencing has accelerated, the need for highly accurate gene prediction systems has grown. Computational systems for identifying genes in prokaryotic genomes have sensitivities of 98-99% or higher (Delcher et al., Nucleic Acids Res., 27, 4636-4641, 1999). These accuracy figures are calculated by comparing the locations of verified stop codons to the predictions. Determining the accuracy of start codon prediction is more problematic, however, due to the relatively small number of start sites that have been confirmed by independent, non-computational methods. Nonetheless, the accuracy of gene finders at predicting the exact gene boundaries at both the 5' and 3' ends of genes is of critical importance for microbial genome annotation, especially in light of the important signaling information that is sometimes found on the 5' end of a protein coding region. In this paper we propose a probabilistic method to improve the accuracy of gene identification systems at finding precise translation start sites. The new system, RBSfinder, is tested on a validated set of genes from Escherichia coli, for which it improves the accuracy of start site locations predicted by computational gene finding systems from the range 67-77% to 90% correct.

Definition of the gene content of the human genome: the need for deep experimental verification

Based on the analysis of the drafts of the human genome sequence, it is being speculated that our species may possess an unexpectedly low number of genes. The quality of the drafts, the impossibility of accurate gene prediction and the lack of sufficient transcript sequence data, however, render such speculations very premature. The complexity of human gene structure requires additional and extensive experimental verification of transcripts that may result in major revisions of these early estimates of the number of human genes.

An assessment of gene prediction accuracy in large DNA sequences.

One of the first useful products from the human genome will be a set of predicted genes. Besides its intrinsic scientific interest, the accuracy and completeness of this data set is of considerable importance for human health and medicine. Though progress has been made on computational gene identification in terms of both methods and accuracy evaluation measures, most of the sequence sets in which the programs are tested are short genomic sequences, and there is concern that these accuracy measures may not extrapolate well to larger, more challenging data sets. Given the absence of experimentally verified large genomic data sets, we constructed a semiartificial test set comprising a number of short single-gene genomic sequences with randomly generated intergenic regions. This test set, which should still present an easier problem than real human genomic sequence, mimics the approximately 200kb long BACs being sequenced. In our experiments with these longer genomic sequences, the accuracy of GENSCAN, one of the most accurate ab initio gene prediction programs, dropped significantly, although its sensitivity remained high. Conversely, the accuracy of similarity-based programs, such as GENEWISE, PROCRUSTES, and BLASTX was not affected significantly by the presence of random intergenic sequence, but depended on the strength of the similarity to the protein homolog. As expected, the accuracy dropped if the models were built using more distant homologs, and we were able to quantitatively estimate this decline. However, the specificities of these techniques are still rather good even when the similarity is weak, which is a desirable characteristic for driving expensive follow-up experiments. Our experiments suggest that though gene prediction will improve with every new protein that is discovered and through improvements in the current set of tools, we still have a long way to go before we can decipher the precise exonic structure of every gene in the human genome using purely computational methodology.

Using GeneWise in the Drosophila annotation experiment.

The GeneWise method for combining gene prediction and homology searches was applied to the 2.9-Mb region from Drosophila melanogaster. The results from the Genome Annotation Assessment Project (GASP) showed that GeneWise provided reasonably accurate gene predictions. Further investigation indicates that many of the incorrect gene predictions from GeneWise were due to transposons with valid protein-coding genes and the remaining cases are pseudogenes or possible annotation oversights.

Ab initio Gene Finding in Drosophila Genomic DNA

Ab initio gene identification in the genomic sequence of Drosophila melanogaster was obtained using (human gene predictor) and Fgenesh programs that have organism-specific parameters for human, Drosophila, plants, yeast, and nematode. We did not use information about cDNA/EST in most predictions to model a real situation for finding new genes because information about complete cDNA is often absent or based on very small partial fragments. We investigated the accuracy of gene prediction on different levels and designed several schemes to predict an unambiguous set of genes (annotation CGG1), a set of reliable exons (annotation CGG2), and the most complete set of exons (annotation CGG3). For 49 genes, protein products of which have clear homologs in protein databases, predictions were recomputed by Fgenesh+ program. The first annotation serves as the optimal computational description of new sequence to be presented in a database. Reliable exons from the second annotation serve as good candidates for selecting the PCR primers for experimental work for gene structure verification. Our results shows that we can identify approximately 90% of coding nucleotides with 20% false positives. At the exon level we accurately predicted 65% of exons and 89% including overlapping exons with 49% false positives. Optimizing accuracy of prediction, we designed a gene identification scheme using Fgenesh, which provided sensitivity (Sn) = 98% and specificity (Sp) = 86% at the base level, Sn = 81% (97% including overlapping exons) and Sp = 58% at the exon level and Sn = 72% and Sp = 39% at the gene level (estimating sensitivity on std1 set and specificity on std3 set). In general, these results showed that computational gene prediction can be a reliable tool for annotating new genomic sequences, giving accurate information on 90% of coding sequences with 14% false positives. However, exact gene prediction (especially at the gene level) needs additional improvement using gene prediction algorithms. The program was also tested for predicting genes of human Chromosome 22 (the last variant of Fgenesh can analyze the whole chromosome sequence). This analysis has demonstrated that the 88% of manually annotated exons in Chromosome 22 were among the ab initio predicted exons. The suite of gene identification programs is available through the WWW server of Computational Genomics Group at http://genomic.sanger.ac.uk/gf. html.

Optimal spliced alignment of homologous cDNA to a genomic DNA template.

Supplementary cDNA or EST evidence is often decisive for discriminating between alternative gene predictions derived from computational sequence inspection by any of a number of requisite programs. Without additional experimental effort, this approach must rely on the occurrence of cognate ESTs for the gene under consideration in available, generally incomplete, EST collections for the given species. In some cases, particular exon assignments can be supported by sequence matching even if the cDNA or EST is produced from non-cognate genomic DNA, including different loci of a gene family or homologous loci from different species. However, marginally significant sequence matching alone can also be misleading. We sought to develop an algorithm that would simultaneously score for predicted intrinsic splice site strength and sequence matching between the genomic DNA template and a related cDNA or EST. In this case, weakly predicted splice sites may be chosen for the optimal scoring spliced alignment on the basis of surrounding sequence matching. Strongly predicted splice sites will enter the optimal spliced alignment even without strong sequence matching. We designed a novel algorithm that produces the optimal spliced alignment of a genomic DNA with a cDNA or EST based on scoring for both sequence matching and intrinsic splice site strength. By example, we demonstrate that this combined approach appears to improve gene prediction accuracy compared with current methods that rely only on either search by content and signal or on sequence similarity. The algorithm is available as a C subroutine and is implemented in the SplicePredictor and GeneSeqer programs. The source code is available via anonymous ftp from ftp. zmdb.iastate.edu. Both programs are also implemented as a Web service at http://gremlin1.zool.iastate.edu/cgi-bin/s p.cgiand http://gremlin1.zool.iastate.edu/cgi-bin/g s.cgi, respectively. vbrendel@iastate.edu

Gene prediction and gene classes in Arabidopsis thaliana

Gene prediction methods for eukaryotic genomes still are not fully satisfying. One way to improve gene prediction accuracy, proven to be relevant for prokaryotes, is to consider more than one model of genes. Thus, we used our classification of Arabidopsis thaliana genes in two classes (CU 1 and CU 2), previously delineated according to statistical features, in the GeneMark gene identification program. For each gene class, as well as for the two classes combined, a Markov model was developed (respectively, GM-CU 1, GM-CU 2 and GM-all) and then used on a test set of 168 genes to compare their respective efficiency. We concluded from this analysis that GM-CU 1 is more sensitive than GM-CU 2 which seems to be more specific to a gene type. Besides, GM-all does not give better results than GM-CU 1 and combining results from GM-CU 1 and GM-CU 2 greatly improve prediction efficiency in comparison with predictions made with GM-all only. Thus, this work confirms the necessity to consider more than one gene model for gene prediction in eukaryotic genomes, and to look for gene classes in order to build these models.

Bacterial start site prediction.

With the growing number of completely sequenced bacterial genes, accurate gene prediction in bacterial genomes remains an important problem. Although the existing tools predict genes in bacterial genomes with high overall accuracy, their ability to pinpoint the translation start site remains unsatisfactory. In this paper, we present a novel approach to bacterial start site prediction that takes into account multiple features of a potential start site, viz., ribosome binding site (RBS) binding energy, distance of the RBS from the start codon, distance from the beginning of the maximal ORF to the start codon, the start codon itself and the coding/non-coding potential around the start site. Mixed integer programing was used to optimize the discriminatory system. The accuracy of this approach is up to 90%, compared to 70%, using the most common tools in fully automated mode (that is, without expert human post-processing of results). The approach is evaluated using Bacillus subtilis, Escherichia coli and Pyrococcus furiosus. These three genomes cover a broad spectrum of bacterial genomes, since B.subtilis is a Gram-positive bacterium, E.coli is a Gram-negative bacterium and P. furiosus is an archaebacterium. A significant problem is generating a set of 'true' start sites for algorithm training, in the absence of experimental work. We found that sequence conservation between P. furiosus and the related Pyrococcus horikoshii clearly delimited the gene start in many cases, providing a sufficient training set.

SST Versus EST in Gene Recognition

The expressed sequence tag (EST) data provide a powerful tool for identification of transcribed DNA sequences. However, as EST are relatively short, many exons are poorly covered by EST, thus reducing the utility of EST data. Recently, signature sequence tag (SST) fingerprints were proposed as an alternative to EST fingerprints. Given a fingerprint set of probes, SST of a clone is a subset of probes from the fingerprint set that hybridize with the clone. We demonstrate that besides being a powerful technique for screening cDNA libraries, SST technology provides for very accurate gene predictions. Even with a small fingerprint set (600-800 probes), SST-based gene recognition outperforms many conventional and EST-based methods. The increase in the size of the fingerprint set to 1500 probes provides almost perfect gene recognition. Even more importantly, SST-based gene predictions miss very few exons and, therefore, provide an opportunity to bypass the cDNA sequencing step on the way from finished genomic sequence to mutation detection in gene-hunting projects. Because SST data can be obtained in a highly parallel and inexpensive way, SST technology has a potential of complementing EST technology for gene hunting.

Combining diverse evidence for gene recognition in completely sequenced bacterial genomes

Analysis of a newly sequenced bacterial genome starts with identification of protein-coding genes. Functional assignment of proteins requires the exact knowledge of protein N-termini. We present a new program ORPHEUS that identifies candidate genes and accurately predicts gene starts. The analysis starts with a database similarity search and identification of reliable gene fragments. The latter are used to derive statistical characteristics of protein-coding regions and ribosome-binding sites and to predict the complete set of genes in the analyzed genome. In a test on Bacillus subtilis and Escherichia coli genomes, the program correctly identified 93.3% (resp. 96.3%) of experimentally annotated genes longer than 100 codons described in the PIR-International database, and for these genes 96.3% (83.9%) of starts were predicted exactly. Furthermore, 98.9% (99.1%) of genes longer than 100 codons annotated in GenBank were found, and 92.9% (75.7%) of predicted starts coincided with the feature table description. Finally, for the complete gene complements of B.subtilis and E.coli, including genes shorter than 100 codons, gene prediction accuracy was 88.9 and 87.1%, respectively, with 94.2 and 76.7% starts coinciding with the existing annotation.

Combining diverse evidence for gene recognition in completely sequenced bacterial genomes.

Analysis of a newly sequenced bacterial genome starts with identification of protein-coding genes. Functional assignment of proteins requires the exact knowledge of protein N-termini. We present a new program ORPHEUS that identifies candidate genes and accurately predicts gene starts. The analysis starts with a database similarity search and identification of reliable gene fragments. The latter are used to derive statistical characteristics of protein-coding regions and ribosome-binding sites and to predict the complete set of genes in the analyzed genome. In a test on Bacillus subtilis and Escherichia coli genomes, the program correctly identified 93.3% (resp. 96.3%) of experimentally annotated genes longer than 100 codons described in the PIR-International database, and for these genes 96.3% (83.9%) of starts were predicted exactly. Furthermore, 98.9% (99.1%) of genes longer than 100 codons annotated in GenBank were found, and 92.9% (75.7%) of predicted starts coincided with the feature table description. Finally, for the complete gene complements of B.subtilis and E.coli , including genes shorter than 100 codons, gene prediction accuracy was 88.9 and 87.1%, respectively, with 94.2 and 76.7% starts coinciding with the existing annotation.

EbEST: an automated tool using expressed sequence tags to delineate gene structure.

Large numbers of expressed sequence tags (ESTs) continue to fill public and private databases with partial cDNA sequences. However, using this huge amount of ESTs to facilitate gene finding in genomic sequence imposes a challenge, especially to wet-lab scientists who often have limited computing resources. In an effort to consolidate the information hidden in the vast number of ESTs into a readable and manageable format, we have developed EbEST-a program that automates the process of using ESTs to help delineate gene structure in long stretches of genomic sequence. The EbEST program consists of three functional modules-the first module separates homologous ESTs into clusters and identifies the most informative ESTs within each cluster; the second module uses the informative ESTs to perform gapped alignment and to predict the exon-intron boundary; and the third module generates text file and graphic outputs that illustrate the orientation, exonic structure, and untranslated regions (UTRs) of putative genes in the genomic sequence being analyzed. Evaluation of EbEST with 176 human genes from the ALLSEQ set indicated that it performed in-line with several existing gene finding programs, but was more tolerant to sequencing errors. Furthermore, when EbEST was challenged with query sequences that harbor more than one gene, it suffered only a slight drop in performance, whereas the performance of the other programs evaluated decreased more. EbEST may be used as a stand-alone tool to annotate human genomic sequences with EST-derived gene elements, or can be used in conjunction with computational gene-recognition programs to increase the accuracy of gene prediction. [EbBEST is available at http://EbEST.ifrc.mcw.edu]

Las Vegas algorithms for gene recognition: suboptimal and error-tolerant spliced alignment.

Recently, Gelfand, Mironov and Pevzner (1996) proposed a spliced alignment approach to gene recognition that provides 99% accurate recognition of human genes if a related mammalian protein is available. However, even 99% accurate gene predictions are insufficient for automated sequence annotation in large-scale sequencing projects and therefore have to be complemented by experimental gene verification. One hundred percent accurate gene predictions would lead to a substantial reduction of experimental work on gene identification. Our goal is to develop an algorithm that either predicts an exon assembly with accuracy sufficient for sequence annotation or warns a biologist that the accuracy of a prediction is insufficient and further experimental work is required. We study suboptimal and error-tolerant spliced alignment problems as the first steps towards such an algorithm, and report an algorithm which provides 100% accurate recognition of human genes in 37% of cases (if a related mammalian protein is available). In 52% of genes, the algorithm predicts at least one exon with 100% accuracy.