Predicting ligand-binding function in families of bacterial receptors

Abstract

The three-dimensional fold of a new protein sequence can often be inferred directly from sequence homology to a protein of known structure. The function of a new protein sequence is more difficult to predict, however, since homologues can have different molecular and cellular functions. To develop and automate computational methods for determining molecular function, we have analyzed ligand-binding specificity in two related families of binding proteins. One of these families includes Escherichia coli lactose repressor and ribose-binding protein, and the other includes E. coli sulfate-and phosphate-binding proteins. These proteins have similar folds but varying specificity, binding many different small molecules, including mono-and disaccharides, purines, oxyanions, ferric iron, and polyamines. Starting from template structural alignments, alignments of over 90 sequences per family were generated by iterative database searches with hidden Markov models. Phylogenetic trees were made of full-length sequences and of subsets of residues lining the binding cleft, to determine whether subbranches of the trees correlate with ligand-binding preference. Automated analyses of residues in the binding pocket were also used to predict ligand-binding function for many uncharacterized database sequences and to identify specific side chain-ligand contacts in proteins without solved structures. Our results demonstrate the utility of anchoring functional annotation within a protein family context. T he need for computational assignment of gene function is becoming more pressing as the number of new sequences far surpasses our ability to perform experiments to determine their functions. Initiatives for targeted three-dimensional structure determination for proteins of unknown fold (1, 2), coupled with advances in threading and other fold prediction methods, will soon lead to correct automated fold assignment for most new protein sequences. These advances will not as readily lead to correct functional annotation, however, since many different molecular and cellular functions may be associated with the same protein fold.The functional annotation of a new protein is often transferred from the homologous protein with the highest BLAST score. Because of the modularity of protein domains and divergence of function in paralogous proteins, this practice has led to many incorrect annotations and the propagation of annotation errors in sequence databases (3-5). Annotation accuracy may be improved by considering the evolutionary history of each protein family (6) or by requiring a match to particular functional residues in a sequence motif, in addition to overall homology, in order for functional annotation to be transferred. These methods generally require careful analysis of a particular family of sequences and have found the most success in recognizing conserved catalytic residues in the active sites of enzymes (7-10).Many proteins, however, have molecular functions that are defined by noncatalytic interactions with ligands or other prote...

Sequence, Structure and Energetic Determinants of Phosphopeptide Selectivity of SH2 Domains

“…The basic strategy is to map the physical chemical properties of the binding site of these complexes onto their sequence and in this way to obtain a binding-site signature that can be used to assign binding function to family members for which no structures are available. Our approach is related to that reported in the study by Johnson & Church,5 who analyzed ligand-binding specificity in bacterial periplasmic binding proteins on the basis of multiple sequence alignments and the analysis of residues in the binding pocket in proteins of known structure. The relationship of our method to that of Johnson & Church will be considered below.…”

Comparison of protein active site structures for functional annotation of proteins and drug design

“…This is typically accomplished by developing structural descriptors of active sites for defi ned protein functional classes and then fi tting these structural templates to novel folds to identify putative active sites and annotate the hypothetical proteins. A variety of approaches are being applied that include aligning structures to match a few consensus or enzymatic catalytic residues, [14][15][16][17][18][19][20][21][22][23] identifi cation of cavities consistent with shapes of known ligands, 24 a sequence independent force fi eld to extract common active site features, 25 theoretical prediction of titration curves, 26 using chemical properties and electrostatic potentials of amino acid residues consistent with active site characteristics, 27,28 neural network analysis of spatial clustering of residues, 29 and conserved residues from multiple sequence alignments (phylogenetic motifs). 20,30 Nevertheless, direct experimental observation of protein-ligand interactions are a more reliable mechanism for the proper and accurate identifi cation of protein active sites.…”

SDPpred: a tool for prediction of amino acid residues that determine differences in functional specificity of homologous proteins

“…Several approaches exploit information about protein structure or functional sites (1,2). Some methods use only protein sequences (3)(4)(5)(6)(7)(8).…”