Dataset Information

Classifier assessment and feature selection for recognizing short coding sequences of human genes.

ABSTRACT: With the ever-increasing pace of genome sequencing, there is a great need for fast and accurate computational tools to automatically identify genes in these genomes. Although great progress has been made in the development of gene-finding algorithms during the past decades, there is still room for further improvement. In particular, the issue of recognizing short exons in eukaryotes is still not solved satisfactorily. This article is devoted to assessing various linear and kernel-based classification algorithms and selecting the best combination of Z-curve features for further improvement of the issue. Eight state-of-the-art linear and kernel-based supervised pattern recognition techniques were used to identify the short (21-192?bp) coding sequences of human genes. By measuring the prediction accuracy, the tradeoff between sensitivity and specificity and the time consumption, partial least squares (PLS) and kernel partial least squares (KPLS) algorithms were verified to be the most optimal linear and kernel-based classifiers, respectively. A surprising result was that, by making good use of the interpretability of the PLS and the Z-curve methods, 93 Z-curve features were proved to be the best selective combination. Using them, the average recognition accuracy was improved as high as 7.7% by means of KPLS when compared with what was obtained by the Fisher discriminant analysis using 189 Z-curve variables (Gao and Zhang, 2004 ). The used codes are freely available from the following approaches (implemented in MATLAB and supported on Linux and MS Windows): (1) SVM: http://www.support-vector-machines.org/SVM_soft.html. (2) GP: http://www.gaussianprocess.org. (3) KPLS and KFDA: Taylor, J.S., and Cristianini, N. 2004. Kernel Methods for Pattern Analysis. Cambridge University Press, Cambridge, UK. (4) PLS: Wise, B.M., and Gallagher, N.B. 2011. PLS-Toolbox for use with MATLAB: ver 1.5.2. Eigenvector Technologies, Manson, WA. Supplementary Material for this article is available at www.liebertonline.com/cmb.

SUBMITTER: Song K

PROVIDER: S-EPMC3298678 | biostudies-literature | 2012 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Classifier assessment and feature selection for recognizing short coding sequences of human genes.

Song Kai K Zhang Ze Z Tong Tuo-Peng TP Wu Fang F

Journal of computational biology : a journal of computational molecular cell biology 20120301 3

With the ever-increasing pace of genome sequencing, there is a great need for fast and accurate computational tools to automatically identify genes in these genomes. Although great progress has been made in the development of gene-finding algorithms during the past decades, there is still room for further improvement. In particular, the issue of recognizing short exons in eukaryotes is still not solved satisfactorily. This article is devoted to assessing various linear and kernel-based classific ...[more]

PMID: 22401589

Similar Datasets

Project description:Comparison of polymorphism at synonymous and non-synonymous sites in protein-coding DNA can provide evidence for selective constraint. Non-coding DNA that forms part of the regulatory landscape presents more of a challenge since there is not such a clear-cut distinction between sites under stronger and weaker selective constraint. Here, we consider putative regulatory elements termed Conserved Non-coding Elements (CNEs) defined by their high level of sequence identity across all vertebrates. Some mutations in these regions have been implicated in developmental disorders; we analyse CNE polymorphism data to investigate whether such deleterious effects are widespread in humans. Single nucleotide variants from the HapMap and 1000 Genomes Projects were mapped across nearly 2000 CNEs. In the 1000 Genomes data we find a significant excess of rare derived alleles in CNEs relative to coding sequences; this pattern is absent in HapMap data, apparently obscured by ascertainment bias. The distribution of polymorphism within CNEs is not uniform; we could identify two categories of sites by exploiting deep vertebrate alignments: stretches that are non-variant, and those that have at least one substitution. The conserved category has fewer polymorphic sites and a greater excess of rare derived alleles, which can be explained by a large proportion of sites under strong purifying selection within humans--higher than that for non-synonymous sites in most protein coding regions, and comparable to that at the strongly conserved trans-dev genes. Conversely, the more evolutionarily labile CNE sites have an allele frequency distribution not significantly different from non-synonymous sites. Future studies should exploit genome-wide re-sequencing to obtain better coverage in selected non-coding regions, given the likelihood that mutations in evolutionarily conserved enhancer sequences are deleterious. Discovery pipelines should validate non-coding variants to aid in identifying causal and risk-enhancing variants in complex disorders, in contrast to the current focus on exome sequencing.

Project description:The aim of the present study was to investigate the feature genes in metastatic breast cancer samples. A total of 5 expression profiles of metastatic breast cancer samples were downloaded from the Gene Expression Omnibus database, which were then analyzed using the MetaQC and MetaDE packages in R language. The feature genes between metastasis and non‑metastasis samples were screened under the threshold of P<0.05. Based on the protein‑protein interactions (PPIs) in the Biological General Repository for Interaction Datasets, Human Protein Reference Database and Biomolecular Interaction Network Database, the PPI network of the feature genes was constructed. The feature genes identified by topological characteristics were then used for support vector machine (SVM) classifier training and verification. The accuracy of the SVM classifier was then evaluated using another independent dataset from The Cancer Genome Atlas database. Finally, function and pathway enrichment analyses for genes in the SVM classifier were performed. A total of 541 feature genes were identified between metastatic and non‑metastatic samples. The top 10 genes with the highest betweenness centrality values in the PPI network of feature genes were Nuclear RNA Export Factor 1, cyclin‑dependent kinase 2 (CDK2), myelocytomatosis proto‑oncogene protein (MYC), Cullin 5, SHC Adaptor Protein 1, Clathrin heavy chain, Nucleolin, WD repeat domain 1, proteasome 26S subunit non‑ATPase 2 and telomeric repeat binding factor 2. The cyclin‑dependent kinase inhibitor 1A (CDKN1A), E2F transcription factor 1 (E2F1), and MYC interacted with CDK2. The SVM classifier constructed by the top 30 feature genes was able to distinguish metastatic samples from non‑metastatic samples [correct rate, specificity, positive predictive value and negative predictive value >0.89; sensitivity >0.84; area under the receiver operating characteristic curve (AUROC) >0.96]. The verification of the SVM classifier in an independent dataset (35 metastatic samples and 143 non‑metastatic samples) revealed an accuracy of 94.38% and AUROC of 0.958. Cell cycle associated functions and pathways were the most significant terms of the 30 feature genes. A SVM classifier was constructed to assess the possibility of breast cancer metastasis, which presented high accuracy in several independent datasets. CDK2, CDKN1A, E2F1 and MYC were indicated as the potential feature genes in metastatic breast cancer.

Dataset Information

Classifier assessment and feature selection for recognizing short coding sequences of human genes.

Publications

Classifier assessment and feature selection for recognizing short coding sequences of human genes.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets