A new hidden Markov model method (SAM-T98) for nding remote homologs of protein sequences is described and evaluated. The method begins with a single target sequence and iteratively builds a hidden Markov model (hmm) from the sequence and homologs found using the hmm for database search. SAM-T98 is also used to construct model libraries automatically from sequences in structural databases. We evaluate the SAM-T98 method with four datasets. Three of the test sets are fold-recognition tests, where the correct answers are determined by structural similarity. The fourth uses a curated database. The method is compared against wu-blastp and against double-blast, a two-step method similar to ISS, but using blast instead of fasta. Results SAM-T98 had the fewest errors in all tests| dramatically so for the fold-recognition tests. At the minimum-error point on the SCOP-domains test, SAM-T98 got 880 true positives and 68 false positives, double-blast got 533 true positives with 71 false positives, and wu-blastp got 353 true positives with 24 false positives. The method is optimized to recognize superfamilies, and would require parameter adjustment to be used to nd family or fold relationships. One key to the performance of the hmm method is a new score-normalization technique that compares the score to the score with a reversed model rather than to a uniform null model. Availability A World Wide Web server, as well as information on obtaining the Sequence Alignment and PREPRINT to appear in Bioinformatics, 1999

A new method for detecting remote protein homologies is introduced and shown to perform well in classifying protein domains by SCOP superfamily. The method is a variant of support vector machines using a new kernel function. The kernel function is derived from a generative statistical model for a protein family, in this case a hidden Markov model. This general approach of combining generative models like HMMs with discriminative methods such as support vector machines may have applications in other areas of biosequence analysis as well.

The sequences of related proteins can diverge beyond the point where their relationship can be recognised by pairwise sequence comparisons. In attempts to overcome this limitation, methods have been developed that use as a query, not a single sequence, but sets of related sequences or a representation of the characteristics shared by related sequences. Here we describe an assessment of three of these methods: the SAM-T98 implementation of a hidden Markov model procedure; PSI-BLAST; and the intermediate sequence search (ISS) procedure. We determined the extent to which these procedures can detect evolutionary relationships between the members of the sequence database PDBD40-J. This database, derived from the structural classification of proteins (SCOP), contains the sequences of proteins of known structure whose sequence identities with each other are 40 % or less. The evolutionary relationships that exist between those that have low sequence identities were found by the examination of their structural details and, in many cases, their functional

Hidden Markov models (HMMs) are a highly effective means of modeling a family of unaligned sequences or a common motif within a set of unaligned sequences. The trained HMM can then be used for discrimination or multiple alignment. The basic mathematical description of an HMM and its expectation-maximization training procedure is relatively straight-forward. In this paper, we review the mathematical extensions and heuristics that move the method from the theoretical to the practical. Then, we experimentally analyze the effectiveness of model regularization, dynamic model modification, and optimization strategies. Finally it is demonstrated on the SH2 domain how a domain can be found from unaligned sequences using a special model type. The experimental work was completed with the aid of the Sequence Alignment and Modeling software suite. 1 Introduction Since their introduction to the computational biology community (Haussler et al., 1993; Krogh et al., 1994a), hidden Markov models (HMMs...

A new method, called the Fisher kernel method, for detecting remote protein homologies is introduced and shown to perform well in classifying protein domains by SCOP superfamily. The method is a variant of support vector machines using a new kernel function. The kernel function is derived from a hidden Markov model. The general approach of combining generative models like HMMs with discriminative methods such as support vector machines may have applications in other areas of biosequence analysis as well.

This paper presents the mathematical foundations of Dirichlet mixtures, which have been used to improve database search results for homologous sequences, when a variable number of sequences from a protein family or domain are known. We present a method for condensing the information in a protein database into a mixture of Dirichlet densities. These mixtures are designed to be combined with observed amino acid frequencies, to form estimates of expected amino acid probabilities at each position in a profile, hidden Markov model, or other statistical model. These estimates give a statistical model greater generalization capacity, such that remotely related family members can be more reliably recognized by the model. Dirichlet mixtures have been shown to outperform substitution matrices and other methods for computing these expected amino acid distributions in database search, resulting in fewer false positives and false negatives for the families tested. This paper corrects a previously p...

Motivation: Modeling families of related biological sequences using Hidden Markov models (HMMs), although increasingly widespread, faces at least one major problem: because of the complexity of these mathematical models, they require a relatively large training set in order to accurately characterize a given family. For families in which there are few known sequences, a standard linear HMM contains too many parameters to be trained adequately. Results: This work attempts to solve that problem by generating smaller HMMs which precisely model only the conserved regions of the family. These HMMs are constructed from motif models generated by the EM algorithm using the MEME software. Because motif-based HMMs have relatively few parameters, they can be trained using smaller data sets. Studies of shortchain alcohol dehydrogenases and 4Fe-4S ferredoxins support the claim that motif-based HMMs exhibit increased sensitivity and selectivity in database searches, especially when training sets co...

The ProteinIDDqNDA ion Resource(PIu is an integrated public resource of protein informatics that supports genomic and proteomic research and scientific discovery.PI maintains the Protein Sequence Database (PSD), an annotated protein database containing over 283 000 sequences covering the entire taxonomic range. Family classification is used for sensitive identification, consistent annotation, and detection of annotation errors. The superfamily curation defines signature domain architecture and categorizes memberships to improve automated classification. To increase the amount of experimental annotation, the PI has developed a bibliography system for literature searching, mapping, and user submission, and has conducted retrospective attribution of citations for experimental features. PI also maintains NREF, a non-redundant reference database, and iProClass, an integrated database of protein family, function, and structure information.PI-NREF provides a timely and comprehensive collection of protein sequences, currently consisting of more than 1 000 000 entries fromPI -PSD, SWIA208 OT, TrEMBL, RefSeq, GenPept, and PDB. The PI web site (http://pir.georgetown.edu) connects data analysis tools to underlying databases for information retrieval and knowledge discovery, with functionalities for interactive queries, combinations of sequence and text searches, and sorting and visual exploration of search results. The FTP site provides free download for PSD and NREF biweekly releases and auxiliary databases and files.

The Protein Information Resource (PIR) serves as an integrated public resource of functional annotation of protein data to support genomic/proteomic research and scientific discovery. The PIR, in collaboration with the Munich Information Center for Protein Sequences (MIPS) and the Japan International Protein Information Database (JIPID), produces the PIR-International Protein Sequence Database (PSD), the major annotated protein sequence database in the public domain, containing about 250 000 proteins. To improve protein annotation and the coverage of experimentally validated data, a bibliography submission system is developed for scientists to submit, categorize and retrieve literature information. Comprehensive protein information is available from i ProClass, which includes family classification at the superfamily, domain and motif levels, structural and functional features of proteins, as well as crossreferences to over 40 biological databases. To provide timely and comprehensive protein data with source attribution, we have introduced a non-redundant reference protein database, PIR-NREF. The database consists of about 800 000 proteins collected from PIR-PSD, SWISS-PROT, TrEMBL, GenPept, RefSeq and PDB, with composite protein names and literature data. To promote database interoperability, we provide XML data distribution and open database schema, and adopt common ontologies. The PIR web site (http://pir.georgetown.edu/) features data mining and sequence analysis tools for information retrieval and functional identification of proteins based on both sequence and annotation information. The PIR databases and other files are also available by FTP (ftp://nbrfa.georgetown.edu/pir_databases).

Motivation: Statistical sequence comparison techniques, such as hidden Markov models and generalized pro les, calculate the probability that a sequence was generated by a given model. Log-odds scoring is a means of evaluating this probability by comparing it to a null hypothesis, usually a simpler statistical model intended to represent the universe of sequences as a whole, rather than the group of interest. Such scoring leads to two immediate questions: what should the null model be, and what threshold of log-odds score should be deemed a match to the model. Results: This paper experimentally analyses these two issues. Within the context of the Sequence Alignment and Modeling software suite (SAM), we consider a variety ofnull models and suitable thresholds. Additionally, we consider HMMer's log-odds scoring and SAM's original Zscoring method. Among the null model choices, a simple looping null model that emits characters according to the geometric mean of the character probabilities in the columns modeled by the HMM performs well or best across all four discrimination experiments.