Computational protein function prediction: are we making progress?

The computational prediction of gene and protein function is rapidly gaining ground as a central undertaking in computational biology. Making sense of the flood of genomic data requires fast and reliable annotation. Many ingenious algorithms have been devised to infer a protein's function from its amino acid sequence, 3D structure and chromosomal location of the encoding genes. However, there are significant challenges in assessing how well these programs perform. In this article we explore those challenges and review our own attempt at assessing the performance of those programs. We conclude that the task is far from complete and that a critical assessment of the performance of function prediction programs is necessary to make true progress in computational function prediction.

The status of structural genomics defined through the analysis of current targets and structures.

Structural genomics--large-scale macromolecular 3-dimenional structure determination--is unique in that major participants report scientific progress on a weekly basis. The target database (TargetDB) maintained by the Protein Data Bank (http://targetdb.pdb.org) reports this progress through the status of each protein sequence (target) under consideration by the major structural genomics centers worldwide. Hence, TargetDB provides a unique opportunity to analyze the potential impact that this major initiative provides to scientists interested in the sequence-structure-function-disease paradigm. Here we report such an analysis with a focus on: (i) temporal characteristics--how is the project doing and what can we expect in the future? (ii) target characteristics--what are the predicted functions of the proteins targeted by structural genomics and how biased is the target set when compared to the PDB and to predictions across complete genomes? (iii) structures solved--what are the characteristics of structures solved thus far and what do they contribute? The analysis required a more extensive database of structure predictions using different methods integrated with data from other sources. This database, associated tools and related data sources are available from http://spam.sdsc.edu.

The apoptosis database.

The apoptosis database is a public resource for researchers and students interested in the molecular biology of apoptosis. The resource provides functional annotation, literature references, diagrams/images, and alternative nomenclatures on a set of proteins having 'apoptotic domains'. These are the distinctive domains that are often, if not exclusively, found in proteins involved in apoptosis. The initial choice of proteins to be included is defined by apoptosis experts and bioinformatics tools. Users can browse through the web accessible lists of domains, proteins containing these domains and their associated homologs. The database can also be searched by sequence homology using basic local alignment search tool, text word matches of the annotation, and identifiers for specific records. The resource is available at http://www.apoptosis-db.org and is updated on a regular basis.

Analysis of the mouse transcriptome based on functional annotation of 60,770 full-length cDNAs.

Only a small proportion of the mouse genome is transcribed into mature messenger RNA transcripts. There is an international collaborative effort to identify all full-length mRNA transcripts from the mouse, and to ensure that each is represented in a physical collection of clones. Here we report the manual annotation of 60,770 full-length mouse complementary DNA sequences. These are clustered into 33,409 'transcriptional units', contributing 90.1% of a newly established mouse transcriptome database. Of these transcriptional units, 4,258 are new protein-coding and 11,665 are new non-coding messages, indicating that non-coding RNA is a major component of the transcriptome. 41% of all transcriptional units showed evidence of alternative splicing. In protein-coding transcripts, 79% of splice variations altered the protein product. Whole-transcriptome analyses resulted in the identification of 2,431 sense-antisense pairs. The present work, completely supported by physical clones, provides the most comprehensive survey of a mammalian transcriptome so far, and is a valuable resource for functional genomics.

Fold predictions for bacterial genomes.

Fold assignments for newly sequenced genomes belong to the most important and interesting applications of the booming field of protein structure prediction. We present a brief survey and a discussion of such assignments completed to date, using as an example several fold assignment projects for proteins from the Escherichia coli genome. This review focuses on steps that are necessary to go beyond the simple assignment projects and into the development of tools extending our understanding of functions of proteins in newly sequenced genomes. This paper also discusses several problems seldom addressed in the literature, such as the problem of domain prediction and complementary predictions (e.g., transmembrane regions and flexible regions) and cross-correlation of predictions from different servers. The influence of sequence and structure database growth on prediction success is also addressed. Finally, we discuss the perspectives of the field in the context of massive sequence and structure determination projects, as well as the development of novel prediction methods.

CLAN, a novel human CED-4-like gene.

Proteins governing cell death form the basis of many normal processes and contribute to the pathogenesis of many diseases when dysregulated. Here we report the cloning of a novel human CED-4-like gene, CLAN, and several of its alternatively spliced isoforms. These caspase-associated recruitment domain (CARD)-containing proteins are expressed at varying degrees in normal human tissues and may contribute to a number of intracellular processes including apoptosis, cytokine processing, and NF-kappa B activation. The CARD of the CLAN proteins binds a number of other CARD-containing proteins including caspase-1, BCL10, NOD2, and NAC. Once their physiologic functions are uncovered, CLAN proteins may prove to be valuable therapeutic targets.

Characterization of matrix metalloproteinase-26, a novel metalloproteinase widely expressed in cancer cells of epithelial origin.

Identification of expanding roles for matrix metalloproteinases (MMPs) in complex regulatory processes of tissue remodelling has stimulated the search for genes encoding proteinases with unique functions, regulation and expression patterns. By using a novel cloning strategy, we identified three previously unknown human MMPs, i.e. MMP-21, MMP-26 and MMP-28, in comprehensive gene libraries. The present study is focused on the gene and the protein of a novel MMP, MMP-26. Our findings show that MMP-26 is specifically expressed in cancer cells of epithelial origin, including carcinomas of lung, prostate and breast. Several unique structural and regulatory features, including an unusual 'cysteine-switch' motif, discriminate broad-spectrum MMP-26 from most other MMPs. MMP-26 efficiently cleaves fibrinogen and extracellular matrix proteins, including fibronectin, vitronectin and denatured collagen. Protein sequence, minimal modular domain structure, exon-intron mapping and computer modelling demonstrate similarity between MMP-26 and MMP-7 (matrilysin). However, substrate specificity and transcriptional regulation, as well as the functional role of MMP-26 and MMP-7 in cancer, are likely to be distinct. Despite these differences, matrilysin-2 may be a suitable trivial name for MMP-26. Our observations suggest an important specific function for MMP-26 in tumour progression and angiogenesis, and confirm and extend the recent findings of other authors [Park, Ni, Gerkema, Liu, Belozerov and Sang (2000) J. Biol. Chem. 275, 20540--20544; Uría and López-Otín (2000) Cancer Res. 60, 4745--4751; de Coignac, Elson, Delneste, Magistrelli, Jeannin, Aubry, Berthier, Schmitt, Bonnefoy and Gauchat (2000) Eur. J. Biochem. 267, 3323--3329].

Surface map comparison: studying function diversity of homologous proteins.

A simplified protein surface cartography approach has been developed to assist in the analysis of surface features in homologous families, and thus to predict conservation or divergence of protein functions and protein-protein interaction patterns. A spherical approximation of protein surface was used, with a focus on charged and hydrophobic residues. The resulting surface map allows for qualitative analysis and comparison of surfaces of proteins, but can also be used to define a simple numerical measure of map similarity between two or more proteins. The latter was shown to be useful for function based classifications within large protein families. Surface map analysis was tested on several test cases: haemoglobins, death domains and TRAF domains. It was shown that surface map comparison allows a better function prediction than general sequence analysis methods and can reproduce known examples of functional variation within a divergent group of proteins. In another example, we predict novel, unexpected sets of common functional properties for seemingly distant members of a large group of divergent proteins. The method was also shown to be robust enough to allow using protein models from comparative modelling instead of experimental structures.

TUCAN, an antiapoptotic caspase-associated recruitment domain family protein overexpressed in cancer.

Caspase-associated recruitment domains (CARDs) are protein interaction domains that participate in activation or suppression of CARD-carrying members of the caspase family of apoptosis-inducing proteases. A novel CARD-containing protein was identified that is overexpressed in some types of cancer and that binds and suppresses activation of procaspase-9, which we term TUCAN (tumor-up-regulated CARD-containing antagonist of caspase nine). The CARD domain of TUCAN selectively binds itself and procaspase-9. TUCAN interferes with binding of Apaf1 to procaspase-9 and suppresses caspase activation induced by the Apaf1 activator, cytochrome c. Overexpression of TUCAN in cells by stable or transient transfection inhibits apoptosis and caspase activation induced by Apaf1/caspase-9-dependent stimuli, including Bax, VP16, and staurosporine, but not by Apaf1/caspase-9-independent stimuli, Fas and granzyme B. High levels of endogenous TUCAN protein were detected in several tumor cell lines and in colon cancer specimens, correlating with shorter patient survival. Thus, TUCAN represents a new member of the CARD family that selectively suppresses apoptosis induced via the mitochondrial pathway for caspase activation.

Clustering of highly homologous sequences to reduce the size of large protein databases.

We present a fast and flexible program for clustering large protein databases at different sequence identity levels. It takes less than 2 h for the all-against-all sequence comparison and clustering of the non-redundant protein database of over 560,000 sequences on a high-end PC. The output database, including only the representative sequences, can be used for more efficient and sensitive database searches.

A diverse family of proteins containing tumor necrosis factor receptor-associated factor domains.

We have identified three new tumor necrosis factor-receptor associated factor (TRAF) domain-containing proteins in humans using bioinformatics approaches, including: MUL, the product of the causative gene in Mulibrey Nanism syndrome; USP7 (HAUSP), an ubiquitin protease; and SPOP, a POZ domain-containing protein. Unlike classical TRAF family proteins involved in TNF family receptor (TNFR) signaling, the TRAF domains (TDs) of MUL, USP7, and SPOP are located near the NH(2) termini or central region of these proteins, rather than carboxyl end. MUL and USP7 are capable of binding in vitro via their TDs to all of the previously identified TRAF family proteins (TRAF1, TRAF2, TRAF3, TRAF4, TRAF5, and TRAF6), whereas the TD of SPOP interacts weakly with TRAF1 and TRAF6 only. The TD of MUL also interacted with itself, whereas the TDs of USP7 and SPOP did not self-associate. Analysis of various MUL and USP7 mutants by transient transfection assays indicated that the TDs of these proteins are necessary and sufficient for suppressing NF-kappaB induction by TRAF2 and TRAF6 as well as certain TRAF-binding TNF family receptors. In contrast, the TD of SPOP did not inhibit NF-kappaB induction. Immunofluorescence confocal microscopy indicated that MUL localizes to cytosolic bodies, with targeting to these structures mediated by a RBCC tripartite domain within the MUL protein. USP7 localized predominantly to the nucleus, in a TD-dependent manner. Data base searches revealed multiple proteins containing TDs homologous to those found in MUL, USP7, and SPOP throughout eukaryotes, including yeast, protists, plants, invertebrates, and mammals, suggesting that this branch of the TD family arose from an ancient gene. We propose the moniker TEFs (TD-encompassing factors) for this large family of proteins.

Bcl-B, a novel Bcl-2 family member that differentially binds and regulates Bax and Bak.

A novel human member of the Bcl-2 family was identified, Bcl-B, which is closest in amino acid sequence homology to the Boo (Diva) protein. The Bcl-B protein contains four Bcl-2 homology (BH) domains (BH1, BH2, BH3, BH4) and a predicted carboxyl-terminal transmembrane (TM) domain. The BCL-B mRNA is widely expressed in adult human tissues. The Bcl-B protein binds Bcl-2, Bcl-X(L), and Bax but not Bak. In transient transfection assays, Bcl-B suppresses apoptosis induced by Bax but not Bak. Deletion of the TM domain of Bcl-B impairs its association with intracellular organelles and diminishes its anti-apoptotic function. Bcl-B thus displays a unique pattern of selectivity for binding and regulating the function of other members of the Bcl-2 family.

PAAD - a new protein domain associated with apoptosis, cancer and autoimmune diseases.

A new protein domain was found in several proteins involved in apoptosis, inflammation, cancer and immune responses. Its location within these proteins and predicted fold suggests that it functions as a protein-protein interaction domain, possibly uniting different signaling pathways.

Bcl-G, a novel pro-apoptotic member of the Bcl-2 family.

A new member of the Bcl-2 family was identified, Bcl-G. The human BCL-G gene consists of 6 exons, resides on chromosome 12p12, and encodes two proteins through alternative mRNA splicing, Bcl-G(L) (long) and Bcl-G(S) (short) consisting of 327 and 252 amino acids in length, respectively. Bcl-G(L) and Bcl-G(S) have identical sequences for the first 226 amino acids but diverge thereafter. Among the Bcl-2 homology (BH) domains previously recognized in Bcl-2 family proteins, the BH3 domain is found in both Bcl-G(L) and Bcl-G(S), but only the longer Bcl-G(L) protein possesses a BH2 domain. Bcl-G(L) mRNA is expressed widely in adult human tissues, whereas Bcl-G(S) mRNA was found only in testis. Overexpression of Bcl-G(L) or Bcl-G(S) in cells induced apoptosis although Bcl-G(S) was far more potent than Bcl-G(L). Apoptosis induction by Bcl-G(S) depended on the BH3 domain and was suppressed by coexpression of anti-apoptotic Bcl-X(L) protein. Bcl-X(L) also coimmunoprecipitated with Bcl-G(S) but not with mutants of Bcl-G(S) in which the BH3 domain was deleted or mutated or with Bcl-G(L). Bcl-G(S) was predominantly localized to cytosolic organelles, whereas Bcl-G(L) was diffusely distributed throughout the cytosol. A mutant of Bcl-G(L) in which the BH2 domain was deleted displayed increased apoptotic activity and coimmunoprecipitated with Bcl-X(L), suggesting that the BH2 domain autorepresses Bcl-G(L).

A novel enhancer of the Apaf1 apoptosome involved in cytochrome c-dependent caspase activation and apoptosis.

Apaf1/CED4 family members play central roles in apoptosis regulation as activators of caspase family cell death proteases. These proteins contain a nucleotide-binding (NB) self-oligomerization domain and a caspase recruitment domain (CARD). A novel human protein was identified, NAC, that contains an NB domain and CARD. The CARD of NAC interacts selectively with the CARD domain of Apaf1, a caspase-activating protein that couples mitochondria-released cytochrome c (cyt-c) to activation of cytosolic caspases. Cyt-c-mediated activation of caspases in cytosolic extracts and in cells is enhanced by overexpressing NAC and inhibited by reducing NAC using antisense/DNAzymes. Furthermore, association of NAC with Apaf1 is cyt c-inducible, resulting in a mega-complex (>1 MDa) containing both NAC and Apaf1 and correlating with enhanced recruitment and proteolytic processing of pro-caspase-9. NAC also collaborates with Apaf1 in inducing caspase activation and apoptosis in intact cells, whereas fragments of NAC representing only the CARD or NB domain suppress Apaf1-dependent apoptosis induction. NAC expression in vivo is associated with terminal differentiation of short lived cells in epithelia and some other tissues. The ability of NAC to enhance Apaf1-apoptosome function reveals a novel paradigm for apoptosis regulation.

Search for a new description of protein topology and local structure.

A novel description of protein structure in terms of the generalized secondary structure elements (GSSE) is proposed. GSSE's are defined as fragments of the protein structure where the chain doesn't radically change its direction. In this new language, global protein topology becomes a particular arrangement of the relatively small number of large, rod like GSSE's. Protein topology can be described by an adjacency matrix giving information, which GSSE's are close in space to each other and defining a graph, where GSSE's are equivalent to vertices and interactions between them to edges. The information about the local structure is translated into the local density of pseudo-Calpha atoms along the chain and the curvature of the chain. This new description has a number of interesting and useful features. For instance, enumeration theorems of graph theory can be used to estimate a number of possible topologies for a protein built from a given number of elements. Different topologies, including novel ones, can be generated from the known by various permutations of elements. Many new regularities in protein structures become suddenly visible in a new description. A new local structure description is more amenable to predictions and easier to use in fold predictions.

Sensitive sequence comparison as protein function predictor.

Protein function assignments based on postulated homology as recognized by high sequence similarity are used routinely in genome analysis. Improvements in sensitivity of sequence comparison algorithms got to the point, that proteins with previously undetectable sequence similarity, such as for instance 10-15% of identical residues, sometimes can be classified as similar. What is the relation between such proteins? Is it possible that they are homologous? What is the practical significance of detecting such similarities? A simplified analysis of the relation between sequence similarity and function similarity is presented here for the well-characterized proteins from the E. coli genome. Using a simple measure of functional similarity based on E.C. classification of enzymes, it is shown that it correlates well with sequence similarity measured by statistical significance of the alignment score. Proteins, similar by this standard, even in cases of low sequence identity, have a much larger chance of having similar function than the randomly chosen protein pairs. Interesting exceptions to these rules are discussed.

Protein structure prediction in biology and medicine

Protein structure prediction from sequence has made rapid advances in the last few years, gaining increased attention after the success of the CASP meetings. Specifically, threading predictions were shown to work for many cases previously thought to be outside of the scope of prediction methods. The papers in this session address a wide scope of topics, ranging from techniques for validation of prediction methods and further improvements of threading algorithms, to specific applications of protein structure predictions in biology.

AhpF can be dissected into two functional units: tandem repeats of two thioredoxin-like folds in the N-terminus mediate electron transfer from the thioredoxin reductase-like C-terminus to AhpC.

AhpF, the flavin-containing component of the Salmonella typhimurium alkyl hydroperoxide reductase system, catalyzes the NADH-dependent reduction of an active-site disulfide bond in the other component, AhpC, which in turn reduces hydroperoxide substrates. The amino acid sequence of the C-terminus of AhpF is 35% identical to that of thioredoxin reductase (TrR) from Escherichia coli. AhpF contains an additional 200-residue N-terminal domain possessing a second redox-active disulfide center also required for AhpC reduction. Our studies indicate that this N-terminus contains a tandem repeat of two thioredoxin (Tr)-like folds, the second of which contains the disulfide redox center. Structural and catalytic properties of independently expressed fragments of AhpF corresponding to the TrR-like C-terminus (F[208-521]) and the 2Tr-like N-terminal domain (F[1-202]) have been addressed. Enzymatic assays, reductive titrations, and circular dichroism studies of the fragments indicate that each folds properly and retains many functional properties. Electron transfer between F[208-521] and F[1-202] is, however, relatively slow (4 x 10(4) M(-)(1) s(-)(1) at 25 degrees C) and nonsaturable up to 100 microM F[1-202]. TrR is nearly as efficient at F[1-202] reduction as is F[208-521], although neither the latter fragment, nor intact AhpF, can reduce Tr. An engineered mutant AhpC substrate with a fluorophore attached via a disulfide bond has been used to demonstrate that only F[1-202], and not F[208-521], is capable of electron transfer to AhpC, thereby establishing the direct role this N-terminal domain plays in mediating electron transfer between the TrR-like part of AhpF and AhpC.

From fold to function predictions: an apoptosis regulator protein BID.

With the rapidly increasing pace of genome sequencing projects and the resulting flood of predicted amino acid sequences of uncharacterized proteins, protein sequence analysis, and in particular, protein structure prediction is quickly gaining in importance. Prediction algorithms can be used for preliminary annotation of newly sequenced proteins and, at least in some cases, provide insights into their function and specific mode of action. Such annotations for several microbial genomes were performed by several groups and placed in public domain for evaluation. An example presented in this work comes from a related project of structural and functional predictions for proteins involved in the process of controlled cell death (apoptosis). The BID protein belongs to an important class of regulators of apoptosis identified by short sequence motifs. Here, several fold prediction methods are used to build a series of three-dimensional models. Structure analysis of the models with reference to the biological data available allows selection of the most appropriate model. It is found that the most likely structural model of BID is built on the structure of Bcl-X(L). The model is discussed in terms of experimental data on specific proteolytic cleavage of BID and its effect on BID interactions with other proteins and membranes.