Ortholog identification is used in gene functional annotation, species phylogeny estimation, phylogenetic profile construction and many other analyses. Bioinformatics methods for ortholog identification are commonly based on pairwise protein sequence comparisons between whole genomes. Phylogenetic methods of ortholog identification have also been developed; these methods can be applied to protein data sets sharing a common domain architecture or which share a single functional domain but differ outside this region of homology. While promiscuous domains represent a challenge to all orthology prediction methods, overall structural similarity is highly correlated with proximity in a phylogenetic tree, conferring a degree of robustness to phylogenetic methods. In this article, we review the issues involved in orthology prediction when data sets include sequences with structurally heterogeneous domain architectures, with particular attention to automated methods designed for high-throughput application, and present a case study to illustrate the challenges in this area.

MOTIVATION: The identification of catalytic residues is a key step in understanding the function of enzymes. While a variety of computational methods have been developed for this task, accuracies have remained fairly low. The best existing method exploits information from sequence and structure to achieve a precision (the fraction of predicted catalytic residues that are catalytic) of 18.5% at a corresponding recall (the fraction of catalytic residues identified) of 57% on a standard benchmark. Here we present a new method, Discern, which provides a significant improvement over the state-of-the-art through the use of statistical techniques to derive a model with a small set of features that are jointly predictive of enzyme active sites. RESULTS: In cross-validation experiments on two benchmark datasets from the Catalytic Site Atlas and CATRES resources containing a total of 437 manually curated enzymes spanning 487 SCOP families, Discern increases catalytic site recall between 12% and 20% over methods that combine information from both sequence and structure, and by >or=50% over methods that make use of sequence conservation signal only. Controlled experiments show that Discern's improvement in catalytic residue prediction is derived from the combination of three ingredients: the use of the INTREPID phylogenomic method to extract conservation information; the use of 3D structure data, including features computed for residues that are proximal in the structure; and a statistical regularization procedure to prevent overfitting.

Structural phylogenomics refers to the combined use of evolutionary and structural information in a bioinformatics analysis. The term phylogenomics refers to two distinct tasks: reconstructing a species phylogeny using multiple genes (for a review, see [1]) and predicting protein function by estimating the evolutionary history of a family of related sequences (i.e., a gene tree or multi-gene tree including gene duplication events) [2]–[4]. In this “Getting Started” article, we focus on the latter task, restricting our discussion to the construction and analysis of phylogenetic trees for amino acid data, and including protein structure data and structure prediction to improve the accuracy of functional annotation. We address the following questions: Why perform a complicated structural phylogenomic analysis when simpler approaches are available? What are the fundamental underlying assumptions of this approach, and what are the implications of any conflicts with these assumptions? What technical challenges do we need to address to achieve the full potential of these ideas?

BACKGROUND: Identifying the catalytic residues in enzymes can aid in understanding the molecular basis of an enzyme's function and has significant implications for designing new drugs, identifying genetic disorders, and engineering proteins with novel functions. Since experimentally determining catalytic sites is expensive, better computational methods for identifying catalytic residues are needed. RESULTS: We propose ResBoost, a new computational method to learn characteristics of catalytic residues. The method effectively selects and combines rules of thumb into a simple, easily interpretable logical expression that can be used for prediction. We formally define the rules of thumb that are often used to narrow the list of candidate residues, including residue evolutionary conservation, 3D clustering, solvent accessibility, and hydrophilicity. ResBoost builds on two methods from machine learning, the AdaBoost algorithm and Alternating Decision Trees, and provides precise control over the inherent trade-off between sensitivity and specificity. We evaluated ResBoost using cross-validation on a dataset of 100 enzymes from the hand-curated Catalytic Site Atlas (CSA). CONCLUSION: ResBoost achieved 85% sensitivity for a 9.8% false positive rate and 73% sensitivity for a 5.7% false positive rate. ResBoost reduces the number of false positives by up to 56% compared to the use of evolutionary conservation scoring alone. We also illustrate the ability of ResBoost to identify recently validated catalytic residues not listed in the CSA.

MOTIVATION: Identification of functionally important residues in proteins plays a significant role in biological discovery. Here, we present INTREPID--an information-theoretic approach for functional site identification that exploits the information in large diverse multiple sequence alignments (MSAs). INTREPID uses a traversal of the phylogeny in combination with a positional conservation score, based on Jensen-Shannon divergence, to rank positions in an MSA. While knowledge of protein 3D structure can significantly improve the accuracy of functional site identification, since structural information is not available for a majority of proteins, INTREPID relies solely on sequence information. We evaluated INTREPID on two tasks: predicting catalytic residues and predicting specificity determinants. RESULTS: In catalytic residue prediction, INTREPID provides significant improvements over Evolutionary Trace, ConSurf as well as over a baseline global conservation method on a set of 100 manually curated enzymes from the Catalytic Site Atlas. In particular, INTREPID is able to better predict catalytic positions that are not globally conserved and hence, attains improved sensitivity at high values of specificity. We also investigated the performance of INTREPID as a function of the evolutionary divergence of the protein family. We found that INTREPID is better able to exploit the diversity in such families and that accuracy improves when homologs with very low sequence identity are included in an alignment. In specificity determinant prediction, when subtype information is known, INTREPID-SPEC, a variant of INTREPID, attains accuracies that are competitive with other approaches for this task. AVAILABILITY: INTREPID is available for 16919 families in the PhyloFacts resource (http://phylogenomics.berkeley.edu/phylofacts).

Function prediction by homology is widely used to provide preliminary functional annotations for genes for which experimental evidence of function is unavailable or limited. This approach has been shown to be prone to systematic error, including percolation of annotation errors through sequence databases. Phylogenomic analysis avoids these errors in function prediction but has been difficult to automate for high-throughput application. To address this limitation, we present a computationally efficient pipeline for phylogenomic classification of proteins. This pipeline uses the SCI-PHY (Subfamily Classification in Phylogenomics) algorithm for automatic subfamily identification, followed by subfamily hidden Markov model (HMM) construction. A simple and computationally efficient scoring scheme using family and subfamily HMMs enables classification of novel sequences to protein families and subfamilies. Sequences representing entirely novel subfamilies are differentiated from those that can be classified to subfamilies in the input training set using logistic regression. Subfamily HMM parameters are estimated using an information-sharing protocol, enabling subfamilies containing even a single sequence to benefit from conservation patterns defining the family as a whole or in related subfamilies. SCI-PHY subfamilies correspond closely to functional subtypes defined by experts and to conserved clades found by phylogenetic analysis. Extensive comparisons of subfamily and family HMM performances show that subfamily HMMs dramatically improve the separation between homologous and non-homologous proteins in sequence database searches. Subfamily HMMs also provide extremely high specificity of classification and can be used to predict entirely novel subtypes. The SCI-PHY Web server at http://phylogenomics.berkeley.edu/SCI-PHY/ allows users to upload a multiple sequence alignment for subfamily identification and subfamily HMM construction. Biologists wishing to provide their own subfamily definitions can do so. Source code is available on the Web page. The Berkeley Phylogenomics Group PhyloFacts resource contains pre-calculated subfamily predictions and subfamily HMMs for more than 40,000 protein families and domains at http://phylogenomics.berkeley.edu/phylofacts/.

Phylogenomic inference of protein (or gene) function attempts to address the question, “What function does this protein perform?” in an evolutionary context. As originally outlined by Jonathan Eisen [1–3], phylogenomic inference of protein function is a multistep process involving selection of homologs, multiple sequence alignment (MSA), and phylogenetic tree construction; overlaying annotations on the tree topology; discriminating between orthologs and paralogs; and—finally—inferring the function of a protein based on the orthologs identified by this process and the annotations retrieved. Figure 1 shows an example of using annotated subfamily groupings to infer function, in a manner similar to [1]. One of us, while at Celera Genomics, separately came up with a similar approach for the functional classification of the human genome [4], based on the automated identification of functional subfamilies using the SCI-PHY algorithm and the use of subfamily hidden Markov models (HMMs) to classify novel sequences [5,6]. Our experiences over the past several years in developing computational pipelines for automating phylogenomic inference at the genome scale [7]—and the challenges we have faced in this effort—motivate this paper. Figure 1 Phylogenomic Analysis of Protein Function Using Subfamily Annotation In practice, phylogenomic inference of gene function is not often used. Far from it. The majority of novel sequences are assigned a putative function through the use of annotation transfer from the top hits in a database search. In our analysis of over 300,000 proteins in the UniProt database, only 3% of proteins with informative annotations (i.e., those not labelled as “hypothetical” or “unknown”) had experimental support for their annotations; 97% were annotated using electronic evidence alone. These annotations are uploaded to GenBank, where they persist even if they are eventually determined to be in error. The systematic errors associated with this annotation protocol have been pointed out by numerous investigators over the years [8–10]. The root causes of these errors are these: Gene duplication. This enables protein superfamilies to innovate novel functions on the same structural template, so that the top database hit may have a function distinct from the query. Domain shuffling. Domain fusion and fission events add an additional layer of complexity, as a query and database hit may share only a local region of homology and thus have entirely different molecular functions and structures. Propagation of existing errors in database annotations. This is particularly pernicious, as existing annotation errors are seldom detected and, even if detected, are not necessarily corrected. Evolutionary distance. Two proteins can share a common ancestor and domain structure, yet have very different functions simply due to their presence in very divergently related species. Phylogenomic analysis, properly applied, avoids these errors and provides a mechanism for detecting existing database annotation errors [3,7]. Why then is phylogenomic inference not used more widely? We believe this is due to four reasons. First, the actual frequency of annotation error is not known, so the gravity of the situation is not recognized. Second, phylogenomic inference is a much more complicated endeavor than a simple database search and requires significantly more expertise and computing resources. It is therefore not easily applied at the genome scale. Third, millions of dollars and years of effort have been poured into developing computational annotation systems that depend on annotation transfer from top database hits, perhaps overlaid with domain prediction methods such as PFAM or the NCBI CDD [11,12]. Fourth, phylogenomic approaches to protein function prediction have arisen only in the last few years, while database search methods have been available for much longer. Revolutions do not normally take place overnight. These four reasons result in phylogenomic inference being applied on a one-off basis, for a few protein superfamilies here and there. This may be about to change. A variety of software tools and algorithms enabling phylogenomic inference have been developed in recent years (see Table 1). Some of these methods have based annotation transfer on the identification of orthologs [13–15] or of functional subfamilies [6,16–21]. Other groups have used whole-tree analyses [22–24]. Still other groups employ expert knowledge to define functional subtypes and then develop statistical models to allow users to classify novel sequences [25,26]; these expert system-based approaches are unfortunately limited by the scarcity of experimental data for most protein families. Table 1 Resources for Phylogenomic Analysis It is worth examining the assumptions underlying these phylogenomic resources, and phylogenomic inference as a whole.

MOTIVATION: In recent years, several methods have been proposed for aligning two protein sequence profiles, with reported improvements in alignment accuracy and homolog discrimination versus sequence-sequence methods (e.g. BLAST) and profile-sequence methods (e.g. PSI-BLAST). Profile-profile alignment is also the iterated step in progressive multiple sequence alignment algorithms such as CLUSTALW. However, little is known about the relative performance of different profile-profile scoring functions. In this work, we evaluate the alignment accuracy of 23 different profile-profile scoring functions by comparing alignments of 488 pairs of sequences with identity < or =30% against structural alignments. We optimize parameters for all scoring functions on the same training set and use profiles of alignments from both PSI-BLAST and SAM-T99. Structural alignments are constructed from a consensus between the FSSP database and CE structural aligner. We compare the results with sequence-sequence and sequence-profile methods, including BLAST and PSI-BLAST. RESULTS: We find that profile-profile alignment gives an average improvement over our test set of typically 2-3% over profile-sequence alignment and approximately 40% over sequence-sequence alignment. No statistically significant difference is seen in the relative performance of most of the scoring functions tested. Significantly better results are obtained with profiles constructed from SAM-T99 alignments than from PSI-BLAST alignments. AVAILABILITY: Source code, reference alignments and more detailed results are freely available at http://phylogenomics.berkeley.edu/profilealignment/

MOTIVATION: Alignments of two multiple-sequence alignments, or statistical models of such alignments (profiles), have important applications in computational biology. The increased amount of information in a profile versus a single sequence can lead to more accurate alignments and more sensitive homolog detection in database searches. Several profile-profile alignment methods have been proposed and have been shown to improve sensitivity and alignment quality compared with sequence-sequence methods (such as BLAST) and profile-sequence methods (e.g. PSI-BLAST). Here we present a new approach to profile-profile alignment we call Comparison of Alignments by Constructing Hidden Markov Models (HMMs) (COACH). COACH aligns two multiple sequence alignments by constructing a profile HMM from one alignment and aligning the other to that HMM. RESULTS: We compare the alignment accuracy of COACH with two recently published methods: Yona and Levitt's prof_sim and Sadreyev and Grishin's COMPASS. On two sets of reference alignments selected from the FSSP database, we find that COACH is able, on average, to produce alignments giving the best coverage or the fewest errors, depending on the chosen parameter settings. AVAILABILITY: COACH is freely available from www.drive5.com/lobster

MOTIVATION: Protein families evolve a multiplicity of functions through gene duplication, speciation and other processes. As a number of studies have shown, standard methods of protein function prediction produce systematic errors on these data. Phylogenomic analysis--combining phylogenetic tree construction, integration of experimental data and differentiation of orthologs and paralogs--has been proposed to address these errors and improve the accuracy of functional classification. The explicit integration of structure prediction and analysis in this framework, which we call structural phylogenomics, provides additional insights into protein superfamily evolution. RESULTS: Results of protein functional classification using phylogenomic analysis show fewer expected false positives overall than when pairwise methods of functional classification are employed. We present an overview of the motivations and fundamental principles of phylogenomic analysis, new methods developed for the key tasks, benchmark datasets for these tasks (when available) and suggest procedures to increase accuracy. We also discuss some of the methods used in the Celera Genomics high-throughput phylogenomic classification of the human genome. AVAILABILITY: Software tools from the Berkeley Phylogenomics Group are available at http://phylogenomics.berkeley.edu

MOTIVATION: Aligning multiple proteins based on sequence information alone is challenging if sequence identity is low or there is a significant degree of structural divergence. We present a novel algorithm (SATCHMO) that is designed to address this challenge. SATCHMO simultaneously constructs a tree and a set of multiple sequence alignments, one for each internal node of the tree. The alignment at a given node contains all sequences within its sub-tree, and predicts which positions in those sequences are alignable and which are not. Aligned regions therefore typically get shorter on a path from a leaf to the root as sequences diverge in structure. Current methods either regard all positions as alignable (e.g. ClustalW), or align only those positions believed to be homologous across all sequences (e.g. profile HMM methods); by contrast SATCHMO makes different predictions of alignable regions in different subgroups. SATCHMO generates profile hidden Markov models at each node; these are used to determine branching order, to align sequences and to predict structurally alignable regions. RESULTS: In experiments on the BAliBASE benchmark alignment database, SATCHMO is shown to perform comparably to ClustalW and the UCSC SAM HMM software. Results using SATCHMO to identify protein domains are demonstrated on potassium channels, with implications for the mechanism by which tumor necrosis factor alpha affects potassium current. AVAILABILITY: The software is available for download from http://www.drive5.com/lobster/index.htm

We present a method for condensing the information in multiple alignments of proteins into a mixture of Dirichlet densities over amino acid distributions. Dirichlet mixture densities 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, so that remotely related family members can be more reliably recognized by the model. This paper corrects the previously published formula for estimating these expected probabilities, and contains complete derivations of the Dirichlet mixture formulas, methods for optimizing the mixtures to match particular databases, and suggestions for efficient implementation.