Three-stage prediction of protein beta-sheets by neural networks,
alignments and graph algorithms
Protein beta-sheets
play a fundamental role in protein structure, function, evolution and bioengineering. Accurate prediction and assembly of protein beta-sheets, however, remains challenging because protein beta-sheets require formation of hydrogen bonds between linearly distant residues. Previous approaches for predicting beta-sheet topological features, such as beta-strand
alignments, in general have not exploited the global covariation and constraints characteristic of beta-sheet architectures.
Results: We propose a modular approach to the problem of predicting/assembling protein beta-sheets in a chain by integrating both local and global constraints in three steps. The first step uses recursive neural networks to predict pairing probabilities for all pairs of interstrand beta-residues from profile, secondary structure and solvent accessibility information. The second step applies dynamic programming techniques to these probabilities to derive binding pseudoenergies and optimal alignments between all pairs of beta-strands. Finally, the third step uses graph matching algorithms to predict the beta-sheet architecture of the protein by optimizing the global pseudoenergy while enforcing strong global beta-strand pairing constraints. The approach is evaluated using cross-validation methods on a large non-homologous dataset and yields significant improvements over previous methods.
Improved detection of DNA motifs using a self-organized
clustering of familial binding profiles
One of the limiting factors in deciphering transcriptional regulatory networks is the effectiveness of motif-finding software. An emerging avenue for improving motif-finding accuracy aims to incorporate generalized binding constraints of related transcription factors (TFs), named familial binding profiles (FBPs), as priors in motif identification methods. A motif-finder can thus be biased towards finding motifs from a particular TF family. However, current motif-finders allow only a single FBP to be used as a prior in a given motif-finding run. In addition, current FBP construction methods are based on manual clustering of position specific scoring matrices (PSSMs) according to the known structural properties of the TF proteins. Manual clustering assumes that the binding preferences of structurally similar TFs will also be similar. This assumption is not true, at least not for some TF families. Automatic PSSM clustering methods are thus required for augmenting the usefulness of FBPs.
Results: A novel method is developed for automatic clustering of PSSM models. The resulting FBPs are incorporated into the SOMBRERO motif-finder, significantly improving its performance when finding motifs related to those that have been incorporated. SOMBRERO is thus the only existing de novo motif-finder that can incorporate knowledge of all known PSSMs in a given motif-finding run.
Availability: The methods outlined will be incorporated into the next release of SOMBRERO, which is available from
Alignments anchored on genomic landmarks can aid in the identification of regulatory elements
The transcription start site (TSS) has been located for an increasing number of genes across several organisms. Statistical tests have shown that some cis-acting regulatory elements have positional preferences with respect to the TSS, but few strategies have emerged for locating elements by their positional preferences. This paper elaborates such a strategy. First, we align promoter regions without gaps, anchoring the alignment on each promoter's TSS. Second, we apply a novel word-specific mask. Third, we apply a clustering test related to gapless BLAST statistics. The test examines whether any specific word is placed unusually consistently with respect to the TSS. Finally, our program A-GLAM, an extension of the GLAM program, uses significant wordpositions as new anchors to realign the sequences. A Gibbs sampling algorithm then locates putative cis-acting regulatory elements. Usually, Gibbs sampling requires a preliminary masking step, to avoid convergence onto a dominant but uninteresting signal from a DNA repeat. However, since the positional anchors focus A-GLAM on the motif of interest, masking DNA repeats during Gibbs sampling becomes unnecessary.
Results: In a set of human DNA sequences with experimentally characterized TSSs, the placement of 791 octonucleotide words was unusually consistent (multiple test corrected P < 0.05). Alignments anchored on these words sometimes located statistically significant motifs inaccessible to GLAM or AlignACE.
NOTE: The A-GLAM program and a list of statistically significant words are available at ftp://ftp.ncbi.nih.gov/pub/spouge/papers/archive/AGLAM