Comparing, Ranking, and Filtering Motifs with Character Classes: Application to Biological Sequences Analysis

Linear motifs mediate a wide variety of cellular functions, which makes their characterization in protein sequences crucial to understanding cellular systems. However, the short length and degenerate nature of linear motifs make their discovery a difficult problem. Here, we introduce MotifHound, an algorithm particularly suited for the discovery of small and degenerate linear motifs. MotifHound performs an exact and exhaustive enumeration of all motifs present in proteins of interest, including all of their degenerate forms, and scores the overrepresentation of each motif based on its occurrence in proteins of interest relative to a background (e.g., proteome) using the hypergeometric distribution. To assess MotifHound, we benchmarked it together with state-of-the-art algorithms. The benchmark consists of 11,880 sets of proteins from S. cerevisiae; in each set, we artificially spiked-in one motif varying in terms of three key parameters, (i) number of occurrences, (ii) length and (iii) the number of degenerate or “wildcard” positions. The benchmark enabled the evaluation of the impact of these three properties on the performance of the different algorithms. The results showed that MotifHound and SLiMFinder were the most accurate in detecting degenerate linear motifs. Interestingly, MotifHound was 15 to 20 times faster at comparable accuracy and performed best in the discovery of highly degenerate motifs. We complemented the benchmark by an analysis of proteins experimentally shown to bind the FUS1 SH3 domain from S. cerevisiae. Using the full-length protein partners as sole information, MotifHound recapitulated most experimentally determined motifs binding to the FUS1 SH3 domain. Moreover, these motifs exhibited properties typical of SH3 binding peptides, e.g., high intrinsic disorder and evolutionary conservation, despite the fact that none of these properties were used as prior information. MotifHound is available (http://michnick.bcm.umontreal.ca or http://tinyurl.com/motifhound) together with the benchmark that can be used as a reference to assess future developments in motif discovery.

Sequence motif discovery algorithms are an important part of the computational biologist's toolkit. The purpose of motif discovery is to discover patterns in biopolymer (nucleotide or protein) sequences to better understand the structure and function of the molecules the sequences represent. This chapter provides an overview of the use of sequence motif discovery in biology and a general guide to the use of motif discovery algorithms. This chapter examines the types of biological features that DNA and protein motifs can represent and their usefulness. This chapter also defines what sequence motifs are, how they are represented, and general techniques for discovering them. The primary focus of the chapter is on one aspect of motif discovery: discovering motifs in a set of unaligned DNA or protein sequences. This chapter also provides the steps useful for checking the biological validity and investigating the function of sequence motifs using methods such as motif scanning-searching for matches to motifs in a given sequence or a database of sequences. A discussion of some limitations of motif discovery concludes the chapter.

Sequence motif discovery algorithms are an important part of the computational biologist's toolkit. The purpose of motif discovery is to discover patterns in biopolymer (nucleotide or protein) sequences in order to better understand the structure and function of the molecules the sequences represent. This chapter provides an overview of the use of sequence motif discovery in biology and a general guide to the use of motif discovery algorithms. The chapter discusses the types of biological features that DNA and protein motifs can represent and their usefulness. It also defines what sequence motifs are, how they are represented, and general techniques for discovering them. The primary focus is on one aspect of motif discovery: discovering motifs in a set of unaligned DNA or protein sequences. Also presented are steps useful for checking the biological validity and investigating the function of sequence motifs using methods such as motif scanning--searching for matches to motifs in a given sequence or a database of sequences. A discussion of some limitations of motif discovery concludes the chapter.

Identification of conserved domains or motifs in proteins may aid in the localization and analysis of important structural and functional regions. We report here a protein sequence motif, called the bromodomain (1), that has been found in six genes from humans (CCG1 and RING3), Drosophila (fsh and brm), and yeast (SPT7 and SNF2). The fsh and brm genes are required maternally for proper expression of certain homeotic genes (1,2). The SPT7 and SNF2 genes of Saccharomyces cerevisiae encode transcriptional activators (3,4). The SNF2 and brm proteins have extensive sequence homology (1). It is not clear whether the two human genes are involved in processes. CCG1 is a DNA-binding protein that complements temperature sensitive mutations that cause cell cycle arrest in Gl (5). The RING3 gene is a newly identified human gene of unknown function, mapping to the class II region of the human major histocompatibility locus, that has substantial homology to the fsh gene (6). Each of these proteins has one or two copies of the 61 —63 amino acid bromodomain (Figure 1). The sequence identity is highest (~ 80%) between the corresponding fsh and RING3 repeats. More typically, any two motifs show 25-40% identity, with 50-60% identity between two repeats within the same protein. There are seven invariant residues, four of which are aromatic amino acids, and numerous conservative substitutions. The location of the motif(s) within the individual proteins is variable, and two motifs may be present in tandem or separated by sequences unrelated to the motif. Secondary structure prediction methods were applied to the bromodomain and revealed two strongly predicted amphipathic a helices followed by reverse turns. Four of the invariant residues are located within the turns, and the other three are located among four highly conserved proline residues at the N terminus. The functional significance of the bromodomain is unknown, and experiments with the two yeast proteins indicate that it may be dispensable in some cases (6; P.Tan and F.W., unpublished data). However, the widespread conservation of the bromodomain suggests that it is important for some aspects of protein function. We speculate that the hydrophobic surfaces of the helices and the invariant hydrophobic residues could serve as sites of intramolecular or intermolecular protein-protein interaction. Such interactions could influence the assembly or activity of multicomponent complexes involved in transcription activation or other cellular processes.

Data Mining is the process to extract hidden predictive information from large databases. In Bioinformatics, data mining enables researchers to meet the challenge of mining large amount of biomolecular data to discover real knowledge. Major research efforts done in the area of bioinformatics involves sequence analysis, protein structure prediction and gene finding. Proteins are said to be prominent molecules in our cells. They involve virtually in all cell functions. The activities and functions of proteins can be determined by protein sequence motifs. These protein motifs are identified from the segments of protein sequences. All segments may not be important to produce good motif patterns. The generated sequence segments do not have classes or labels. Hence, unsupervised segment selection technique is adopted to select significant segments. Therefore Singular Value Decomposition (SVD) entropy method is adopted to select significant sequence segments. In this proposed work, weighted K-Means and Adaptive Fuzzy C-Means have been applied to the selected segments to generate granules, since large amount of segments cannot be grouped or clustered as such. Each granules generated by weighted K-Means algorithm are further clustered by using the K-Means algorithm and granules generated by Adaptive Fuzzy C-Means algorithm are clustered by using Weighted K-Means. The two proposed models are compared with K-Means granular computing model. The experimental results show that Adaptive Fuzzy C-Means with Weighted K-Means technique produces better results than K-Means and weighted K-Means granular computing methods.

BackgroundA key step in the regulation of gene expression is the sequence-specific binding of transcription factors (TFs) to their DNA recognition sites. However, elucidating TF binding site (TFBS) motifs in higher eukaryotes has been challenging, even when employing cross-species sequence conservation. We hypothesized that for human and mouse, many orthologous genes expressed in a similarly tissue-specific manner in both human and mouse gene expression data, are likely to be co-regulated by orthologous TFs that bind to DNA sequence motifs present within noncoding sequence conserved between these genomes.ResultsWe performed automated motif searching and merging across four different motif finding algorithms, followed by filtering of the resulting motifs for those that contain blocks of information content. Applying this motif finding strategy to conserved noncoding regions surrounding co-expressed tissue-specific human genes allowed us to discover both previously known, and many novel candidate, regulatory DNA motifs in all 18 tissue-specific expression clusters that we examined. For previously known TFBS motifs, we observed that if a TF was expressed in the specified tissue of interest, then in most cases we identified a motif that matched its TRANSFAC motif; conversely, of all those discovered motifs that matched TRANSFAC motifs, most of the corresponding TF transcripts were expressed in the tissue(s) corresponding to the expression cluster for which the motif was found.ConclusionOur results indicate that the integration of the results from multiple motif finding tools identifies and ranks highly more known and novel motifs than does the use of just one of these tools. In addition, we believe that our simultaneous enrichment strategies helped to identify likely human cis regulatory elements. A number of the discovered motifs may correspond to novel binding site motifs for as yet uncharacterized tissue-specific TFs. We expect this strategy to be useful for identifying motifs in other metazoan genomes.

Advances in mass spectrometry-based proteomics have yielded a substantial mapping of the tyrosine phosphoproteome and thus provided an important step toward a systematic analysis of intracellular signaling networks in higher eukaryotes. In this study we decomposed an uncharacterized proteomics data set of 481 unique phosphotyrosine (Tyr(P)) peptides by sequence similarity to known ligands of the Src homology 2 (SH2) and the phosphotyrosine binding (PTB) domains. From 20 clusters we extracted 16 known and four new interaction motifs. Using quantitative mass spectrometry we pulled down Tyr(P)-specific binding partners for peptides corresponding to the extracted motifs. We confirmed numerous previously known interaction motifs and found 15 new interactions mediated by phosphosites not previously known to bind SH2 or PTB. Remarkably, a novel hydrophobic N-terminal motif ((L/V/I)(L/V/I)pY) was identified and validated as a binding motif for the SH2 domain-containing inositol phosphatase SHIP2. Our decomposition of the in vivo Tyr(P) proteome furthermore suggests that two-thirds of the Tyr(P) sites mediate interaction, whereas the remaining third govern processes such as enzyme activation and nucleic acid binding.

We have searched for the exclusivity of common sequence motifs of the mitochondrial uncoupling proteins (UCP1, UCP2, UCP3, UCP4, BMCP1, and plant UCP [PUMP]) within the gene family of mitochondrial anion carrier proteins. The UCP-specific sequences, "UCP signatures", were found in the first, second, and fourth alpha-helices. First: Ala/Ser-Cys/Thr/n-n/Phe-Ala/Gly-[negatively charged residue]-n/Phe-n/Cys-Thr-Phe/n; second: Gly/Ala-Ile/Leu-Gln/X-[positively charged residue]-NH-n/Cys-Ser/nphi/X-n/Ser-OH/Gly-n-[positively charged residue]-Ile/Met-Gly/Val-n/Thr; fourth: Pro-Asn/ Thr-n-X-[positively charged residue]-Asn/Ser/Ala-n-n-Ile/Leu-n-Asn/Val-Cys/n-n/Thr-[negatively charged residue]-n-n/Thr/Pro-OH/Val (n, nonpolar; phi, aromatic; (positively charged residue/negatively charged residue, charged residue). The second and part of the third signature are also present in the yeast dicarboxylate transporter. The UCP signature excluding BMCP1 was also found in the second matrix segment: [positively charged residue]-(Pro/ del-Leu/del)-[positively charged residue]-phi-X-Gly/Ser-Thr/n-X-NH/[negatively charged residue]-Ala-phi. These UCP signatures are thought to be involved in fatty acid anion binding and translocation.

9] Profile analysis

Finding functional motifs in protein sequences with deep learning and natural language models

The existence of preserved subsequences in a set of related protein sequences suggests that they might play a structural and functional role in protein's mechanisms. Due to its exploratory approach, the mining process tends to deliver a large number of motifs. Therefore it is critical to release methods that identify relevant significant motifs. Many measures of interest and significance have been proposed. However, since motifs have a wide range of applications, how to choose the appropriate significance measures is application dependent. Some measures show consistent results being highly correlated, while others show disagreements. In this paper we review existent measures and study their behavior in order to assist the selection of the most appropriate set of measures. An experimental evaluation of the measures for high quality patterns from the Prosite database is presented

Here, we present DeepBIO, the first-of-its-kind automated and interpretable deep-learning platform for high-throughput biological sequence functional analysis. DeepBIO is a one-stop-shop web service that enables researchers to develop new deep-learning architectures to answer any biological question. Specifically, given any biological sequence data, DeepBIO supports a total of 42 state-of-the-art deep-learning algorithms for model training, comparison, optimization and evaluation in a fully automated pipeline. DeepBIO provides a comprehensive result visualization analysis for predictive models covering several aspects, such as model interpretability, feature analysis and functional sequential region discovery. Additionally, DeepBIO supports nine base-level functional annotation tasks using deep-learning architectures, with comprehensive interpretations and graphical visualizations to validate the reliability of annotated sites. Empowered by high-performance computers, DeepBIO allows ultra-fast prediction with up to million-scale sequence data in a few hours, demonstrating its usability in real application scenarios. Case study results show that DeepBIO provides an accurate, robust and interpretable prediction, demonstrating the power of deep learning in biological sequence functional analysis. Overall, we expect DeepBIO to ensure the reproducibility of deep-learning biological sequence analysis, lessen the programming and hardware burden for biologists and provide meaningful functional insights at both the sequence level and base level from biological sequences alone. DeepBIO is publicly available at https://inner.wei-group.net/DeepBIO.

We analyze the convergence rate of a simplified version of a popular Gibbs sampling method used for statistical discovery of gene regulatory binding motifs in DNA sequences. This sampler satisfies a very strong form of ergodicity (uniform). However, we show that, due to multimodality of the posterior distribution, the rate of convergence often decreases exponentially as a function of the length of the DNA sequence. Specifically, we show that this occurs whenever there is more than one true repeating pattern in the data. In practice there are typically multiple such patterns in biological data, the goal being to detect the most well-conserved and frequently-occurring of these. Our findings match empirical results, in which the motif-discovery Gibbs sampler has exhibited such poor convergence that it is used only for finding modes of the posterior distribution (candidate motifs) rather than for obtaining samples from that distribution. Ours are some of the first meaningful bounds on the convergence rate of a Markov chain method for sampling from a multimodal posterior distribution, as a function of statistical quantities like the number of observations.

BackgroundLong short-term memory (LSTM) is one of the most attractive deep learning methods to learn time series or contexts of input data. Increasing studies, including biological sequence analyses in bioinformatics, utilize this architecture. Amino acid sequence profiles are widely used for bioinformatics studies, such as sequence similarity searches, multiple alignments, and evolutionary analyses. Currently, many biological sequences are becoming available, and the rapidly increasing amount of sequence data emphasizes the importance of scalable generators of amino acid sequence profiles.ResultsWe employed the LSTM network and developed a novel profile generator to construct profiles without any assumptions, except for input sequence context. Our method could generate better profiles than existing de novo profile generators, including CSBuild and RPS-BLAST, on the basis of profile-sequence similarity search performance with linear calculation costs against input sequence size. In addition, we analyzed the effects of the memory power of LSTM and found that LSTM had high potential power to detect long-range interactions between amino acids, as in the case of beta-strand formation, which has been a difficult problem in protein bioinformatics using sequence information.ConclusionWe demonstrated the importance of sequence context and the feasibility of LSTM on biological sequence analyses. Our results demonstrated the effectiveness of memories in LSTM and showed that our de novo profile generator, SPBuild, achieved higher performance than that of existing methods for profile prediction of beta-strands, where long-range interactions of amino acids are important and are known to be difficult for the existing window-based prediction methods. Our findings will be useful for the development of other prediction methods related to biological sequences by machine learning methods.

GR-BUCHER