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

Abstract

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 approximate to 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

A Machine Learning Approach To Identify Hydrogenosomal Proteins in Trichomonas vaginalis

“…This approach allows us to extract information from various feature combinations in order to identify patterns within a known learning set (bait) and perform subsequent predictions on an unknown data set (prey). Machine learning algorithms have been used for biological data mining, including applications for prediction of protein targeting signals (see reference 70 for a review) or protein-protein interactions (37), and finding protein-encoding genes (72) and noncoding RNAs (51) within completely sequenced genomes. Using this approach we predicted and subsequently validated experimentally new hydrogenosomal proteins, some of which do not carry N-terminal targeting motifs.…”

A Comprehensive Review of Emerging Computational Methods for Gene Identification

“…A SVM-based two-layer approach [94] consists of independent SVM signal and content detectors, and hidden semi-Markov (HSM) SVMs. The first layer is SVM feature recognition, while the second layer is gene structure reconstruction.…”

“…Machine learning (ML) methods, particularly SVM-based methods, have been a significant methodology in genome analysis for solving a wide range of problems, including gene finding [94,198]. Recently, machine learning based methods [169,199] are being further used for gene prediction in metagenomic fragments.…”

Gene function prediction with knowledge from gene ontology

“…In order to improve the classification accuracy, many approaches have been proposed from the perspective of machine learning and pattern recognition Guyon et al, 2002;Lee and Zhang, 2006;Nevins and Potti, 2007;Liu and Huang, 2008;Schweikert et al, 2009;Zheng et al, 2009;Cai et al, 2010;Leung and Hung, 2010;Zare et al, 2011;Zheng et al, 2011;Wang et al, 2012). Despite of the success achieved by these advanced techniques, the improvement for the classification accuracy remains limited, because they only deal with the data obtained from the biological experiments, which contains noise and missing values.…”