Superfamily and family analyses provide an effective tool for the functional classification of proteins, but must be automated for use on large datasets. We describe a 'gold standard' set of enzyme superfamilies, clustered according to specific sequence, structure, and functional criteria, for use in the validation of family and superfamily clustering methods. The gold standard set represents four fold classes and differing clustering difficulties, and includes five superfamilies, 91 families, 4,887 sequences and 282 structures.

The radical SAM superfamily contains over 100,000 homologous enzymes that catalyze a remarkably broad range of reactions required for life, including metabolism, nucleic acid modification, and biogenesis of cofactors. While the highly conserved SAM-binding motif responsible for formation of the key 5'-deoxyadenosyl radical intermediate is a key structural feature that simplifies identification of superfamily members, our understanding of their structure-function relationships is complicated by the modular nature of their structures, which exhibit varied and complex domain architectures. To gain new insight about these relationships, we classified the entire set of sequences into similarity-based subgroups that could be visualized using sequence similarity networks. This superfamily-wide analysis reveals important features that had not previously been appreciated from studies focused on one or a few members. Functional information mapped to the networks indicates which members have been experimentally or structurally characterized, their known reaction types, and their phylogenetic distribution. Despite the biological importance of radical SAM chemistry, the vast majority of superfamily members have never been experimentally characterized in any way, suggesting that many new reactions remain to be discovered. In addition to 20 subgroups with at least one known function, we identified additional subgroups made up entirely of sequences of unknown function. Importantly, our results indicate that even general reaction types fail to track well with our sequence similarity-based subgroupings, raising major challenges for function prediction for currently identified and new members that continue to be discovered. Interactive similarity networks and other data from this analysis are available from the Structure-Function Linkage Database.

Protein function identification remains a significant problem. Solving this problem at the molecular functional level would allow mechanistic determinant identification-amino acids that distinguish details between functional families within a superfamily. Active site profiling was developed to identify mechanistic determinants. DASP and DASP2 were developed as tools to search sequence databases using active site profiling. Here, TuLIP (Two-Level Iterative clustering Process) is introduced as an iterative, divisive clustering process that utilizes active site profiling to separate structurally characterized superfamily members into functionally relevant clusters. Underlying TuLIP is the observation that functionally relevant families (curated by Structure-Function Linkage Database, SFLD) self-identify in DASP2 searches; clusters containing multiple functional families do not. Each TuLIP iteration produces candidate clusters, each evaluated to determine if it self-identifies using DASP2. If so, it is deemed a functionally relevant group. Divisive clustering continues until each structure is either a functionally relevant group member or a singlet. TuLIP is validated on enolase and glutathione transferase structures, superfamilies well-curated by SFLD. Correlation is strong; small numbers of structures prevent statistically significant analysis. TuLIP-identified enolase clusters are used in DASP2 GenBank searches to identify sequences sharing functional site features. Analysis shows a true positive rate of 96%, false negative rate of 4%, and maximum false positive rate of 4%. F-measure and performance analysis on the enolase search results and comparison to GEMMA and SCI-PHY demonstrate that TuLIP avoids the over-division problem of these methods. Mechanistic determinants for enolase families are evaluated and shown to correlate well with literature results.

The funding section was mistakenly not included in this paper. The article has now been updated to include this section. The publisher apologises for this error.

With ever-increasing amounts of sequence data available in both the primary literature and sequence repositories, there is a bottleneck in annotating molecular function to a sequence. This article describes the biocuration process and methods used in the structure-function linkage database (SFLD) to help address some of the challenges. We discuss how the hierarchy within the SFLD allows us to infer detailed functional properties for functionally diverse enzyme superfamilies in which all members are homologous, conserve an aspect of their chemical function and have associated conserved structural features that enable the chemistry. Also presented is the Enzyme Structure-Function Ontology (ESFO), which has been designed to capture the relationships between enzyme sequence, structure and function that underlie the SFLD and is used to guide the biocuration processes within the SFLD. http://sfld.rbvi.ucsf.edu/.

To study the substrate specificity of enzymes, we use the amidohydrolase and enolase superfamilies as model systems; members of these superfamilies share a common TIM barrel fold and catalyze a wide range of chemical reactions. Here, we describe a collaboration between the Enzyme Specificity Consortium (ENSPEC) and the New York SGX Research Center for Structural Genomics (NYSGXRC) that aims to maximize the structural coverage of the amidohydrolase and enolase superfamilies. Using sequence- and structure-based protein comparisons, we first selected 535 target proteins from a variety of genomes for high-throughput structure determination by X-ray crystallography; 63 of these targets were not previously annotated as superfamily members. To date, 20 unique amidohydrolase and 41 unique enolase structures have been determined, increasing the fraction of sequences in the two superfamilies that can be modeled based on at least 30% sequence identity from 45% to 73%. We present case studies of proteins related to uronate isomerase (an amidohydrolase superfamily member) and mandelate racemase (an enolase superfamily member), to illustrate how this structure-focused approach can be used to generate hypotheses about sequence–structure–function relationships.