CATH: an expanded resource to predict protein function through structure and sequence.
Dawson NL, Lewis TE, Das S, Lees JG, Lee D, Ashford P, Orengo CA, Sillitoe I. · 2017
Read CATH: an expanded resource to predict protein function through structure and sequence. by Dawson NL, Lewis TE, Das S, Lees JG, Lee D, Ashford P, Orengo CA, Sillitoe I. (2017) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5210570.txt.
Introduction
Natalie L. Dawson, Tony E. Lewis, Sayoni Das, Jonathan G. Lees, David Lee, Paul Ashford, Christine A. Orengo, and Ian Sillitoe reported the CATH-Gene3D update in Nucleic Acids Research in 2017. CATH sorts protein domains, the compact folding units of proteins, by shape and family ties, so a new sequence can be matched to relatives and its likely job predicted.
Version 4.1 holds over 300,000 domain structures and more than 53 million protein domains grouped into 2,737 superfamilies, twice the predicted domains of the prior release. A daily-updated companion adds early calls for over 100,000 more domains. New work includes wider structure and sequence coverage, larger functional families, a machine-learning tool that assigns domains to superfamilies on its own, better search with alignments, and redesigned web pages and downloads.
Catalog record
| CREATOR | Dawson NL, Lewis TE, Das S, Lees JG, Lee D, Ashford P, Orengo CA, Sillitoe I. |
|---|---|
| YEAR | 2017 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
CATH: an expanded resource to predict protein function through structure and sequence By Natalie L. Dawson, Tony E. Lewis, Sayoni Das, Jonathan G. Lees, David Lee, Paul Ashford, Christine A. Orengo, Ian Sillitoe Published in: Nucleic Acids Research DOI: 10.1093/nar/gkw1098 ABSTRACT The latest version of the CATH-Gene3D protein structure classification database has recently been released (version 4.1, http://www.cathdb.info). The resource comprises over 300 000 domain structures and over 53 million protein domains classified into 2737 homologous superfamilies, doubling the number of predicted protein domains in the previous version. The daily-updated CATH-B, which contains our very latest domain assignment data, provides putative classifications for over 100 000 additional protein domains. This article describes developments to the CATH-Gene3D resource over the last two years since the publication in 2015, including: significant increases to our structural and sequence coverage; expansion of the functional families in CATH; building a support vector machine (SVM) to automatically assign domains to superfamilies; improved search facilities to return alignments of query sequences against multiple sequence alignments; the redesign of the web pages and download site. INTRODUCTION CATH-Gene3D, established in the mid 1990s, is a publicly-accessible, online resource providing a protein domain structure classification (1) (http://www.cathdb.info). CATH identifies protein domain structures within protein three-dimensional (3D) structures from the Protein Data Bank (PDB) (2) and assigns domains sharing evolutionary similarities to the same superfamily within the CATH hierarchical structure classification. The structural hierarchy allows these superfamilies to be organised within structural space, the letters CATH represent the levels in that hierarchy: Class, Architecture, Topology and Homologous superfamily. Additional protein domain sequences with no known structure are identified from UniProtKB (3) and Ensembl (4) protein sequences and classified within our sister resource Gene3D (5). Hidden Markov model (HMM) technology (6) is then employed to create profiles, or evolutionary fingerprints, seeded from these CATH structural domains. All known protein sequences are then scanned against these HMMs and where significant matches are found, domain boundaries and evolutionary relationships are inferred onto the query protein sequence. CATH-Gene3D has recently become an ELIXIR-endorsed resource. The ELIXIR programme has a number of objectives in developing bioinformatics resources, including: training, developing data standards, collaborations, developing tools and supporting innovation (https://www.elixir-europe.org/). The CATH-Gene3D resource is contributing to a current ELIXIR research programme, EXCELERATE, in which the CATH-Gene3D HMMs are being used to assign domain structure and function annotations for metagenome sequences in the marine metagenome data use case. CATH V4.1 RELEASE HIGHLIGHTS Since the last release of CATH (v4.0), an additional ∼73 000 domains of known structure and over 4000 sequence families (i.e. families of homologous domains with at least 35% sequence identity) have been incorporated. CATH-B Creating official releases of CATH (e.g. v4.1) is an important part of our resource pipeline: it provides a static target to which other resources (both internal and external) can consistently reference their own data. Once the domain definitions and classifications for a particular CATH release have been ‘frozen’, the release protocol goes on to generate additional calculations, annotations and analyses. The extra data provided from these steps are integral both for internal research projects in the future and the backwards compatibility required by the external scientific community. However, this release process also increases the time between the very latest structures being released in the PDB and the time annotations are made available in the latest release of CATH-Gene3D. To address this, we also provide access to the very latest putative CATH annotations in CATH-B. This list of domain definitions and superfamily assignments is updated daily (1) and is available to download in a number of different formats. Over 414 000 domains are present in CATH-B, adding an additional ∼100 000 domains to the v4.1 release of CATH-Gene3D. There has also been a significant increase in the number of homologous superfamilies putatively classified in CATH as a result of the new support vector machine (SVM) method (see ‘New Data in CATH’ Section). Since a number of additional checks and analyses are performed when a release is frozen, it is possible that some assignments may change from CATH-B (e.g. additional evidence may allow two superfamilies to be merged). However, in practice, only 0.15% of all domain boundaries and 0.02% of all superfamily assignments were seen to change during the last release process (from CATH-B to v4.1). As such, these ‘putative’ assignments in CATH-B should be considered safe to use. CATH FunFams FunFams or Functional Families are sub-clusters of domains (structures and sequences) within homologous superfamilies that are likely to have highly similar structures and functions. FunFams are generated by agglomerative hierarchical clustering of domain relatives within a superfamily using the Genome Modelling and Model Annotation (GeMMA) algorithm (7) followed by optimal partitioning of the resulting superfamily clustering tree by the automated family classification method, FunFHMMer (8). FunFHMMer uses evolutionary patterns within cluster sequence alignments such as highly conserved positions and specificity-determining positions to distinguish between clusters that are likely to share the same function and vice-versa, thereby determining an optimal cut of the superfamily tree. Sub-classification of the CATH superfamilies into FunFams helps to capture the functional diversity within a superfamily and allows comparison of functional sites between relatives. This provides valuable insights into evolutionary mechanisms underlying changes in function (9,10). It also helps to provide function annotations of greater precision for more uncharacterised proteins than other domain-based resources (8). Functional annotations predicted using CATH FunFam assignments were ranked among the top five function prediction methods (out of 126 methods) in the Critical Assessment of Function Annotation experiment 2 (11) for both ‘molecular function’ and ‘biological process’ prediction. To update the FunFams, the CATH superfamilies were first populated with updated sequence domains from our sister resource Gene3D (v14). These domain sequences were obtained by scanning UniProtKB and Ensembl sequence data against the CATH v4.1 HMMs (constructed for each S35 cluster representative in a CATH superfamily using the HMMER3 software suite (6)). Domains assigned to a given CATH superfamily in v4.1 were mapped to equivalent domains assigned to FunFams in v4.0. Since it is possible that domain boundaries change between versions of CATH (due to updated software, updated HMMs, additional sequence data, etc.), any domain that was seen to change significantly (i.e. overlap < 70%) was removed from the existing FunFam. These mapping issues affected ∼17% of the Gene3D sequences assigned in v4.1. Many of the domain sequences that changed significantly came from singleton FunFams (i.e. FunFams with a single member) or FunFams with low populations, which resulted in removing ∼17 000 FunFams. As a result, CATH-Gene3D v4.1 now comprises 92 882 FunFams from 2737 superfamilies. Approximately 12% of the FunFams in v4.1 have increased in population by 10% or more since v4.0. After refreshing FunFams with the updated domain sequences, new FunFam HMMs were used to search against the pool of novel domain sequences. Inclusion thresholds were determined for each FunFam as described in (8) and new domains were added to existing FunFams if the match surpassed the inclusion threshold. CATH currently classifies over 21 million domain sequences into FunFams within superfamilies. Since functional terms have been integrated from Gene3D (i.e. Gene Ontology (GO) (12) and EC classification (13) terms), this constitutes 27 459 unique functional terms associated with FunFam sequences (4504 different