The Erbin PDZ domain binds with high affinity and specificity to the carboxyl termini of delta-catenin and ARVCF

Erbin is a recently described member of the LAP (leucine-rich repeat and PDZ domain) protein family. We used a C-terminally displayed phage peptide library to identify optimal ligands for the Erbin PDZ domain. Phage-selected peptides were type 1 PDZ ligands that bound with high affinity and specificity to the Erbin PDZ domain in vitro. These peptides most closely resembled the C-terminal PDZ domain-binding motifs of three p120-related catenins: delta-catenin, ARVCF, and p0071 (DSWV-COOH). Analysis of the interactions of the Erbin PDZ domain with synthetic peptides matching the C termini of ARVCF or delta-catenin also demonstrated specific high affinity binding. We characterized the interactions between the Erbin PDZ domain and both ARVCF and delta-catenin in vitro and in vivo. The Erbin PDZ domain co-localized and coprecipitated with ARVCF or delta-catenin complexed with beta-catenin and E/N-cadherin. Mutagenesis and peptide competition experiments showed that the association of Erbin with the cadherin-catenin complex was mediated by the interaction of its PDZ domain with the C-terminal PDZ domain-binding motifs (DSWV-COOH) of ARVCF and delta-catenin. Finally, we showed that endogenous delta-catenin and Erbin co-localized in and co-immunoprecipitated from neurons. These results suggest that delta-catenin and ARVCF may function to mediate the association of Erbin with the junctional cadherin-catenin complex. They also demonstrate that C-terminal phage-display technology can be used to predict physiologically relevant ligands for PDZ domains.     

Alteration of the C-Terminal Ligand Specificity of the Erbin PDZ Domain by Allosteric Mutational Effects

Modulation of protein binding specificity is important for basic biology and for applied science. Here we explore how binding specificity is conveyed in PDZ (postsynaptic density protein-95/discs large/zonula occludens-1) domains, small interaction modules that recognize various proteins by binding to an extended C terminus. Our goal was to engineer variants of the Erbin PDZ domain with altered specificity for the most C-terminal position (position 0) where a Val is strongly preferred by the wild-type domain. We constructed a library of PDZ domains by randomizing residues in direct contact with position 0 and in a loop that is close to but does not contact position 0. We used phage display to select for PDZ variants that bind to 19 peptide ligands differing only at position 0. To verify that each obtained PDZ domain exhibited the correct binding specificity, we selected peptide ligands for each domain. Despite intensive efforts, we were only able to evolve Erbin PDZ domain variants with selectivity for the aliphatic C-terminal side chains Val, Ile and Leu. Interestingly, many PDZ domains with these three distinct specificities contained identical amino acids at positions that directly contact position 0 but differed in the loop that does not contact position 0. Computational modeling of the selected PDZ domains shows how slight conformational changes in the loop region propagate to the binding site and result in different binding specificities. Our results demonstrate that second-sphere residues could be crucial in determining protein binding specificity.     

Novel mode of ligand recognition by the Erbin PDZ domain

Erbin contains a class I PDZ domain that binds to the C-terminal region of the receptor tyrosine kinase ErbB2, a class II ligand. The crystal structure of the human Erbin PDZ bound to the peptide EYLGLDVPV corresponding to the C-terminal residues 1247-1255 of human ErbB2 has been determined at 1.25-A resolution. The Erbin PDZ deviates from the canonical PDZ fold in that it contains a single alpha-helix. The isopropyl group of valine at position -2 of the ErbB2 peptide interacts with the Erbin Val(1351) and displaces the peptide backbone away from the alpha-helix, elucidating the molecular basis of class II ligand recognition by a class I PDZ domain. Strikingly, the phenolic ring of tyrosine -7 enters into a pocket formed by the extended beta 2-beta 3 loop of the Erbin PDZ. Phosphorylation of tyrosine -7 abolishes this interaction but does not affect the binding of the four C-terminal peptidic residues to PDZ, as revealed by the crystal structure of the Erbin PDZ complexed with a phosphotyrosine-containing ErbB2 peptide. Since phosphorylation of tyrosine -7 plays a critical role in ErbB2 function, the selective binding and sequestration of this residue in its unphosphorylated state by the Erbin PDZ provides a novel mechanism for regulation of the ErbB2-mediated signaling and oncogenicity.     

Convergent and divergent ligand specificity among PDZ domains of the LAP and zonula occludens (ZO) families

We present a detailed comparative analysis of the PDZ domains of the human LAP proteins Erbin, Densin-180, and Scribble and the MAGUK ZO-1. Phage-displayed peptide libraries and in vitro affinity assays were used to define ligand binding profiles for each domain. The analysis reveals the importance of interactions with all four C-terminal residues of the ligand, which constitute a core recognition motif, and also the role of interactions with more upstream ligand residues that support and modulate the core binding interaction. In particular, the results highlight the importance of site(-1), which interacts with the penultimate residue of ligand C termini. Site(-1) was found to be monospecific in the Erbin PDZ domain (accepts tryptophan only), bispecific in the first PDZ domain of ZO-1 (accepts tryptophan or tyrosine), and promiscuous in the Scribble PDZ domains. Furthermore, it appears that the level of promiscuity within site(-1) greatly influences the range of potential biological partners and functions that can be associated with each protein. These findings show that subtle changes in binding specificity can significantly alter the range of biological partners for PDZ domains, and the insights enhance our understanding of this diverse family of peptide-binding modules.     

Comprehensive analysis of all evolutionary paths between two divergent PDZ domain specificities

To understand the molecular evolution of functional diversity in protein families, we comprehensively investigated the consequences of all possible mutation combinations separating two peptide‐binding domains with highly divergent specificities. We analyzed the Erbin PDZ domain (Erbin‐PDZ), which exhibits canonical type I specificity, and a synthetic Erbin‐PDZ variant (E‐14) that differs at six positions and exhibits an atypical specificity that closely resembles that of the natural Pdlim4 PDZ domain (Pdlim4‐PDZ). We constructed a panel of 64 PDZ domains covering all possible transitions between Erbin‐PDZ and E‐14 (i.e., the panel contained variants with all possible combinations of either the Erbin‐PDZ or E‐14 sequence at the six differing positions). We assessed the specificity profiles of the 64 PDZ domains using a C‐terminal phage‐displayed peptide library containing all possible genetically encoded heptapeptides. The specificity profiles clustered into six distinct groups, showing that intermediate domains can be nodes for the evolution of divergent functions. Remarkably, three substitutions were sufficient to convert the specificity of Erbin‐PDZ to that of Pdlim4‐PDZ, whereas Pdlim4‐PDZ contains 71 differences relative to Erbin‐PDZ. X‐ray crystallography revealed the structural basis for specificity transition: a single substitution in the center of the binding site, supported by contributions from auxiliary substitutions, altered the main chain conformation of the peptide ligand to resemble that of ligands bound to Pdlim4‐PDZ. Our results show that a very small set of mutations can dramatically alter protein specificity, and these findings support the hypothesis whereby complex protein functions evolve by gene duplication followed by cumulative mutations.     

Comparative structural analysis of the Erbin PDZ domain and the first PDZ domain of ZO-1. Insights into determinants of PDZ domain specificity

We report a structural comparison of the first PDZ domain of ZO-1 (ZO1-PDZ1) and the PDZ domain of Erbin (Erbin-PDZ). Although the binding profile of Erbin-PDZ is extremely specific ([D/E][T/S]WV(COOH)), that of ZO1-PDZ1 is similar ([R/K/S/T][T/S][W/Y][V/I/L](COOH)) but broadened by increased promiscuity for three of the last four ligand residues. Consequently, the biological function of ZO-1 is also broadened, as it interacts with both tight and adherens junction proteins, whereas Erbin is restricted to adherens junctions. Structural analyses reveal that the differences in specificity can be accounted for by two key differences in primary sequence. A reduction in the size of the hydrophobic residue at the base of the site(0) pocket enables ZO1-PDZ1 to accommodate larger C-terminal residues. A single additional difference alters the specificity of both site(-1) and site(-3). In ZO1-PDZ1, an Asp residue makes favorable interactions with both Tyr(-1) and Lys/Arg(-3). In contrast, Erbin-PDZ contains an Arg at the equivalent position, and this side chain cannot accommodate either Tyr(-1) or Lys/Arg(-3) but, instead, interacts favorably with Glu/Asp(-3). We propose a model for ligand recognition that accounts for interactions extending across the entire binding site but that highlights several key specificity switches within the PDZ domain fold.     

Crystal structure of the second PDZ domain of SAP97 in complex with a GluR-A C-terminal peptide

Synaptic targeting of GluR-A subunit-containing glutamate receptors involves an interaction with synapse-associated protein 97 (SAP97). The C-terminus of GluR-A, which contains a class I PDZ ligand motif (-x-Ser/Thr-x-phi-COOH where phi is an aliphatic amino acid) associates preferentially with the second PDZ domain of SAP97 (SAP97(PDZ2)). To understand the structural basis of this interaction, we have determined the crystal structures of wild-type and a SAP97(PDZ2) variant in complex with an 18-mer C-terminal peptide (residues 890-907) of GluR-A and of two variant PDZ2 domains in unliganded state at 1.8-2.44 A resolutions. SAP97(PDZ2) folds to a compact globular domain comprising six beta-strands and two alpha-helices, a typical architecture for PDZ domains. In the structure of the peptide complex, only the last four C-terminal residues of the GluR-A are visible, and align as an antiparallel beta-strand in the binding groove of SAP97(PDZ2). The free carboxylate group and the aliphatic side chain of the C-terminal leucine (Leu907), and the hydroxyl group of Thr905 of the GluR-A peptide are engaged in essential class I PDZ interactions. Comparison between the free and complexed structures reveals conformational changes which take place upon peptide binding. The betaAlpha-betaBeta loop moves away from the C-terminal end of alphaB leading to a slight opening of the binding groove, which may better accommodate the peptide ligand. The two conformational states are stabilized by alternative hydrogen bond and coulombic interactions of Lys324 in betaAlpha-betaBeta loop with Asp396 or Thr394 in betaBeta. Results of in vitro binding and immunoprecipitation experiments using a PDZ motif-destroying L907A mutation as well as the insertion of an extra alanine residue between the C-terminal Leu907 and the stop codon are also consistent with a 'classical' type I PDZ interaction between SAP97 and GluR-A C-terminus.     

Structural Basis for Asymmetric Association of the βPIX Coiled Coil and Shank PDZ

βPIX (p21-activated kinase interacting exchange factor) and Shank/ProSAP protein form a complex acting as a protein scaffold that integrates signaling pathways and regulates postsynaptic structure. Complex formation is mediated by the C-terminal PDZ binding motif of βPIX and the Shank PDZ domain. The coiled-coil (CC) domain upstream of the PDZ binding motif allows multimerization of βPIX, which is important for its physiological functions. We have solved the crystal structure of the βPIX CC-Shank PDZ complex and determined the stoichiometry of complex formation. The βPIX CC forms a 76-Å-long parallel CC trimer. Despite the fact that the βPIX CC exposes three PDZ binding motifs in the C-termini, the βPIX trimer associates with a single Shank PDZ. One of the C-terminal ends of the CC forms an extensive β-sheet interaction with the Shank PDZ, while the other two ends are not involved in ligand binding and form random coils. The two C-terminal ends of βPIX have significantly lower affinity than the first PDZ binding motif due to the steric hindrance in the C-terminal tails, which results in binding of a single PDZ domain to the βPIX trimer. The structure shows canonical class I PDZ binding with a β-sheet interaction extending to position − 6 of βPIX. The βB-βC loop of Shank PDZ undergoes a conformational change upon ligand binding to form the β-sheet interaction and to accommodate the bulky side chain of Trp − 5. This structural study provides a clear picture of the molecular recognition of the PDZ ligand and the asymmetric association of βPIX CC and Shank PDZ.     

Signaling pathways of PDZ2 domain: a molecular dynamics interaction correlation analysis

PDZ domains are found in many signaling proteins. One of their functions is to provide scaffolds for forming membrane-associated protein complexes by binding to the carboxyl termini of their partners. PDZ domains are thought also to play a signal transduction role by propagating the information that binding has occurred to remote sites. In this study, a molecular dynamics (MD) simulation-based approach, referred to as an interaction correlation analysis, is applied to the PDZ2 domain to identify the possible signal transduction pathways. A residue correlation matrix is constructed from the interaction energy correlations between all residue pairs obtained from the MD simulations. Two continuous interaction pathways, starting at the ligand binding pocket, are identified by a hierarchical clustering analysis of the residue correlation matrix. One pathway is mainly localized at the N-terminal side of helix alpha1 and the adjacent C-terminus of loop beta1-beta2. The other pathway is perpendicular to the central beta-sheet and extends toward the side of PDZ2 domain opposite to the ligand binding pocket. The results complement previous studies based on multiple sequence analysis, NMR, and MD simulations. Importantly, they reveal the energetic origin of the long-range coupling. The PDZ2 results, as well as the earlier rhodopsin analysis, show that the interaction correlation analysis is a robust approach for determining pathways of intramolecular signal transduction.     

Quantification of PDZ domain specificity, prediction of ligand affinity and rational design of super-binding peptides

Transient macromolecular complexes are often formed by protein-protein interaction domains (e.g. PDZ, SH2, SH3, WW) which recognize linear sequence motifs with in vitro affinities typically in the micromolar range. The analysis of the resulting interaction networks requires a quantification of domain specificity and selectivity towards all possible ligands with physiologically relevant affinity. As representative examples, we determined specificity as a function of ligand sequence-dependent affinity contributions by statistical analysis of peptide library screens for the AF6, ERBIN and SNA1 (alpha-1-syntrophin) PDZ domains. For this purpose, the three PDZ domains were first screened for binding with a peptide library comprising 6223 human C termini created by SPOT synthesis. Based on the detected ligand preferences, we designed focused peptide libraries (profile libraries). These libraries were used to quantify the affinity contributions of the four C-terminal ligand residues by means of ANOVA models (analysis of variance) relating the C-terminal ligand sequences to the corresponding dissociation constants. Our models agreed well with experimentally determined dissociation constants and allowed us to design super binding peptides. The latter were shown experimentally to bind to their cognate PDZ domains with the highest affinity. In addition, we determined structure-activity relationships and thereby rationalized the position-specific affinity contributions. Furthermore, we used the statistical models to predict the dissociation constants for the complete ligand sequence space and thus determined the specificity overlap for the three investigated PDZ domains (). Altogether, we present an efficient method for profiling protein-protein interaction domains that provides a biophysical picture of specificity and selectivity. This approach allows the rational design of functional experiments and provides a basis for simulating interaction networks in the field of systems biology.     

An allosteric intramolecular PDZ-PDZ interaction modulates PTP-BL PDZ2 binding specificity

PDZ (acronym of the synapse-associated protein PSD-95/SAP90, the septate junction protein Discs-large, and the tight junction protein ZO-1) domains are abundant small globular protein interaction domains that mainly recognize the carboxyl termini of their target proteins. Detailed knowledge on PDZ domain binding specificity is a prerequisite for understanding the interaction networks they establish. We determined the binding preference of the five PDZ domains in the protein tyrosine phosphatase PTP-BL by screening a random C-terminal peptide lambda phage display library. Interestingly, the potential of PDZ2 to interact with class III-type ligands was found to be modulated by the presence of PDZ1. Structural studies revealed a direct and specific interaction of PDZ1 with a surface on PDZ2 that is opposite the peptide binding groove. Long-range allosteric effects that cause structural changes in the PDZ2 peptide binding groove thus explain the altered PDZ2 binding preference. Our results experimentally corroborate that the molecular embedding of PDZ domains is an important determinant of their ligand binding specificity.     

High‐energy water sites determine peptide binding affinity and specificity of PDZ domains

PDZ domains have well known binding preferences for distinct C‐terminal peptide motifs. For most PDZ domains, these motifs are of the form [S/T]‐W‐[I/L/V]. Although the preference for S/T has been explained by a specific hydrogen bond interaction with a histidine in the PDZ domain and the (I/L/V) is buried in a hydrophobic pocket, the mechanism for Trp specificity at the second to last position has thus far remained unknown. Here, we apply a method to compute the free energies of explicit water molecules and predict that potency gained by Trp binding is due to a favorable release of high‐energy water molecules into bulk. The affinities of a series of peptides for both wild‐type and mutant forms of the PDZ domain of Erbin correlate very well with the computed free energy of binding of displaced waters, suggesting a direct relationship between water displacement and peptide affinity. Finally, we show a correlation between the magnitude of the displaced water free energy and the degree of Trp‐sensitivity among subtypes of the HTRA PDZ family, indicating a water‐mediated mechanism for specificity of peptide binding.     

A specificity map for the PDZ domain family

PDZ domains are protein-protein interaction modules that recognize specific C-terminal sequences to assemble protein complexes in multicellular organisms. By scanning billions of random peptides, we accurately map binding specificity for approximately half of the over 330 PDZ domains in the human and Caenorhabditis elegans proteomes. The domains recognize features of the last seven ligand positions, and we find 16 distinct specificity classes conserved from worm to human, significantly extending the canonical two-class system based on position -2. Thus, most PDZ domains are not promiscuous, but rather are fine-tuned for specific interactions. Specificity profiling of 91 point mutants of a model PDZ domain reveals that the binding site is highly robust, as all mutants were able to recognize C-terminal peptides. However, many mutations altered specificity for ligand positions both close and far from the mutated position, suggesting that binding specificity can evolve rapidly under mutational pressure. Our specificity map enables the prediction and prioritization of natural protein interactions, which can be used to guide PDZ domain cell biology experiments. Using this approach, we predicted and validated several viral ligands for the PDZ domains of the SCRIB polarity protein. These findings indicate that many viruses produce PDZ ligands that disrupt host protein complexes for their own benefit, and that highly pathogenic strains target PDZ domains involved in cell polarity and growth.     

Molecular determinants for the complex binding specificity of the PDZ domain in PICK1

PICK1 (protein interacting with C kinase 1) contains a single PDZ domain known to mediate interaction with the C termini of several receptors, transporters, ion channels, and kinases. In contrast to most PDZ domains, the PICK1 PDZ domain interacts with binding sequences classifiable as type I (terminating in (S/T)XPhi; X, any residue) as well as type II (PhiXPhi; Phi, any hydrophobic residue). To enable direct assessment of the affinity of the PICK1 PDZ domain for its binding partners we developed a purification scheme for PICK1 and a novel quantitative binding assay based on fluorescence polarization. Our results showed that the PICK1 PDZ domain binds the type II sequence presented by the human dopamine transporter (-WLKV) with an almost 15-fold and >100-fold higher affinity than the type I sequences presented by protein kinase Calpha (-QSAV) and the beta(2)-adrenergic receptor (-DSLL), respectively. Mutational analysis of Lys(83) in the alphaB1 position of the PDZ domain suggested that this residue mimics the function of hydrophobic residues present in this position in regular type II PDZ domains. The PICK1 PDZ domain was moreover found to prefer small hydrophobic residues in the C-terminal P(0) position of the ligand. Molecular modeling predicted a rank order of (Val > Ile > Leu) that was verified experimentally with up to a approximately 16-fold difference in binding affinity between a valine and a leucine in P(0). The results define the structural basis for the unusual binding pattern of the PICK1 PDZ domain by substantiating the critical role of the alphaB1 position (Lys(83)) and of discrete side chain differences in position P(0) of the ligands.     

Redox-regulated affinity of the third PDZ domain in the phosphotyrosine phosphatase PTP-BL for cysteine-containing target peptides

PDZ domains are protein-protein interaction modules that are crucial for the assembly of structural and signalling complexes. They specifically bind to short C-terminal peptides and occasionally to internal sequences that structurally resemble such peptide termini. The binding of PDZ domains is dominated by the residues at the P(0) and P(-2) position within these C-terminal targets, but other residues are also important in determining specificity. In this study, we analysed the binding specificity of the third PDZ domain of protein tyrosine phosphatase BAS-like (PTP-BL) using a C-terminal combinatorial peptide phage library. Binding of PDZ3 to C-termini is preferentially governed by two cysteine residues at the P(-1) and P(-4) position and a valine residue at the P(0) position. Interestingly, we found that this binding is lost upon addition of the reducing agent dithiothrietol, indicating that the interaction is disulfide-bridge-dependent. Site-directed mutagenesis of the single cysteine residue in PDZ3 revealed that this bridge formation does not occur intermolecularly, between peptide and PDZ3 domain, but rather is intramolecular. These data point to a preference of PTP-BL PDZ3 for cyclic C-terminal targets, which may suggest a redox state-sensing role at the cell cortex.     

The binding of the PDZ tandem of syntenin to target proteins

PDZ domains are among the most abundant protein modules in the known genomes. Their main function is to provide scaffolds for membrane-associated protein complexes by binding to the cytosolic, C-terminal fragments of receptors, channels, and other integral membrane proteins. Here, using both heteronuclear NMR and single crystal X-ray diffraction, we show how peptides with different sequences, including those corresponding to the C-termini of syndecan, neurexin, and ephrin B, can simultaneously bind to both PDZ domains of the scaffolding protein syntenin. The PDZ2 domain binds these peptides in the canonical fashion, and an induced fit mechanism allows for the accommodation of a range of side chains in the P(0) and P(-)(2) positions. However, binding to the PDZ1 domain requires that the target peptide assume a noncanonical conformation. These data help explain how syntenin, and perhaps other PDZ-containing proteins, may preferentially bind to dimeric and clustered targets, and provide a mechanistic explanation for the previously reported cooperative ligand binding by syntenin's two PDZ domains.     

The PDZ1 domain of SAP90. Characterization of structure and binding.

The structural features of the PDZ1 domain of the synapse-associated protein SAP90 have been characterized by NMR. A comparison with the structures of the PDZ2 and PDZ3 domains of SAP90 illustrates significant differences, which may account for the unique binding properties of these homologous domains. Within the postsynaptic density, SAP90 functions as a molecular scaffold with a number of the protein-protein interactions mediated through the PDZ1 domain. Here, using fluorescence anisotropy and NMR chemical shift analysis, we have characterized the association of PDZ1 to the C-terminal peptides of the GluR6 subunit of the kainate receptor, voltage-gated K(+) channel Kv1.4, and microtubule-associate protein CRIPT, all of which are known to associate with SAP90. The latter two, which possess the consensus sequence for binding to PDZ domains (T/S-X-V-oh), have low micromolar binding affinities (1.5-15 microm). The C terminus of GluR6, RLPGKETMA-oh, lacking the consensus sequence, binds to PDZ1 of SAP90 with an affinity of 160 microm. The NMR data illustrate that although all three peptides occupy the binding groove capped by the GLGF loop of PDZ1, specific differences are present, consistent with the variation in binding affinities.     

Single-amino acid substitutions alter the specificity and affinity of PDZ domains for their ligands

PDZ domains are modular protein-protein interaction domains that bind to specific C-terminal sequences of membrane proteins and/or to other PDZ domains. Certain PDZ domains in PSD-95 and syntrophins interact with C-terminal peptide ligands and heterodimerize with the extended nNOS PDZ domain. The capacity to interact with nNOS correlates with the presence of a Lys residue in the carboxylate- binding loop of these PDZ domains. Here, we report that substitution of an Arg for Lys-165 in PSD-95 PDZ2 disrupted its interaction with nNOS, but not with the C terminus of the Shaker-type K(+) channel Kv1.4. The same mutation affected nNOS binding to alpha1- and beta1-syntrophin PDZ domains to a lesser extent, due in part to the stabilizing effect of tertiary interactions with the canonical nNOS PDZ domain. PDZ domains with an Arg in the carboxylate-binding loop do not bind nNOS; however, substitution with Lys or Ala was able to confer nNOS binding. Our results indicate that the carboxylate-binding loop Lys or Arg is a critical determinant of nNOS binding and that the identity of this residue can profoundly alter one mode of PDZ recognition without affecting another. We also analyzed the effects of mutating Asp-143, a residue in the alphaB helix of alpha1-syntrophin that forms a tertiary contact with the nNOS PDZ domain. This residue is important for both nNOS and C-terminal peptide binding and confers a preference for peptides with a positively charged residue at position -4. On this basis, we have identified the C terminus of the Kir2.1 channel as a possible binding partner for syntrophin PDZ domains. Together, our results demonstrate that single-amino acid substitutions alter the specificity and affinity of PDZ domains for their ligands.     

Inferring PDZ domain multi-mutant binding preferences from single-mutant data

Many important cellular protein interactions are mediated by peptide recognition domains. The ability to predict a domain's binding specificity directly from its primary sequence is essential to understanding the complexity of protein-protein interaction networks. One such recognition domain is the PDZ domain, functioning in scaffold proteins that facilitate formation of signaling networks. Predicting the PDZ domain's binding specificity was a part of the DREAM4 Peptide Recognition Domain challenge, the goal of which was to describe, as position weight matrices, the specificity profiles of five multi-mutant ERBB2IP-1 domains. We developed a method that derives multi-mutant binding preferences by generalizing the effects of single point mutations on the wild type domain's binding specificities. Our approach, trained on publicly available ERBB2IP-1 single-mutant phage display data, combined linear regression-based prediction for ligand positions whose specificity is determined by few PDZ positions, and single-mutant position weight matrix averaging for all other ligand columns. The success of our method as the winning entry of the DREAM4 competition, as well as its superior performance over a general PDZ-ligand binding model, demonstrates the advantages of training a model on a well-selected domain-specific data set.