Accurate Gene Prediction Research Articles

Polyphase filtering with variable mapping rule in protein coding region prediction

Genomic researches are concerned with the study of genomes of organisms. It has become a challenge to the researchers to identify the segments within the DNA sequence that involved in protein synthesis and called coding region of gene. The methods are generally used to identify the segment that relies on period-3 property of genes. This period-3 property easily can be identified by digital signal processing with great accuracy. Prior to DSP application in gene prediction a conversion rule is required which converts symbolic DNA (ATCGTC…) sequence into numerical representations. Accuracy of gene prediction depends on mapping rule. The effectiveness of mapping rule depends on the application area of genomics. Some mapping rule works well in gene prediction may not performed good in genetic disease prediction. Most of the available conversion rules are fixed mapping technique. In this paper a new conversion rule is proposed prior to DSP application and a polyphase filter is used to suppress the noise in the DNA spectrum. The performance of the proposed mapping is compared with existing mapping and also the performance of the polyphase filtering method is compared with existing filtering methods in terms of signal to noise ratio (SNR) and location accuracy.

GeneValidator: identify problems with protein-coding gene predictions.

Summary: Genomes of emerging model organisms are now being sequenced at very low cost. However, obtaining accurate gene predictions remains challenging: even the best gene prediction algorithms make substantial errors and can jeopardize subsequent analyses. Therefore, many predicted genes must be time-consumingly visually inspected and manually curated. We developed GeneValidator (GV) to automatically identify problematic gene predictions and to aid manual curation. For each gene, GV performs multiple analyses based on comparisons to gene sequences from large databases. The resulting report identifies problematic gene predictions and includes extensive statistics and graphs for each prediction to guide manual curation efforts. GV thus accelerates and enhances the work of biocurators and researchers who need accurate gene predictions from newly sequenced genomes.Availability and implementation: GV can be used through a web interface or in the command-line. GV is open-source (AGPL), available at https://wurmlab.github.io/tools/genevalidator.Contact: y.wurm@qmul.ac.ukSupplementary information: Supplementary data are available at Bioinformatics online.

CodingQuarry: highly accurate hidden Markov model gene prediction in fungal genomes using RNA-seq transcripts.

BackgroundThe impact of gene annotation quality on functional and comparative genomics makes gene prediction an important process, particularly in non-model species, including many fungi. Sets of homologous protein sequences are rarely complete with respect to the fungal species of interest and are often small or unreliable, especially when closely related species have not been sequenced or annotated in detail. In these cases, protein homology-based evidence fails to correctly annotate many genes, or significantly improve ab initio predictions. Generalised hidden Markov models (GHMM) have proven to be invaluable tools in gene annotation and, recently, RNA-seq has emerged as a cost-effective means to significantly improve the quality of automated gene annotation. As these methods do not require sets of homologous proteins, improving gene prediction from these resources is of benefit to fungal researchers. While many pipelines now incorporate RNA-seq data in training GHMMs, there has been relatively little investigation into additionally combining RNA-seq data at the point of prediction, and room for improvement in this area motivates this study.ResultsCodingQuarry is a highly accurate, self-training GHMM fungal gene predictor designed to work with assembled, aligned RNA-seq transcripts. RNA-seq data informs annotations both during gene-model training and in prediction. Our approach capitalises on the high quality of fungal transcript assemblies by incorporating predictions made directly from transcript sequences. Correct predictions are made despite transcript assembly problems, including those caused by overlap between the transcripts of adjacent gene loci.Stringent benchmarking against high-confidence annotation subsets showed CodingQuarry predicted 91.3% of Schizosaccharomyces pombe genes and 90.4% of Saccharomyces cerevisiae genes perfectly. These results are 4-5% better than those of AUGUSTUS, the next best performing RNA-seq driven gene predictor tested. Comparisons against whole genome Sc. pombe and S. cerevisiae annotations further substantiate a 4-5% improvement in the number of correctly predicted genes.ConclusionsWe demonstrate the success of a novel method of incorporating RNA-seq data into GHMM fungal gene prediction. This shows that a high quality annotation can be achieved without relying on protein homology or a training set of genes. CodingQuarry is freely available (https://sourceforge.net/projects/codingquarry/), and suitable for incorporation into genome annotation pipelines.

English

The objective of the study was to determine the codon usage bias in the porcine genome and decipher its determinants. To investigate the underlying mechanisms of codon bias, the coding sequence (CDS) from the swine reference sequence (ssc10.2) was extracted using Biomart. An in house built Perl script was used to derive various genomic traits and codon indices. Analysis was done using R statistical package, and correlations and multivariate regressions were performed. We report the existence of codon usage bias that might suggest existence of weak translational selection. The codon bias is feebly related to nucleotide composition (GC%, GC3, CDS length). This study can be explored for designing degenerate primers, necessitate selecting appropriate hosts expression systems to manipulate the expression of target genes in vivo or in vitro and improve the accuracy of gene prediction from genomic sequences thus maximizing the effectiveness of genetic manipulations in synthetic biology. Key words: Coding sequence, synomynous codons, selection, translational mutation, pig genome

Integration of mapped RNA-Seq reads into automatic training of eukaryotic gene finding algorithm.

We present a new approach to automatic training of a eukaryotic ab initio gene finding algorithm. With the advent of Next-Generation Sequencing, automatic training has become paramount, allowing genome annotation pipelines to keep pace with the speed of genome sequencing. Earlier we developed GeneMark-ES, currently the only gene finding algorithm for eukaryotic genomes that performs automatic training in unsupervised ab initio mode. The new algorithm, GeneMark-ET augments GeneMark-ES with a novel method that integrates RNA-Seq read alignments into the self-training procedure. Use of ‘assembled’ RNA-Seq transcripts is far from trivial; significant error rate of assembly was revealed in recent assessments. We demonstrated in computational experiments that the proposed method of incorporation of ‘unassembled’ RNA-Seq reads improves the accuracy of gene prediction; particularly, for the 1.3 GB genome of Aedes aegypti the mean value of prediction Sensitivity and Specificity at the gene level increased over GeneMark-ES by 24.5%. In the current surge of genomic data when the need for accurate sequence annotation is higher than ever, GeneMark-ET will be a valuable addition to the narrow arsenal of automatic gene prediction tools.

A new computational strategy for predicting essential genes

BackgroundDetermination of the minimum gene set for cellular life is one of the central goals in biology. Genome-wide essential gene identification has progressed rapidly in certain bacterial species; however, it remains difficult to achieve in most eukaryotic species. Several computational models have recently been developed to integrate gene features and used as alternatives to transfer gene essentiality annotations between organisms.ResultsWe first collected features that were widely used by previous predictive models and assessed the relationships between gene features and gene essentiality using a stepwise regression model. We found two issues that could significantly reduce model accuracy: (i) the effect of multicollinearity among gene features and (ii) the diverse and even contrasting correlations between gene features and gene essentiality existing within and among different species. To address these issues, we developed a novel model called feature-based weighted Naïve Bayes model (FWM), which is based on Naïve Bayes classifiers, logistic regression, and genetic algorithm. The proposed model assesses features and filters out the effects of multicollinearity and diversity. The performance of FWM was compared with other popular models, such as support vector machine, Naïve Bayes model, and logistic regression model, by applying FWM to reciprocally predict essential genes among and within 21 species. Our results showed that FWM significantly improves the accuracy and robustness of essential gene prediction.ConclusionsFWM can remarkably improve the accuracy of essential gene prediction and may be used as an alternative method for other classification work. This method can contribute substantially to the knowledge of the minimum gene sets required for living organisms and the discovery of new drug targets.

Complexity-Reduced Tumor Classification System using Microarray Gene Expression Dataset

e classification of cancer based on gene expression data is the advancements in DNA Microarray technology and genome sequencing. The important feature is to predict the genes for various diagnosis purposes using such micro-array gene expression dataset and also the gene expressions that are relevant to a particular type of genes. Lot of research works have been carried out to produce a better solution to improve the prediction accuracy of cancer gene prediction. But the analysis results are not up to the convincing level artificial intelligence is exploited to improve the prediction accuracy meanwhile state-of-the art insists necessary enhancements which are essential in the classification module instead in the features module. The enhanced classifier called Principal component analysis used in latter researches is used in this work for the performances comparison of the classifier as a conventional prediction methodology. This work intends to apply the developed classifier and dominant gene prediction methodology to predict extensive set of cancer expression datasets. The experimental study will be carried out by considering the techniques on CNS tumor, colon tumor and ALL_AML Leukemia. The prediction performance of the proposed methodology will be compared against the conventional prediction methodologies and the results will be validated extensively. The method will be implemented in the working platform of MATLAB and the performance will be analysed.

Effects of Sample Size on Differential Gene Expression, Rank Order and Prediction Accuracy of a Gene Signature

Top differentially expressed gene lists are often inconsistent between studies and it has been suggested that small sample sizes contribute to lack of reproducibility and poor prediction accuracy in discriminative models. We considered sex differences (69♂, 65♀) in 134 human skeletal muscle biopsies using DNA microarray. The full dataset and subsamples (n = 10 (5♂, 5♀) to n = 120 (60♂, 60♀)) thereof were used to assess the effect of sample size on the differential expression of single genes, gene rank order and prediction accuracy. Using our full dataset (n = 134), we identified 717 differentially expressed transcripts (p<0.0001) and we were able predict sex with ∼90% accuracy, both within our dataset and on external datasets. Both p-values and rank order of top differentially expressed genes became more variable using smaller subsamples. For example, at n = 10 (5♂, 5♀), no gene was considered differentially expressed at p<0.0001 and prediction accuracy was ∼50% (no better than chance). We found that sample size clearly affects microarray analysis results; small sample sizes result in unstable gene lists and poor prediction accuracy. We anticipate this will apply to other phenotypes, in addition to sex.

Next generation genome annotation with mGene.ngs

An increasingly large number of novel genomes is being sequenced and the task of automatic genome annotation has never been more important. The current revolution in sequencing technologies also allows us to obtain a detailed picture of the whole complement of expressed RNA transcripts. We have developed a novel de novo gene finding system mGene.ngs that combines the benefits of accurate ab initio gene finding with the rich information obtained in RNA sequencing (RNA-seq) experiments. The system is based on the recently developed accurate gene finding system mGene [1], which employs state-of-the-art prediction techniques and which has been shown to perform very well compared to established gene finding systems [2]. In contrast to many HMM-based gene finders, mGene has the conceptual advantage of being very flexible in terms of incorporating heterogeneous input data. The employed inference techniques can exploit the transcriptome information already at the learning stage to appropriately adapt to the relevance of the different evidences. We show that these advantages can be translated into more accurate gene predictions. Moreover, we developed extensions of mGene.ngs to predict and quantify alternative RNA transcripts. To provide de novo genome annotations based on RNA-seq experiments, we first construct a preliminary, highly specific gene set for genes that are well-covered with RNA-seq reads. In a second step, we train predictors for genomic signals on the preliminary gene set. In the third step we train mGene.ngs, using the preliminary gene models while taking advantage of the RNA-seq read coverage and genomic signal predictions. We illustrate the power of our approach for the C. elegans genome and 50M paired-end RNA-seq reads (Illumina; 76nt). Figure 1 shows transcript level evaluation results for all annotated genes (WS200) as a function of the expression level. The ab initio mGene-based system (blue) trained on the annotation achieves an average transcript-level F-score of 49.9%. We achieve a slightly better performance (51.8%) for the de novo annotation system (green) using RNA-seq reads, but without considering the existing genome annotation. If we use the RNA-seq reads and train on the existing annotation (red), we achieve 57.6%, and can therefore take advantage of the previous annotation. We find it remarkable that for medium to high expressed genes the de novo gene predictions are as similar to the genome annotation as the predictions of the system, that has seen parts of the annotation in training. Comparing these results to predictions from the recently published method cufflinks [3] (black) reveals that cufflinks seems not to be able to appropriately adapt to the RNA-seq data at hand. Investigating the contribution of individual features we found that spliced read alignments suggesting introns help most to increase the gene prediction performance; 91.6% of the achieved total improvement is due to spliced read alignments. The read coverage alone is much less informative and only leads to improvements similar to the ones achieved with transcriptome tiling arrays. We employed the developed annotation strategy for the re-annotation of the C. briggsae genome, for which only few transcriptome sequences are available yet. We can show that the new annotation is considerably more accurate than previous ones and additionally includes alternative RNA isoforms. mGene.ngs will be released as open source software on http://mgene.org and is already available as Galaxy-based web-service at http://galaxy.fml.mpg.de. * Correspondence: jonas.behr@tuebingen.mpg.de Friedrich Miescher Laboratory of the Max Planck Society, Tubingen, Germany Full list of author information is available at the end of the article Behr et al. BMC Bioinformatics 2010, 11(Suppl 10):O8 http://www.biomedcentral.com/1471-2105/11/S10/O8

VIGOR, an annotation program for small viral genomes.

BackgroundThe decrease in cost for sequencing and improvement in technologies has made it easier and more common for the re-sequencing of large genomes as well as parallel sequencing of small genomes. It is possible to completely sequence a small genome within days and this increases the number of publicly available genomes. Among the types of genomes being rapidly sequenced are those of microbial and viral genomes responsible for infectious diseases. However, accurate gene prediction is a challenge that persists for decoding a newly sequenced genome. Therefore, accurate and efficient gene prediction programs are highly desired for rapid and cost effective surveillance of RNA viruses through full genome sequencing.ResultsWe have developed VIGOR (Viral Genome ORF Reader), a web application tool for gene prediction in influenza virus, rotavirus, rhinovirus and coronavirus subtypes. VIGOR detects protein coding regions based on sequence similarity searches and can accurately detect genome specific features such as frame shifts, overlapping genes, embedded genes, and can predict mature peptides within the context of a single polypeptide open reading frame. Genotyping capability for influenza and rotavirus is built into the program. We compared VIGOR to previously described gene prediction programs, ZCURVE_V, GeneMarkS and FLAN. The specificity and sensitivity of VIGOR are greater than 99% for the RNA viral genomes tested.ConclusionsVIGOR is a user friendly web-based genome annotation program for five different viral agents, influenza, rotavirus, rhinovirus, coronavirus and SARS coronavirus. This is the first gene prediction program for rotavirus and rhinovirus for public access. VIGOR is able to accurately predict protein coding genes for the above five viral types and has the capability to assign function to the predicted open reading frames and genotype influenza virus. The prediction software was designed for performing high throughput annotation and closure validation in a post-sequencing production pipeline.

MGene: Accurate SVM-based gene finding with an application to nematode genomes

We present a highly accurate gene-prediction system for eukaryotic genomes, called mGene. It combines in an unprecedented manner the flexibility of generalized hidden Markov models (gHMMs) with the predictive power of modern machine learning methods, such as Support Vector Machines (SVMs). Its excellent performance was proved in an objective competition based on the genome of the nematode Caenorhabditis elegans. Considering the average of sensitivity and specificity, the developmental version of mGene exhibited the best prediction performance on nucleotide, exon, and transcript level for ab initio and multiple-genome gene-prediction tasks. The fully developed version shows superior performance in 10 out of 12 evaluation criteria compared with the other participating gene finders, including Fgenesh++ and Augustus. An in-depth analysis of mGene's genome-wide predictions revealed that approximately 2200 predicted genes were not contained in the current genome annotation. Testing a subset of 57 of these genes by RT-PCR and sequencing, we confirmed expression for 24 (42%) of them. mGene missed 300 annotated genes, out of which 205 were unconfirmed. RT-PCR testing of 24 of these genes resulted in a success rate of merely 8%. These findings suggest that even the gene catalog of a well-studied organism such as C. elegans can be substantially improved by mGene's predictions. We also provide gene predictions for the four nematodes C. briggsae, C. brenneri, C. japonica, and C. remanei. Comparing the resulting proteomes among these organisms and to the known protein universe, we identified many species-specific gene inventions. In a quality assessment of several available annotations for these genomes, we find that mGene's predictions are most accurate.

Improving Gene-finding in Chlamydomonas reinhardtii:GreenGenie2

BackgroundThe availability of whole-genome sequences allows for the identification of the entire set of protein coding genes as well as their regulatory regions. This can be accomplished using multiple complementary methods that include ESTs, homology searches and ab initio gene predictions. Previously, the Genie gene-finding algorithm was trained on a small set of Chlamydomonas genes and shown to improve the accuracy of gene prediction in this species compared to other available programs. To improve ab initio gene finding in Chlamydomonas, we assemble a new training set consisting of over 2,300 cDNAs by assembling over 167,000 Chlamydomonas EST entries in GenBank using the EST assembly tool PASA.ResultsThe prediction accuracy of our cDNA-trained gene-finder, GreenGenie2, attains 83% sensitivity and 83% specificity for exons on short-sequence predictions. We predict about 12,000 genes in the version v3 Chlamydomonas genome assembly, most of which (78%) are either identical to or significantly overlap the published catalog of Chlamydomonas genes [1]. 22% of the published catalog is absent from the GreenGenie2 predictions; there is also a fraction (23%) of GreenGenie2 predictions that are absent from the published gene catalog. Randomly chosen gene models were tested by RT-PCR and most support the GreenGenie2 predictions.ConclusionThese data suggest that training with EST assemblies is highly effective and that GreenGenie2 is a valuable, complementary tool for predicting genes in Chlamydomonas reinhardtii.

The effect of sequencing errors on metagenomic gene prediction

BackgroundGene prediction is an essential step in the annotation of metagenomic sequencing reads. Since most metagenomic reads cannot be assembled into long contigs, specialized statistical gene prediction tools have been developed for short and anonymous DNA fragments, e.g. MetaGeneAnnotator and Orphelia. While conventional gene prediction methods have been subject to a benchmark study on real sequencing reads with typical errors, such a comparison has not been conducted for specialized tools, yet. Their gene prediction accuracy was mostly measured on error free DNA fragments.ResultsIn this study, Sanger and pyrosequencing reads were simulated on the basis of models that take all types of sequencing errors into account. All metagenomic gene prediction tools showed decreasing accuracy with increasing sequencing error rates. Performance results on an established metagenomic benchmark dataset are also reported. In addition, we demonstrate that ESTScan, a tool for sequencing error compensation in eukaryotic expressed sequence tags, outperforms some metagenomic gene prediction tools on reads with high error rates although it was not designed for the task at hand.ConclusionThis study fills an important gap in metagenomic gene prediction research. Specialized methods are evaluated and compared with respect to sequencing error robustness. Results indicate that the integration of error-compensating methods into metagenomic gene prediction tools would be beneficial to improve metagenome annotation quality.

Genome wide survey, discovery and evolution of repetitive elements in three Entamoeba species

BackgroundIdentification and mapping of repetitive elements is a key step for accurate gene prediction and overall structural annotation of genomes. During the assembly and annotation of three highly repetitive amoeba genomes, Entamoeba histolytica, Entamoeba dispar, and Entamoeba invadens, we performed comparative sequence analysis to identify and map all class I and class II transposable elements in their sequences.ResultsHere, we report the identification of two novel Entamoeba-specific repeats: ERE1 and ERE2; ERE1 is spread across the three genomes and associated with different repeats in a species-specific manner, while ERE2 is unique to E. histolytica. We also report the identification of two novel subfamilies of LINE and SINE retrotransposons in E. dispar and provide evidence for how the different LINE and SINE subfamilies evolved in these species. Additionally, we found a putative transposase-coding gene in E. histolytica and E. dispar related to the mariner transposon Hydargos from E. invadens. The distribution of transposable elements in these genomes is markedly skewed with a tendency of forming clusters. More than 70% of the three genomes have a repeat density below their corresponding average value indicating that transposable elements are not evenly distributed. We show that repeats and repeat-clusters are found at syntenic break points between E. histolytica and E. dispar and hence, could work as recombination hot spots promoting genome rearrangements.ConclusionThe mapping of all transposable elements found in these parasites shows that repeat coverage is up to three times higher than previously reported. LINE, ERE1 and mariner elements were present in the common ancestor to the three Entamoeba species while ERE2 was likely acquired by E. histolytica after its separation from E. dispar. We demonstrate that E. histolytica and E. dispar share their entire repertoire of LINE and SINE retrotransposons and that Eh_SINE3/Ed_SINE1 originated as a chimeric SINE from Eh/Ed_SINE2 and Eh_SINE1/Ed_SINE3. Our work shows that transposable elements are organized in clusters, frequently found at syntenic break points providing insights into their contribution to chromosome instability and therefore, to genomic variation and speciation in these parasites.

NGASP--the nematode genome annotation assessment project.

BackgroundWhile the C. elegans genome is extensively annotated, relatively little information is available for other Caenorhabditis species. The nematode genome annotation assessment project (nGASP) was launched to objectively assess the accuracy of protein-coding gene prediction software in C. elegans, and to apply this knowledge to the annotation of the genomes of four additional Caenorhabditis species and other nematodes. Seventeen groups worldwide participated in nGASP, and submitted 47 prediction sets across 10 Mb of the C. elegans genome. Predictions were compared to reference gene sets consisting of confirmed or manually curated gene models from WormBase.ResultsThe most accurate gene-finders were 'combiner' algorithms, which made use of transcript- and protein-alignments and multi-genome alignments, as well as gene predictions from other gene-finders. Gene-finders that used alignments of ESTs, mRNAs and proteins came in second. There was a tie for third place between gene-finders that used multi-genome alignments and ab initio gene-finders. The median gene level sensitivity of combiners was 78% and their specificity was 42%, which is nearly the same accuracy reported for combiners in the human genome. C. elegans genes with exons of unusual hexamer content, as well as those with unusually many exons, short exons, long introns, a weak translation start signal, weak splice sites, or poorly conserved orthologs posed the greatest difficulty for gene-finders.ConclusionThis experiment establishes a baseline of gene prediction accuracy in Caenorhabditis genomes, and has guided the choice of gene-finders for the annotation of newly sequenced genomes of Caenorhabditis and other nematode species. We have created new gene sets for C. briggsae, C. remanei, C. brenneri, C. japonica, and Brugia malayi using some of the best-performing gene-finders.

A new method to compute K-mer frequencies and its application to annotate large repetitive plant genomes

BackgroundThe challenges of accurate gene prediction and enumeration are further aggravated in large genomes that contain highly repetitive transposable elements (TEs). Yet TEs play a substantial role in genome evolution and are themselves an important subject of study. Repeat annotation, based on counting occurrences of k-mers, has been previously used to distinguish TEs from low-copy genic regions; but currently available software solutions are impractical due to high memory requirements or specialization for specific user-tasks.ResultsHere we introduce the Tallymer software, a flexible and memory-efficient collection of programs for k-mer counting and indexing of large sequence sets. Unlike previous methods, Tallymer is based on enhanced suffix arrays. This gives a much larger flexibility concerning the choice of the k-mer size. Tallymer can process large data sizes of several billion bases. We used it in a variety of applications to study the genomes of maize and other plant species. In particular, Tallymer was used to index a set of whole genome shotgun sequences from maize (B73) (total size 109 bp.). We analyzed k-mer frequencies for a wide range of k. At this low genome coverage (≈ 0.45×) highly repetitive 20-mers constituted 44% of the genome but represented only 1% of all possible k-mers. Similar low-complexity was seen in the repeat fractions of sorghum and rice. When applying our method to other maize data sets, High-C0t derived sequences showed the greatest enrichment for low-copy sequences. Among annotated TEs, the most highly repetitive were of the Ty3/gypsy class of retrotransposons, followed by the Ty1/copia class, and DNA transposons. Among expressed sequence tags (EST), a notable fraction contained high-copy k-mers, suggesting that transposons are still active in maize. Retrotransposons in Mo17 and McC cultivars were readily detected using the B73 20-mer frequency index, indicating their conservation despite extensive rearrangement across cultivars. Among one hundred annotated bacterial artificial chromosomes (BACs), k-mer frequency could be used to detect transposon-encoded genes with 92% sensitivity, compared to 96% using alignment-based repeat masking, while both methods showed 92% specificity.ConclusionThe Tallymer software was effective in a variety of applications to aid genome annotation in maize, despite limitations imposed by the relatively low coverage of sequence available. For more information on the software, see .

Reranking candidate gene models with cross-species comparison for improved gene prediction

BackgroundMost gene finders score candidate gene models with state-based methods, typically HMMs, by combining local properties (coding potential, splice donor and acceptor patterns, etc). Competing models with similar state-based scores may be distinguishable with additional information. In particular, functional and comparative genomics datasets may help to select among competing models of comparable probability by exploiting features likely to be associated with the correct gene models, such as conserved exon/intron structure or protein sequence features.ResultsWe have investigated the utility of a simple post-processing step for selecting among a set of alternative gene models, using global scoring rules to rerank competing models for more accurate prediction. For each gene locus, we first generate the K best candidate gene models using the gene finder Evigan, and then rerank these models using comparisons with putative orthologous genes from closely-related species. Candidate gene models with lower scores in the original gene finder may be selected if they exhibit strong similarity to probable orthologs in coding sequence, splice site location, or signal peptide occurrence. Experiments on Drosophila melanogaster demonstrate that reranking based on cross-species comparison outperforms the best gene models identified by Evigan alone, and also outperforms the comparative gene finders GeneWise and Augustus+.ConclusionReranking gene models with cross-species comparison improves gene prediction accuracy. This straightforward method can be readily adapted to incorporate additional lines of evidence, as it requires only a ranked source of candidate gene models.

CDNA sequences reveal considerable gene prediction inaccuracy in the Plasmodium falciparum genome

BackgroundThe completion of the Plasmodium falciparum genome represents a milestone in malaria research. The genome sequence allows for the development of genome-wide approaches such as microarray and proteomics that will greatly facilitate our understanding of the parasite biology and accelerate new drug and vaccine development. Designing and application of these genome-wide assays, however, requires accurate information on gene prediction and genome annotation. Unfortunately, the genes in the parasite genome databases were mostly identified using computer software that could make some erroneous predictions.ResultsWe aimed to obtain cDNA sequences to examine the accuracy of gene prediction in silico. We constructed cDNA libraries from mixed blood stages of P. falciparum parasite using the SMART cDNA library construction technique and generated 17332 high-quality expressed sequence tags (EST), including 2198 from primer-walking experiments. Assembly of our sequence tags produced 2548 contigs and 2671 singletons versus 5220 contigs and 5910 singletons when our EST were assembled with EST in public databases. Comparison of all the assembled EST/contigs with predicted CDS and genomic sequences in the PlasmoDB database identified 356 genes with predicted coding sequences fully covered by EST, including 85 genes (23.6%) with introns incorrectly predicted. Careful automatic software and manual alignments found an additional 308 genes that have introns different from those predicted, with 152 new introns discovered and 182 introns with sizes or locations different from those predicted. Alternative spliced and antisense transcripts were also detected. Matching cDNA to predicted genes also revealed silent chromosomal regions, mostly at subtelomere regions.ConclusionOur data indicated that approximately 24% of the genes in the current databases were predicted incorrectly, although some of these inaccuracies could represent alternatively spliced transcripts, and that more genes than currently predicted have one or more additional introns. It is therefore necessary to annotate the parasite genome with experimental data, although obtaining complete cDNA sequences from this parasite will be a formidable task due to the high AT nature of the genome. This study provides valuable information for genome annotation that will be critical for functional analyses.

Intron length distributions and gene prediction

Accurate gene prediction in eukaryotes is a difficult and subtle problem. Here we point out a useful feature of expected distributions of spliceosomal intron lengths. Since introns are removed from transcripts prior to translation, intron lengths are not expected to respect coding frame, thus the number of genomic introns that are a multiple of three bases (‘3n introns’) should be similar to the number that are a multiple of three plus one bases (or plus two bases). Skewed predicted intron length distributions thus suggest systematic errors in intron prediction. For instance, a genome-wide excess of 3n introns suggests that many internal exonic sequences have been incorrectly called introns, whereas a deficit of 3n introns suggests that many 3n introns that lack stop codons have been mistaken for exonic sequence. A survey of genomic annotations for 29 diverse eukaryotic species showed that skew in intron length distributions is a common problem. We discuss several examples of skews in genome-wide intron length distributions that indicate systematic problems with gene prediction. We suggest that evaluation of length distributions of predicted introns is a fast and simple method for detecting a variety of possible systematic biases in gene prediction or even problems with genome assemblies, and discuss ways in which these insights could be incorporated into genome annotation protocols.

Gene prediction in eukaryotes with a generalized hidden Markov model that uses hints from external sources.

BackgroundIn order to improve gene prediction, extrinsic evidence on the gene structure can be collected from various sources of information such as genome-genome comparisons and EST and protein alignments. However, such evidence is often incomplete and usually uncertain. The extrinsic evidence is usually not sufficient to recover the complete gene structure of all genes completely and the available evidence is often unreliable. Therefore extrinsic evidence is most valuable when it is balanced with sequence-intrinsic evidence.ResultsWe present a fairly general method for integration of external information. Our method is based on the evaluation of hints to potentially protein-coding regions by means of a Generalized Hidden Markov Model (GHMM) that takes both intrinsic and extrinsic information into account. We used this method to extend the ab initio gene prediction program AUGUSTUS to a versatile tool that we call AUGUSTUS+. In this study, we focus on hints derived from matches to an EST or protein database, but our approach can be used to include arbitrary user-defined hints. Our method is only moderately effected by the length of a database match. Further, it exploits the information that can be derived from the absence of such matches. As a special case, AUGUSTUS+ can predict genes under user-defined constraints, e.g. if the positions of certain exons are known. With hints from EST and protein databases, our new approach was able to predict 89% of the exons in human chromosome 22 correctly.ConclusionSensitive probabilistic modeling of extrinsic evidence such as sequence database matches can increase gene prediction accuracy. When a match of a sequence interval to an EST or protein sequence is used it should be treated as compound information rather than as information about individual positions.