Biomedical subjects

Timothy L Bailey

Publications and source records attributed to Timothy L Bailey.

11 recordsLinked to original sources

PubMed2006

MEME: discovering and analyzing DNA and protein sequence motifs.

MEME (Multiple EM for Motif Elicitation) is one of the most widely used tools for searching for novel 'signals' in sets of biological sequences. Applications include the discovery of new transcription factor binding sites and protein domains. MEME works by searching for repeated, ungapped sequence patterns that occur in the DNA or protein sequences provided by the user. Users can perform MEME searches via the web server hosted by the National Biomedical Computation Resource (http://meme.nbcr.net) and several mirror sites. Through the same web server, users can also access the Motif Alignment and Search Tool to search sequence databases for matches to motifs encoded in several popular formats. By clicking on buttons in the MEME output, users can compare the motifs discovered in their input sequences with databases of known motifs, search sequence databases for matches to the motifs and display the motifs in various formats. This article describes the freely accessible web server and its architecture, and discusses ways to use MEME effectively to find new sequence patterns in biological sequences and analyze their significance.

Amino Acid Motifs↗

PubMed2006

Identifying sequence regions undergoing conformational change via predicted continuum secondary structure.

MOTIVATION: Conformational flexibility is essential to the function of many proteins, e.g. catalytic activity. To assist efforts in determining and exploring the functional properties of a protein, it is desirable to automatically identify regions that are prone to undergo conformational changes. It was recently shown that a probabilistic predictor of continuum secondary structure is more accurate than categorical predictors for structurally ambivalent sequence regions, suggesting that such models are suited to characterize protein flexibility. RESULTS: We develop a computational method for identifying regions that are prone to conformational change directly from the amino acid sequence. The method uses the entropy of the probabilistic output of an 8-class continuum secondary structure predictor. Results for 171 unique amino acid sequences with well-characterized variable structure (identified in the 'Macromolecular movements database') indicate that the method is highly sensitive at identifying flexible protein regions, but false positives remain a problem. The method can be used to explore conformational flexibility of proteins (including hypothetical or synthetic ones) whose structure is yet to be determined experimentally. AVAILABILITY: The predictor, sequence data and supplementary studies are available at http://pprowler.itee.uq.edu.au/sspred/ and are free for academic use.

Algorithms↗

PubMed2006

Pseudo-messenger RNA: phantoms of the transcriptome.

The mammalian transcriptome harbours shadowy entities that resist classification and analysis. In analogy with pseudogenes, we define pseudo-messenger RNA to be RNA molecules that resemble protein-coding mRNA, but cannot encode full-length proteins owing to disruptions of the reading frame. Using a rigorous computational pipeline, which rules out sequencing errors, we identify 10,679 pseudo-messenger RNAs (approximately half of which are transposon-associated) among the 102,801 FANTOM3 mouse cDNAs: just over 10% of the FANTOM3 transcriptome. These comprise not only transcribed pseudogenes, but also disrupted splice variants of otherwise protein-coding genes. Some may encode truncated proteins, only a minority of which appear subject to nonsense-mediated decay. The presence of an excess of transcripts whose only disruptions are opal stop codons suggests that there are more selenoproteins than currently estimated. We also describe compensatory frameshifts, where a segment of the gene has changed frame but remains translatable. In summary, we survey a large class of non-standard but potentially functional transcripts that are likely to encode genetic information and effect biological processes in novel ways. Many of these transcripts do not correspond cleanly to any identifiable object in the genome, implying fundamental limits to the goal of annotating all functional elements at the genome sequence level.

Animals↗

PubMed2006

The abundance of short proteins in the mammalian proteome.

Short proteins play key roles in cell signalling and other processes, but their abundance in the mammalian proteome is unknown. Current catalogues of mammalian proteins exhibit an artefactual discontinuity at a length of 100 aa, so that protein abundance peaks just above this length and falls off sharply below it. To clarify the abundance of short proteins, we identify proteins in the FANTOM collection of mouse cDNAs by analysing synonymous and non-synonymous substitutions with the computer program CRITICA. This analysis confirms that there is no real discontinuity at length 100. Roughly 10% of mouse proteins are shorter than 100 aa, although the majority of these are variants of proteins longer than 100 aa. We identify many novel short proteins, including a "dark matter" subset containing ones that lack detectable homology to other known proteins. Translation assays confirm that some of these novel proteins can be translated and localised to the secretory pathway.

Amino Acid Sequence↗

PubMed2006

Discrimination of non-protein-coding transcripts from protein-coding mRNA.

Several recent studies indicate that mammals and other organisms produce large numbers of RNA transcripts that do not correspond to known genes. It has been suggested that these transcripts do not encode proteins, but may instead function as RNAs. However, discrimination of coding and non-coding transcripts is not straightforward, and different laboratories have used different methods, whose ability to perform this discrimination is unclear. In this study, we examine ten bioinformatic methods that assess protein-coding potential and compare their ability and congruency in the discrimination of non-coding from coding sequences, based on four underlying principles: open reading frame size, sequence similarity to known proteins or protein domains, statistical models of protein-coding sequence, and synonymous versus non-synonymous substitution rates. Despite these different approaches, the methods show broad concordance, suggesting that coding and non-coding transcripts can, in general, be reliably discriminated, and that many of the recently discovered extra-genic transcripts are indeed non-coding. Comparison of the methods indicates reasons for unreliable predictions, and approaches to increase confidence further. Conversely and surprisingly, our analyses also provide evidence that as much as approximately 10% of entries in the manually curated protein database Swiss-Prot are erroneous translations of actually non-coding transcripts.

Algorithms↗

PubMed2006

GONOME: measuring correlations between GO terms and genomic positions.

BACKGROUND: Current methods to find significantly under- and over-represented gene ontology (GO) terms in a set of genes consider the genes as equally probable "balls in a bag", as may be appropriate for transcripts in micro-array data. However, due to the varying length of genes and intergenic regions, that approach is inappropriate for deciding if any GO terms are correlated with a set of genomic positions. RESULTS: We present an algorithm--GONOME--that can determine which GO terms are significantly associated with a set of genomic positions given a genome annotated with (at least) the starts and ends of genes. We show that certain GO terms may appear to be significantly associated with a set of randomly chosen positions in the human genome if gene lengths are not considered, and that these same terms have been reported as significantly over-represented in a number of recent papers. This apparent over-representation disappears when gene lengths are considered, as GONOME does. For example, we show that, when gene length is taken into account, the term "development" is not significantly enriched in genes associated with human CpG islands, in contradiction to a previous report. We further demonstrate the efficacy of GONOME by showing that occurrences of the proteosome-associated control element (PACE) upstream activating sequence in the S. cerevisiae genome associate significantly to appropriate GO terms. An extension of this approach yields a whole-genome motif discovery algorithm that allows identification of many other promoter sequences linked to different types of genes, including a large group of previously unknown motifs significantly associated with the terms 'translation' and 'translational elongation'. CONCLUSION: GONOME is an algorithm that correctly extracts over-represented GO terms from a set of genomic positions. By explicitly considering gene size, GONOME avoids a systematic bias toward GO terms linked to large genes. Inappropriate use of existing algorithms that do not take gene size into account has led to erroneous or suspect conclusions. Reciprocally GONOME may be used to identify new features in genomes that are significantly associated with particular categories of genes.

Algorithms↗

PubMed2006

Prediction of protein continuum secondary structure with probabilistic models based on NMR solved structures.

BACKGROUND: The structure of proteins may change as a result of the inherent flexibility of some protein regions. We develop and explore probabilistic machine learning methods for predicting a continuum secondary structure, i.e. assigning probabilities to the conformational states of a residue. We train our methods using data derived from high-quality NMR models. RESULTS: Several probabilistic models not only successfully estimate the continuum secondary structure, but also provide a categorical output on par with models directly trained on categorical data. Importantly, models trained on the continuum secondary structure are also better than their categorical counterparts at identifying the conformational state for structurally ambivalent residues. CONCLUSION: Cascaded probabilistic neural networks trained on the continuum secondary structure exhibit better accuracy in structurally ambivalent regions of proteins, while sustaining an overall classification accuracy on par with standard, categorical prediction methods.

Algorithms↗

PubMed2005

Prediction of protein B-factor profiles.

The polypeptide backbones and side chains of proteins are constantly moving due to thermal motion and the kinetic energy of the atoms. The B-factors of protein crystal structures reflect the fluctuation of atoms about their average positions and provide important information about protein dynamics. Computational approaches to predict thermal motion are useful for analyzing the dynamic properties of proteins with unknown structures. In this article, we utilize a novel support vector regression (SVR) approach to predict the B-factor distribution (B-factor profile) of a protein from its sequence. We explore schemes for encoding sequences and various settings for the parameters used in SVR. Based on a large dataset of high-resolution proteins, our method predicts the B-factor distribution with a Pearson correlation coefficient (CC) of 0.53. In addition, our method predicts the B-factor profile with a CC of at least 0.56 for more than half of the proteins. Our method also performs well for classifying residues (rigid vs. flexible). For almost all predicted B-factor thresholds, prediction accuracies (percent of correctly predicted residues) are greater than 70%. These results exceed the best results of other sequence-based prediction methods.

Algorithms↗

PubMed2005

Assessing computational tools for the discovery of transcription factor binding sites.

The prediction of regulatory elements is a problem where computational methods offer great hope. Over the past few years, numerous tools have become available for this task. The purpose of the current assessment is twofold: to provide some guidance to users regarding the accuracy of currently available tools in various settings, and to provide a benchmark of data sets for assessing future tools.

Amino Acid Motifs↗

PubMed2003

Searching for statistically significant regulatory modules.

MOTIVATION: The regulatory machinery controlling gene expression is complex, frequently requiring multiple, simultaneous DNA-protein interactions. The rate at which a gene is transcribed may depend upon the presence or absence of a collection of transcription factors bound to the DNA near the gene. Locating transcription factor binding sites in genomic DNA is difficult because the individual sites are small and tend to occur frequently by chance. True binding sites may be identified by their tendency to occur in clusters, sometimes known as regulatory modules. RESULTS: We describe an algorithm for detecting occurrences of regulatory modules in genomic DNA. The algorithm, called mcast, takes as input a DNA database and a collection of binding site motifs that are known to operate in concert. mcast uses a motif-based hidden Markov model with several novel features. The model incorporates motif-specific p-values, thereby allowing scores from motifs of different widths and specificities to be compared directly. The p-value scoring also allows mcast to only accept motif occurrences with significance below a user-specified threshold, while still assigning better scores to motif occurrences with lower p-values. mcast can search long DNA sequences, modeling length distributions between motifs within a regulatory module, but ignoring length distributions between modules. The algorithm produces a list of predicted regulatory modules, ranked by E-value. We validate the algorithm using simulated data as well as real data sets from fruitfly and human. AVAILABILITY: http://meme.sdsc.edu/MCAST/paper

Algorithms↗

PubMed2002

Estimating and evaluating the statistics of gapped local-alignment scores.

We present a novel maximum-likelihood-based algorithm for estimating the distribution of alignment scores from the scores of unrelated sequences in a database search. Using a new method for measuring the accuracy of p-values, we show that our maximum-likelihood-based algorithm is more accurate than existing regression-based and lookup table methods. We explore a more sophisticated way of modeling and estimating the score distributions (using a two-component mixture model and expectation maximization), but conclude that this does not improve significantly over simply ignoring scores with small E-values during estimation. Finally, we measure the classification accuracy of p-values estimated in different ways and observe that inaccurate p-values can, somewhat paradoxically, lead to higher classification accuracy. We explain this paradox and argue that statistical accuracy, not classification accuracy, should be the primary criterion in comparisons of similarity search methods that return p-values that adjust for target sequence length.

Algorithms↗