BACKGROUND: Hidden Markov Models (HMMs) are a powerful tool for protein domain identification. The Pfam database notably provides a large collection of HMMs which are widely used for the annotation of proteins in new sequenced organisms. In Pfam, each domain family is represented by a curated multiple sequence alignment from which a profile HMM is built. In spite of their high specificity, HMMs may lack sensitivity when searching for domains in divergent organisms. This is particularly the case for species with a biased amino-acid composition, such as P. falciparum, the main causal agent of human malaria. In this context, fitting HMMs to the specificities of the target proteome can help identify additional domains. RESULTS: Using P. falciparum as an example, we compare approaches that have been proposed for this problem, and present two alternative methods. Because previous attempts strongly rely on known domain occurrences in the target species or its close relatives, they mainly improve the detection of domains which belong to already identified families. Our methods learn global correction rules that adjust amino-acid distributions associated with the match states of HMMs. These rules are applied to all match states of the whole HMM library, thus enabling the detection of domains from previously absent families. Additionally, we propose a procedure to estimate the proportion of false positives among the newly discovered domains. Starting with the Pfam standard library, we build several new libraries with the different HMM-fitting approaches. These libraries are first used to detect new domain occurrences with low E-values. Second, by applying the Co-Occurrence Domain Discovery (CODD) procedure we have recently proposed, the libraries are further used to identify likely occurrences among potential domains with higher E-values. CONCLUSION: We show that the new approaches allow identification of several domain families previously absent in the P. falciparum proteome and the Apicomplexa phylum, and identify many domains that are not detected by previous approaches. In terms of the number of new discovered domains, the new approaches outperform the previous ones when no close species are available or when they are used to identify likely occurrences among potential domains with high E-values. All predictions on P. falciparum have been integrated into a dedicated website which pools all known/new annotations of protein domains and functions for this organism. A software implementing the two proposed approaches is available at the same address: http://www.lirmm.fr/~terrapon/HMMfit/

MOTIVATION: Hidden Markov models (HMMs) have proved to be a powerful tool for protein domain identification in newly sequenced organisms. However, numerous domains may be missed in highly divergent proteins. This is the case for Plasmodium falciparum proteins, the main causal agent of human malaria. RESULTS: We propose a method to improve the sensitivity of HMM domain detection by exploiting the tendency of the domains to appear preferentially with a few other favorite domains in a protein. When sequence information alone is not sufficient to warrant the presence of a particular domain, our method enables its detection on the basis of the presence of other Pfam or InterPro domains. Moreover, a shuffling procedure allows us to estimate the false discovery rate associated with the results. Applied to P. falciparum, our method identifies 585 new Pfam domains (versus the 3683 already known domains in the Pfam database) with an estimated error rate <20%. These new domains provide 387 new Gene Ontology (GO) annotations to the P. falciparum proteome. Analogous and congruent results are obtained when applying the method to related Plasmodium species (P. vivax and P. yoelii). AVAILABILITY: Supplementary Material and a database of the new domains and GO predictions achieved on Plasmodium proteins are available at http://www.lirmm.fr/~terrapon/codd/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Confidence in pairwise alignments of biological sequences, obtained by various methods such as Blast or Smith-Waterman, is critical for automatic analyses of genomic data. Two statistical models have been proposed. In the asymptotic limit of long sequences, the Karlin-Altschul model is based on the computation of a P-value, assuming that the number of high scoring matching regions above a threshold is Poisson distributed. Alternatively, the Lipman-Pearson model is based on the computation of a Z-value from a random score distribution obtained by a Monte-Carlo simulation. Z-values allow the deduction of an upper bound of the P-value (1/Z-value2) following the TULIP theorem. Simulations of Z-value distribution is known to fit with a Gumbel law. This remarkable property was not demonstrated and had no obvious biological support. RESULTS: We built a model of evolution of sequences based on aging, as meant in Reliability Theory, using the fact that the amount of information shared between an initial sequence and the sequences in its lineage (i.e., mutual information in Information Theory) is a decreasing function of time. This quantity is simply measured by a sequence alignment score. In systems aging, the failure rate is related to the systems longevity. The system can be a machine with structured components, or a living entity or population. "Reliability" refers to the ability to operate properly according to a standard. Here, the "reliability" of a sequence refers to the ability to conserve a sufficient functional level at the folded and maturated protein level (positive selection pressure). Homologous sequences were considered as systems 1) having a high redundancy of information reflected by the magnitude of their alignment scores, 2) which components are the amino acids that can independently be damaged by random DNA mutations. From these assumptions, we deduced that information shared at each amino acid position evolved with a constant rate, corresponding to the information hazard rate, and that pairwise sequence alignment scores should follow a Gumbel distribution, which parameters could find some theoretical rationale. In particular, one parameter corresponds to the information hazard rate. CONCLUSION: Extreme value distribution of alignment scores, assessed from high scoring segments pairs following the Karlin-Altschul model, can also be deduced from the Reliability Theory applied to molecular sequences. It reflects the redundancy of information between homologous sequences, under functional conservative pressure. This model also provides a link between concepts of biological sequence analysis and of systems biology.

BACKGROUND: Popular methods to reconstruct molecular phylogenies are based on multiple sequence alignments, in which addition or removal of data may change the resulting tree topology. We have sought a representation of homologous proteins that would conserve the information of pair-wise sequence alignments, respect probabilistic properties of Z-scores (Monte Carlo methods applied to pair-wise comparisons) and be the basis for a novel method of consistent and stable phylogenetic reconstruction. RESULTS: We have built up a spatial representation of protein sequences using concepts from particle physics (configuration space) and respecting a frame of constraints deduced from pair-wise alignment score properties in information theory. The obtained configuration space of homologous proteins (CSHP) allows the representation of real and shuffled sequences, and thereupon an expression of the TULIP theorem for Z-score probabilities. Based on the CSHP, we propose a phylogeny reconstruction using Z-scores. Deduced trees, called TULIP trees, are consistent with multiple-alignment based trees. Furthermore, the TULIP tree reconstruction method provides a solution for some previously reported incongruent results, such as the apicomplexan enolase phylogeny. CONCLUSION: The CSHP is a unified model that conserves mutual information between proteins in the way physical models conserve energy. Applications include the reconstruction of evolutionary consistent and robust trees, the topology of which is based on a spatial representation that is not reordered after addition or removal of sequences. The CSHP and its assigned phylogenetic topology, provide a powerful and easily updated representation for massive pair-wise genome comparisons based on Z-score computations.

MOTIVATION: Different automatic methods of sequence alignments are routinely used as a starting point for homology searches and function inference. Confidence in an alignment probability is one of the major fundamentals of massive automatic genome-scale pairwise comparisons, for clustering of putative orthologs and paralogs, sequenced genome annotation or multiple-genomic tree constructions. Extreme value distribution based on the Karlin-Altschul model, usually advised for large-scale comparisons are not always valid, particularly in the case of comparisons of non-biased with nucleotide-biased genomes (such that of Plasmodium falciparum). Z-values estimates based on Monte Carlo technics, can be calculated experimentally for any alignment output, whatever the method used. Empirically, a Z-value higher than approximately 8 is supposed reasonable to assess that an alignment score is significant, but this arbitrary figure was never theoretically justified. RESULTS: In this paper, we used the Bienaymé-Chebyshev inequality to demonstrate a theorem of the upper limit of an alignment score probability (or P-value). This theorem implies that a computed Z-value is a statistical test, a single-linkage clustering criterion and that 1/Z-value(2) is an upper limit to the probability of an alignment score whatever the actual probability law is. Therefore, this study provides the missing theoretical link between a Z-value cut-off used for an automatic clustering of putative orthologs and/or paralogs, and the corresponding statistical risk in such genome-scale comparisons (using non-biased or biased genomes).