Leveraging enzyme structure-function relationships for functional inference and experimental design: the structure-function linkage database.

Abstract

The study of mechanistically diverse enzyme superfamilies-collections of enzymes that perform different overall reactions but share both a common fold and a distinct mechanistic step performed by key conserved residues-helps elucidate the structure-function relationships of enzymes. We have developed a resource, the structure-function linkage database (SFLD), to analyze these structure-function relationships. Unique to the SFLD is its hierarchical classification scheme based on linking the specific partial reactions (or other chemical capabilities) that are conserved at the superfamily, subgroup, and family levels with the conserved structural elements that mediate them. We present the results of analyses using the SFLD in correcting misannotations, guiding protein engineering experiments, and elucidating the function of recently solved enzyme structures from the structural genomics initiative. The SFLD is freely accessible at http://sfld.rbvi.ucsf.edu.

Full text

Page 1

Leveraging Enzyme Structure-Function Relationships for Functional Inference and
Experimental Design: The Structure-Function Linkage Database†
Scott C.-H. Pegg,‡Shoshana D. Brown,‡Sunil Ojha,‡Jennifer Seffernick,|Elaine C. Meng,§John H. Morris,§
Patricia J. Chang,‡Conrad C. Huang,§Thomas E. Ferrin,‡,§and Patricia C. Babbitt*,‡,§
Departments of Biopharmaceutical Sciences and Pharmaceutical Chemistry, UniVersity of California, San Francisco,
1700 Fourth Street, San Francisco, California 94143-2250, and Department of Biochemistry, Molecular Biology, and
Biophysics, Biological Process Technology Institute, and Center for Microbial and Plant Genomics, UniVersity of Minnesota,
St. Paul, Minnesota 55108
ReceiVed October 14, 2005; ReVised Manuscript ReceiVed December 8, 2005
ABSTRACT: The study of mechanistically diverse enzyme superfamiliesscollections of enzymes that perform
different overall reactions but share both a common fold and a distinct mechanistic step performed by
key conserved residuesshelps elucidate the structure-function relationships of enzymes. We have
developed a resource, the structure-function linkage database (SFLD), to analyze these structure-function
relationships. Unique to the SFLD is its hierarchical classification scheme based on linking the specific
partial reactions (or other chemical capabilities) that are conserved at the superfamily, subgroup, and
family levels with the conserved structural elements that mediate them. We present the results of analyses
using the SFLD in correcting misannotations, guiding protein engineering experiments, and elucidating
the function of recently solved enzyme structures from the structural genomics initiative. The SFLD is
freely accessible at http://sfld.rbvi.ucsf.edu.
An important goal of the study of enzymes is a knowledge
of their sequence, structure, and function relationships that
is deductive, allowing us to rapidly determine without
physical experimentation, the substrate, chemical mechanism,
and product of a given sequence or structure (1). Such
knowledge should also be predictive, enabling us to design
enzymes that will perform desired reactions on substrates
of our choosing. For genome projects, our current abilities
to deduce function rely primarily on how similar a new
protein sequence is to that of a characterized enzyme. These
methods provide what is essentially a binary classification.
(The protein either performs the same reaction as the
characterized enzyme or it does not.) Depending upon the
similarity thresholds used, this approach can produce many
erroneous annotations. The annotation of enzymes by break-
ing down functional representations to correspond to struc-
tural similarities at different levels (e.g., superfamily, sub-
group, and family levels) enhances the precision with which
we can predict functional characteristics and aids in func-
tional inference. In this article we present and demonstrate
the application of a resource for performing such multilevel
annotations, the structure-function linkage database (SFLD)1.
Our strategy begins with the study of how enzymes have
evolved in nature, seeking an understanding of how proteins
of identical fold are used in many contexts to deliver a wide
variety of functions. Horowitz (2, 3) proposed that as an
enzyme evolved, it maintained the ability to bind a particular
substrate although the structural regions of the protein
involved in delivering chemistry changed. Although there
are some instances that appear to fit this early model (4),
there are now many more cases observed in which it appears
that in the evolution of new functions, the conserved aspect
of the enzyme structure-function relationship is not the
ability to bind a specific substrate but rather the ability to
perform a key mechanistic step in the chemical reaction (5,
6). Recent surveys of enzyme structure and function (7-9)
as well as related theoretical analyses of pathway evolution
(10-12) support this paradigm of chemistry-constrained
enzyme evolution. The result of millions of years of this type
of evolution has produced mechanistically diverse enzyme
superfamilies (7), each consisting of homologous enzymes
that perform a wide variety of overall chemical reactions on
an equally wide variety of substrates but maintain a key
mechanistic step mediated by conserved active-site features.
The enolase superfamily (13, 14) provides a good example
of how the analysis of structure-function relationships in
the context of mechanistically diverse superfamilies can lead
to valuable biological insights. The 1000+ proteins of this
superfamily perform many distinct and, often, quite different
†This work was performed with support from NIH grant GM60595
(to PCB), NSF grant 0234768 (to PCB), and NIH resource grant P41-
RR01081 (to TEF).
* To whom correspondence should be addressed. Tel: (415) 476-
3784. Fax: (415) 514-4260. E-mail: babbitt@cgl.ucsf.edu.
‡Department of Biopharmaceutical Sciences, University of Califor-
nia, San Francisco.
§Department of Pharmaceutical Chemistry, University of California,
San Francisco.
|University of Minnesota.
1Abbreviations: SFLD, Structure-Function Linkage Database;
HMM, hidden Markov model; CSA, Catalytic Site Atlas; AEE, L-Ala-
D/L-Glu epimerase; MLEII, chloromuconate lactonizing enzyme, also
called muconate lactonizing enzyme II; OSBS, o-succinylbenzoate
synthase; E. C., Enzyme Commission.
2545
Biochemistry 2006, 45, 2545-2555
10.1021/bi052101l CCC: $33.50© 2006 American Chemical Society
Published on Web 02/04/2006

Page 2

overall chemical reactions. Thirteen of these distinct catalytic
reactions have been experimentally characterized, and ad-
ditional reactions can be predicted from phylogenetic and
other bioinformatic and experimental analyses. Each of these
reactions includes a common mechanistic step, the abstraction
of a proton alpha to a carboxylate, leading to stabilization
of a similar type of enolate anion intermediate. This reaction
is shown in Figure 1. In all of the different superfamily
members, this partial reaction is mediated by active-site
residues conserved throughout the superfamily. Knowledge
of how these enzymes deliver function, that is, how specific
aspects of their chemical mechanisms are delivered via
conserved residues, has allowed us and our collaborators to
correct mistakes in the annotation of members of the
superfamily, to predict the functions of newly sequenced
members, and to design protein engineering experiments in
which two members of the superfamily were modified to
perform the very different overall reaction of a third.
As the rate of information generation from genomic
sequencing and high-throughput structure determination
projects increases, access to such information in the super-
family context becomes a pressing issue (15). What is
required is a computational framework that provides more
than information storage and the tools to access it. That is,
we require a level of organization and curation that allows
users to leverage the data for functional inference and
biological insight. Several well-known databases address
different aspects of this need, storing the raw sequence,
structure, and function data in a manner that allows users to
put it into a useful biological context. For example, the
Superfamily (16) and PFAM (17) databases group related
sequences together and provide hidden Markov models
(HMMs) (18) that can be used in the assignment of homology
to new protein sequences. Databases such as SCOP (19, 20)
and CATH (21) sort proteins into a structural hierarchy,
enabling users to examine distant evolutionary relationships
that are recognizable only from similarities in their 3D
structures. The KEGG (22) and MetaCyc (23, 24) databases
provide a catalog of metabolic pathways and reactions,
including many types of information about their component
enzymes and their small molecule substrates and products.
The more comprehensive database BRENDA (25) provides
both structural and functional information, including the
overall reactions of enzymes as well as their pathways and
structures, which is organized using the overall reaction-
based Enzyme Commission (E. C.) (26) numbering hierar-
chy. Yet none of these, including BRENDA, provides
information about how these properties are linked, particu-
larly, how a given protein sequence or structure performs
its molecular function. For example, using the KEGG or
BRENDA databases, one can easily find nucleotide and
amino acid sequences, crystal structures, and representations
of the overall reaction for the enzyme subtilisin, but no
information is provided regarding the mechanism by which
subtilisin performs proteolytic cleavage using the catalytic
triad Ser-His-Asp (27). This is not surprising, however, as
these databases were not specifically designed to contain
information about how proteins deliver function at the
molecular level but rather as tools that provide biological
insights at higher levels of analysis.
A newer resource, the Catalytic Site Atlas (CSA) (28),
provides a useful first step toward exploring enzyme
sequence-structure-function relationships by storing the
identities of catalytic residues for enzymes with solved crystal
structures. One of the strengths of the CSA is the use of a
strict definition of the term catalytic, which includes only
those residues thought to be directly involved in the reaction
catalyzed by an enzyme. The CSA has recently been
demonstrated as a useful tool for enzyme function prediction
(29). However, the CSA does not yet provide information
regarding the reaction mechanism of enzymes. Helping to
fill this gap, the recently developed databases EzCatDB (30)
and MACiE (31) catalog a broad range of enzyme reactions
and include information regarding the specific amino acids
involved. The MaCiE database also describes each overall
reaction in terms of its constituent steps to aid in the
representation of enzyme mechanism. Still, none of these
databases provide an organizational framework that directly
represents the connections between enzymes with similar
sequence-structure-function relationships.
We describe here a database we have designed to provide
information and analysis of the links between structure and
function information of enzymes, the structure-function
linkage database (SFLD). Specifically, the SFLD captures
structure-function relationships in mechanistically diverse
enzyme superfamilies. It is unique in identifying not only
the overall reactions of specific enzymes but also the partial
reactions (or other chemical capabilities) that represent
conserved functional capabilities shared among highly diver-
gent members of a superfamily. These are the functional attri-
butes that can be specifically correlated with conserved struc-
tural elements (residues or networks of residues) in superfami-
lies that follow the chemistry-constrained model of enzyme
evolution. This database is meant to be a resource of informa-
tion that can be leveraged to aid researchers in the analysis
and engineering of protein function. In particular, correlation
between structure and function at the superfamily level can
be used for rules-based predictions of some functional
capabilities of any new sequence or structure identified as a
member of a superfamily in the database. Finer granularity
assignment of a specific overall reaction, including substrate
specificity, can also be achieved for sequences or structures
that can be confidently assigned to an individual family
within a superfamily. We demonstrate here the use of the
SFLD for correcting misannotations, guiding protein engi-
neering experiments, and elucidating the function of recently
solved enzyme structures. We also describe the organization,
architecture, and searching abilities of the SFLD.
FIGURE 1: Partial reaction conserved in all members of the enolase
superfamily, the enolization of a carboxylic acid via the abstraction
of a proton alpha to a carboxylate leading to metal-assisted
stabilization of an enolic intermediate. Following this initial
conserved step, reactions within the superfamily can go down a
variety of paths, requiring, in some cases, additional partial reactions
as needed to complete the reactions shown. The structure-function
relationships of those subsequent steps are distinguished in the
SFLD at the subgroup and family levels.
2546 Biochemistry, Vol. 45, No. 8, 2006
Pegg et al.

Page 3

MATERIALS AND METHODS
An initial report describing key aspects of the schema and
methods relevant to the issue of structure-function linkage
in mechanistically diverse enzyme superfamilies has been
published (32). For convenience, we provide a summary of
that information at the end of the Results section.
In the SFLD, three levels of functional granularity are
defined. A superfamily is defined as a set of evolutionarily
related proteins that perform in common a fundamental
partial reaction or share a mechanistic attribute using
conserved structural elements. The subgroup designation
distinguishes structural variations in how the aspects of
shared chemistry are delivered in subsets of the superfamily
proteins. A family is defined as a set of proteins that perform
the same overall reaction using an identical mechanism.
Thus, families represent orthologous proteins.
RESULTS
The SFLD not only contains specific structure-function
relationships for a set of homologous enzymes but also
organizes them in a context focused on the shared chemistry
performed by conserved active-site residues. This grouping
of enzymes into functional fold superfamilies provides a
more relevant theoretical representation of structure-function
relationships in mechanistically diverse enzyme superfamilies
than has previously been available. The resulting framework
provides different levels of granularity at which functions
can be assigned and thereby is less prone to the overpredic-
tion of functional characteristics than are representations
based only on overall reactions. Thus, if only the active-site
residues associated with a partial reaction are conserved in
a member of a superfamily, then that protein is annotated
only with the superfamily common partial reaction. Only if
other family-specific structural characteristics are additionally
conserved is a protein annotated with the family-specific
overall reaction. In the sections that follow, examples of how
the information and tools provided in the SFLD can be
leveraged for functional insight are illustrated. Following this,
we provide a short summary of the organization, contents,
and analysis tools in the SFLD.
Correcting Misannotations
The need for rapid, high-throughput annotation of whole
genomes has resulted in increasing reliance upon electronic
annotation schemes that assign functional similarity accord-
ing to sequence similarity. Thus, a new protein sequence is
assigned the function of the annotated protein(s) closest to
it in sequence (typically using a similarity threshold). For
enzymes belonging to a widely divergent superfamily in
which the members may perform many different reactions,
this approach is often problematic because the protein (or
protein family) showing the closest sequence similarity may
not perform the same overall reaction. Such situations may
be more common than initially expected, as recent experi-
ments have demonstrated that even single amino acid changes
in members of diverse enzyme superfamilies are enough to
alter substrate specificity, mechanism, or both (33). In these
cases, the specific correlations between conserved active-
site residues and the aspects of function they mediate can
sometimes be used to identify annotations that have been
incorrectly assigned by high-throughput methods.
For example, the sequence annotated in Genbank as
muconate cycloisomerase from the organism Oceanobacillus
iheyensis (gi 23100420) shows as its closest characterized
homologues the MLEI family of enzymes (members of the
enolase superfamily) in both BLAST and PFAM searches.
This sequence displays those active-site residues generally
conserved across the superfamily to perform the conserved
partial reaction of the abstraction of a proton alpha to a
carboxylate. Yet, it lacks a key glutamic acid residue found
in all experimentally characterized members and orthologs
that are thought to perform the specific overall MLE reaction
(34). Aligning the sequence to other families within the
superfamily, armed with the knowledge of which residues
are conserved at the superfamily level and which ones are
conserved only at the family level, we believe this sequence
is more likely to be a dipeptide epimerase.2
As sequencing projects expand to genomes that represent
more diverse biological niches, we are likely to see a greater
number of sequences that are incorrectly assigned the
function of a distantly related but characterized homologue
via the use of automated high-throughput annotation meth-
ods. Such errors and their propagation into secondary
databases pose a significant problem to biologists at many
levels, from the study of individual enzymes to the larger
analyses of systems biology. Some studies have estimated
the rates of misannotation in public databases to be as high
as 15-30% (35, 36). As shown in the example above,
knowledge of how families and superfamilies of related
enzymes carry out conserved mechanistic steps such as that
provided by the SFLD can be leveraged to provide methods
for correcting such misannotations. To gain a more accurate
estimate of levels of misannotation for mechanistically
diverse enzyme superfamilies in public databases, a system-
atic study of misannotation using the superfamilies in the
SFLD is currently underway in our laboratory.
Protein Engineering
The organization of enzymes into the functional fold
hierarchy used by the SFLD has helped guide protein
engineering experiments designed to investigate potential
pathways for functional evolution. By grouping evolution-
arily related but mechanistically diverse enzymes into
superfamilies according to their conserved mechanistic steps,
we can more easily recognize the structural determinants
associated with substrate specificity and distinguish them
from residues conserved across the entire superfamily. With
a knowledge of how the common mechanistic step is
delivered by the structure, we can also recognize specific
regions of the active site that may be altered to create the
additional functional steps required to perform a new
chemical reaction.
2We note that the closest BLAST hit reported using gi23100420 as
a query against the NCBI “nr” database is gi16078363. This sequence,
nominally annotated in the Genbank header as “similar to chloromu-
conate cycloisomerase,” is more likely to be a dipeptide epimerase. In
the alignment section of the BLAST output, other gi numbers grouped
by the “nr” database as identical to gi16078363 are listed, including
the solved structure of the experimentally characterized dipeptide
epimerase (gi18158850) as well as identical sequences annotated as
“chloromuconate cycloisomerase homologue ykfB” (gi7428454) and
“YkfB” (gi2632019 and gi2633652). Automated annotation methods
typically do not parse the alignment section of the BLAST output, and
are unlikely to be able to resolve the conflicting annotations listed.
Leveraging Enzyme Structure-Function Relationships
Biochemistry, Vol. 45, No. 8, 2006 2547

Page 4

Proof of concept for this idea was recently obtained from
a protein engineering experiment in which two members of
the enolase superfamily were re-engineered to perform a
quite different overall reaction of a third member (33). Both
of the reactions performed by the template enzymes, L-Ala-
D/L-Glu epimerase (AEE) and chloromuconate lactonizing
enzyme (MLEII), and the target reaction, o-succinylbenzoate
synthesis, represent different overall reactions within the
enolase superfamily but share the common mechanistic step
of the abstraction of a proton alpha to a carboxylate. Figure
2A shows the structural conservation of the residues in the
active sites of these enzymes that perform this step. Aside
from this common ability, however, all three of these
enzymes perform quite different overall reactions (Figure
2B). By comparing the structures of the AEE and OSBS
enzymes and considering the known structure-function
relationships both have in common, that is, shared by the
superfamily, contrasted with those unique to each enzyme,
a single active-site mutant of AEE (D297G) was designed
and found to perform the reaction catalyzed by OSBS. This
choice was guided by the knowledge that substrates in the
enolase superfamily must bind in an orientation conducive
to the stabilization of the enolate anion intermediate by the
conserved constellation of residues shown in Figure 2A. A
similar result was obtained via directed-evolution experi-
ments in which a single active-site mutation in MLEII was
found to confer the OSBS function. Although neither the
wild-type AEE nor MLEII performs the OSBS reaction,
enzyme efficiencies for the OSBS reaction of the single-site
mutants are sufficient to complement an OSBS knockout
strain and show measurable levels of OSBS activity (Kcat/
Km(M-1s-1) of 12.5 and 2 × 103for the AEE and MLEII
mutants, respectively). We suggest here that choosing a
starting template for protein engineering that already “knows”
how to perform a fundamental aspect of a new chemical
reaction of interest may increase the success of protein design
strategies in general. Analysis of the structure-function
relationships among members of the enolase and other
mechanistically diverse superfamilies thus can provide guid-
ance for such protein engineering experiments and aid in
our efforts to gain a broader understanding of the structural
determinants of enzyme specificity. The SFLD stores these
relationships in a functional fold superfamily hierarchy that
provides users with the tools to leverage this information
for their own protein engineering work.
Elucidation of Function: Structural Genomics Targets
The notion that the elucidation of a protein’s 3D structure
can aid in the determination of the protein’s function is one
of the principal ideas underlying the vision for the structural
genomics initiative (37) projects. For many proteins, how-
ever, the determination of the structure does not immediately
lead to inference of function. Of the 1605 publicly available
protein structures determined from the structural genomics
initiative (as of June 21, 2005), 590 are annotated as having
unknown functions. Using the HMMs and curated alignments
of the SFLD, we examined all 1605 structures and compared
our functional predictions for these proteins to their annota-
tions in the Protein Data Bank (pdb) (38). We were also
able to predict some functional properties for the structures
whose functions are annotated as unknown. Table 1 shows
all of the structural genomics initiative targets that matched
at least one hidden Markov model in the SFLD and the level-
(s) of granularity (superfamily, subgroup, or family) at which
the target’s function can be described. For each target that
matched a family HMM (families in the SFLD represent
isofunctional groups of enzymes), Table 1 also gives the
fraction of conserved functional residues that align between
the target and the curated family alignment in the SFLD.
Target 1HZD. The functional predictions made using the
HMMs in the SFLD agree with a majority of the annotations
provided by the pdb, with some notable differences. Target
1HZD, although annotated as an RNA-binding homologue
of enoyl-CoA hydratase by the pdb, matches the methyl-
glutaconyl-CoA hydratase family in the crotonase (also called
the enoyl-CoA hydratase) superfamily of the SFLD. Al-
though the literature suggests that the methylglutaconyl-CoA
hydratase function is the most biologically relevant (loss of
this function via mutation in the human gene causes the
disorder 3-methylglutaconic aciduria type I (39)), this
particular protein has been experimentally determined to
perform at least three functions: RNA binding, hydration
of enoyl-CoA, and hydration of methylglutaconyl-CoA (40,
FIGURE 2: (A) Superimposed active sites of crystal structures of
1MUC (muconate lactonizing enzyme), 1JPM (L-Ala-D/L-Glu
epimerase), and 1FHU (o-succinylbenzoate synthase). The highly
conserved active-site residues involved in catalysis of the proton
abstraction step common to all three enzymes are represented in
color (1MUC in red, 1JPM in blue, and 1FHU in purple). (B)
Different overall reactions catalyzed by these enzymes.
2548 Biochemistry, Vol. 45, No. 8, 2006
Pegg et al.

Page 5

41). This target represents a good example of the difficulties
in assigning a function to multifunctional enzymes. The
SFLD adds additional knowledge to the annotation of 1HZD
in the pdb, that is, the assignment of an additional specific
enzymatic function.
Targets 1RVK and 1TZZ. Targets 1RVK and 1TZZ are
annotated in the pdb as having unknown function, but they
both match very well the mandelate racemase (MR) subgroup
of the enolase superfamily in the SFLD. From the coarsest
granularity prediction at the superfamily level, we know that
both proteins perform proton abstraction on a carbon R to a
carboxylate in the unknown substrate. Further, assignment
to the MR subgroup specifies proton abstraction machinery
using a His-Asp dyad for a substrate of R stereochemistry
and proton abstraction machinery usually using a Lys-X-X
(usually Lys) for a substrate of S stereochemistry. Thus, even
though we cannot assign the specific substrates and overall
reactions for either 1RVK or 1TZZ, the SFLD provides
additional information not found in the pdb, including a
context in which to interpret the existence of conserved
active-site residues. Using the conserved structure-function
relationships we associate with the MR subgroup, we have,
along with our collaborators, made further predictions of
overall function that are currently being tested experimentally
by our collaborators.
Targets 1WUE and 1WUF. Targets 1WUE and 1WUF,
both annotated in the pdb as having unknown function, match
the o-succinylbenzoate synthase (OSBS) family of the SFLD.
When aligned with the curated family alignment in the
SFLD, which indicates the location, type, and function of
the conserved residues required to perform the OSBS
reaction, all five (out of five) residues are conserved in both
targets. An examination of the genomic context of these
targets adds confidence to these functional predictions. The
gene encoding 1WUE (in Enterococcus faecalis) lies within
an operon encoding five other genes of the menaquinone
synthesis pathway in which the OSBS reaction is the third
(out of seven) step (42). Whereas the gene encoding 1WUF
(Listeria innocua) does not appear to be localized within an
operon, the other six genes of the menaquinone synthesis
pathway are present in the organism. We note that a BLASTP
(43) search using each of these sequences shows that the
corresponding gi numbers at the NCBI GenBank database
(44) for each of these structures annotates these sequences
as similar to an OSBS (gi 16804558, 1WUE) or as an OSBS
homologue (gi 25514206, 1WUF). The predictions that both
of these enzymes perform the OSBS reaction has recently
been confirmed experimentally (Yew, W. S., and Gerlt, J.
A., personal communication).
Table 1: Structures Solved by the Structural Genomics Initiative that Match Hidden Markov Models of the SFLD
pdb
1j6o
1j6p
pdb annotationsuperfamily
amidohydrolase
amidohydrolase
subgroup family CFRa
Tatd-related deoxyribonuclease
metal-dependent hydrolase of
cytosinedemaniase chlorohydrolase family
collapsin response mediator protein 1
N-acetylglucosamine-6-phosphate
deacetylase
Tatd deoxyribonuclease
Tatd homolog, hydrolase
N-acetylglucosamine-6-phosphate
deacetylase
RNA-binding homologue of
enoyl-CoA hydratase
MenB-naphthoate synthase
uncharacterized-147
uncharacterized-66
1kcx
1o12
amidohydrolase
amidohydrolase
collapsin response mediator
N-acetylglucosamine-6-
phosphate
TatD_MttC
uncharacterized-147
N-acetylglucosamine-6-
phosphate
D-hydantoinaseb
N-acetylglucosamine-6-
phosphate deacetylase
1/6
5/5
1xwy
1yix
1ymy
amidohydrolase
amidohydrolase
amidohydrolase
N-acetylglucosamine-6-
phosphate deacetylase
methylglutaconyl-CoA
hydratase
1,4-dihydroxy-2-naphthoyl-CoA
synthase
enoyl-CoA hydrataseb
5/5
1hzd crotonase7/7
1rjncrotonase 4/4
1uiy
1rvk
1tzz
1wue
1wuf
enoyl-CoA hydratase
hypothetical protein, unknown function
unknown member of enolase superfamily
unknown member of enolase superfamily
member of enolase superfamily,
unknown function
L-fuconate dehydratase
deoxy-D-mannose
-octulosonate 8-phosphate phosphatase
phosphoserine phosphatase
putative Nagd protein
putative phosphatase
4-nitrophenylphosphatase
putative phosphatase
putative mannosyl-3-phosphoglycerate
phosphatase
hydrolase, haloacid
dehalogenase-like family
hypothetical protein,
unknown function
unknown function
metallo protein from glyoxalase family,
unknown function
crotonase
enolase
enolase
enolase
enolase
3/4
mandelate racemase
mandelate racemase
muconate cycloisomerase
muconate cycloisomerase
o-succinylbenzoate synthase
o-succinylbenzoate synthase
5/5
5/5
1yey
1k1e
enolase
HAD
mandelate racemase
phosphatase-like2
L-fuconate dehydratase
deoxy-D-mannose-octulosonate
8-phosphate phosphatase
phosphoserine phosphatase
6/6
6/6
1l7p
1pw5
1te2
1vjr
1wvi
1xvi
HAD
HAD
HAD
HAD
HAD
HAD
phosphatase-like2
phosphatase-like4
phosphatase-like1
phosphatase-like4
phosphatase-like4
phosphatase-like3
6/6
mannosyl-3-phosphoglycerate
phosphatase
4/4
1ydf HADphosphatase-like4
1ys9HADphosphatase-like4
1k4n
1zsw
VOC
VOC
YecM-like
2,6-dichlorohydroquinone
dioxygenase
aCFR: the fraction of conserved active-site residues present when aligned to curated family alignments of the SFLD.bAlthough these protein
sequences match a family HMM in the SFLD, the fact that they are missing at least one functionally important residue suggests that they do not
perform the designated family reaction.
Leveraging Enzyme Structure-Function Relationships
Biochemistry, Vol. 45, No. 8, 2006 2549