Computational Analysis of Antibody Paratopes for Antibody Sequences in Antibody Libraries.

To ensure the functionalities of the antibodies in phage-displayed synthetic antibody libraries, we use computational method to evaluate the designs of the antibody libraries. The computational methodologies developed in our lab for designing antibody library provide rich information on the function of the designed antibody sequences-adequate antibody designs for a specific antigen type should have predicted paratopes for the antigen type. This computational assessment of the designed antibody sequences helps eliminate non-functional designs before proceeding to construct the library designs in the wet lab. As such, only reasonable antibody designs are constructed for antibody discoveries.

Antibody CDR amino acids underlying the functionality of antibody repertoires in recognizing diverse protein antigens.

Antibodies recognize protein antigens with exquisite specificity in a complex aqueous environment, where interfacial waters are an integral part of the antibody-protein complex interfaces. In this work, we elucidate, with computational analyses, the principles governing the antibodies' specificity and affinity towards their cognate protein antigens in the presence of explicit interfacial waters. Experimentally, in four model antibody-protein complexes, we compared the contributions of the interaction types in antibody-protein antigen complex interfaces with the antibody variants selected from phage-displayed synthetic antibody libraries. Evidently, the specific interactions involving a subset of aromatic CDR (complementarity determining region) residues largely form the predominant determinant underlying the specificity of the antibody-protein complexes in nature. The interfacial direct/water-mediated hydrogen bonds accompanying the CDR aromatic interactions are optimized locally but contribute little in determining the epitope location. The results provide insights into the phenomenon that natural antibodies with limited sequence and structural variations in an antibody repertoire can recognize seemingly unlimited protein antigens. Our work suggests guidelines in designing functional artificial antibody repertoires with practical applications in developing novel antibody-based therapeutics and diagnostics for treating and preventing human diseases.

Attitude towards hereditary cancer risk management among women with cancer in Taiwan.

BACKGROUND: Risk management intentions prior to genetic counseling predict risk management uptake following genetic testing. Limited studies examined the attitude and understanding towards genetic counseling/testing in underserved countries. The purposes of this study were to explore knowledge and attitude towards genetic counseling, testing, and risk management for breast and ovarian cancer, and to understand the factors influencing risk management intentions in women with cancer in Taiwan. METHODS: Cross-sectional with correlational design was used in this study. Participants were enrolled for genetic testing based on clinical criteria suspected of having hereditary cancer. Survey was conducted using a standardized questionnaire including (1) demographics and personal/family history of cancer; (2) prior experience or consideration of genetic testing and reasons for not considering; (3) perception and attitude towards genetic counseling; and (4) intentions for risk management with a hypothetical BRCA1 mutation status. Multinomial logistic regression was used to analyze the predictors of participants' intentions for cancer risk management strategies. RESULTS: A total of 430 women with cancer were analyzed in which 51.6% had family history of cancer in first-degree relatives. Only 30.7% had considered genetic testing and 28.4% had known about genetic counseling prior to the study. When prompted with the services of genetic counseling, the attitude towards genetic counseling was fairly positive (score of 19.8 ± 2.9 out of 25). Given hypothetical BRCA1 mutation status, enhanced breast cancer screening with annual breast MRI was much more accepted than cancer risk reducing interventions. More positive attitude towards genetic counseling (each score point increase) was associated with higher odds of intention for breast MRI (OR 1.20, 95% CI 1.09-1.32) and preventive tamoxifen (OR 1.11, 95% CI 1.02-1.22). Having considered genetic testing prior to the study was associated with higher odds of intention for all four risk management strategies: breast MRI (OR 2.99, 95% CI 1.46-6.11), preventive tamoxifen (OR 1.79, 95% CI 1.00-3.17), risk-reducing mastectomy (OR 2.24, 95% CI 1.13-4.42), and risk-reducing salpingo-oophorectomy (OR 2.69, 95% CI 1.27-6.93). CONCLUSION: Knowledge of genetic testing and positive attitude towards genetic counseling were associated with increased willingness to consider cancer risk management strategies for hereditary breast and ovarian cancer syndrome. Given the limited knowledge on genetic testing and counseling in the studied population, increasing public awareness of these services may increase adoption of the risk management strategies.

Eradicating mesothelin-positive human gastric and pancreatic tumors in xenograft models with optimized anti-mesothelin antibody-drug conjugates from synthetic antibody libraries.

Mesothelin (MSLN) is an attractive candidate of targeted therapy for several cancers, and hence there are increasing needs to develop MSLN-targeting strategies for cancer therapeutics. Antibody-drug conjugates (ADCs) targeting MSLN have been demonstrated to be a viable strategy in treating MSLN-positive cancers. However, developing antibodies as targeting modules in ADCs for toxic payload delivery to the tumor site but not to normal tissues is not a straightforward task with many potential hurdles. In this work, we established a high throughput engineering platform to develop and optimize anti-MSLN ADCs by characterizing more than 300 scFv CDR-variants and more than 50 IgG CDR-variants of a parent anti-MSLN antibody as candidates for ADCs. The results indicate that only a small portion of the complementarity determining region (CDR) residues are indispensable in the MSLN-specific targeting. Also, the enhancement of the hydrophilicity of the rest of the CDR residues could drastically increase the overall solubility of the optimized anti-MSLN antibodies, and thus substantially improve the efficacies of the ADCs in treating human gastric and pancreatic tumor xenograft models in mice. We demonstrated that the in vivo treatments with the optimized ADCs resulted in almost complete eradication of the xenograft tumors at the treatment endpoints, without detectable off-target toxicity because of the ADCs' high specificity targeting the cell surface tumor-associated MSLN. The technological platform can be applied to optimize the antibody sequences for more effective targeting modules of ADCs, even when the candidate antibodies are not necessarily feasible for the ADC development due to the antibodies' inferior solubility or affinity/specificity to the target antigen.

A panel of anti-influenza virus nucleoprotein antibodies selected from phage-displayed synthetic antibody libraries with rapid diagnostic capability to distinguish diverse influenza virus subtypes.

Immunoassays based on sandwich immuno-complexes of capture and detection antibodies simultaneously binding to the target analytes have been powerful technologies in molecular analyses. Recent developments in single molecule detection technologies enable the detection limit of the sandwich immunoassays approaching femtomolar (10-15 M), driving the needs of developing sensitive and specific antibodies for ever-increasingly broad applications in detecting and quantifying biomarkers. The key components underlying the sandwich immunoassays are antibody-based affinity reagents, for which the conventional sources are mono- or poly-clonal antibodies from immunized animals. The downsides of the animal-based antibodies as affinity reagents arise from the requirement of months of development timespan and limited choices of antibody candidates due to immunodominance of humoral immune responses in animals. Hence, developing animal antibodies capable of distinguishing highly related antigens could be challenging. To overcome the limitation imposed by the animal immune systems, we developed an in vitro methodology based on phage-displayed synthetic antibody libraries for diverse antibodies as affinity reagents against closely related influenza virus nucleoprotein (NP) subtypes, aiming to differentiating avian influenza virus (H5N1) from seasonal influenza viruses (H1N1 and H3N2), for which the NPs are closely related by 90-94% in terms of pairwise amino acid sequence identity. We applied the methodology to attain, within four weeks, a panel of IgGs with distinguishable specificities against a group of representative NPs with pairwise amino acid sequence identities up to more than 90%, and the antibodies derived from the antibody libraries without further affinity refinement had comparable affinity of mouse antibodies to the NPs with the detection limit less than 1 nM of viral NP from lysed virus with sandwich ELISA. The panel of IgGs were capable of rapidly distinguishing infections due to virulent avian influenza virus from infections of seasonal flu, in responding to a probable emergency scenario where avian influenza virus would be transmissible among humans overlapping with the seasonal influenza infections. The results indicate that the in vitro antibody development methodology enables developing diagnostic antibodies that would not otherwise be available from animal-based antibody technologies.

Evaluating BRCA mutation risk predictive models in a Chinese cohort in Taiwan.

Accurate estimation of carrier probabilities of cancer susceptibility gene mutations is an important part of pre-test genetic counselling. Many predictive models are available but their applicability in the Asian population is uncertain. We evaluated the performance of five BRCA mutation risk predictive models in a Chinese cohort of 647 women, who underwent germline DNA sequencing of a cancer susceptibility gene panel. Using areas under the curve (AUCs) on receiver operating characteristics (ROC) curves as performance measures, the models did comparably well as in western cohorts (BOADICEA 0.75, BRCAPRO 0.73, Penn II 0.69, Myriad 0.68). For unaffected women with family history of breast or ovarian cancer (n = 144), BOADICEA, BRCAPRO, and Tyrer-Cuzick models had excellent performance (AUC 0.93, 0.92, and 0.92, respectively). For women with both personal and family history of breast or ovarian cancer (n = 241), all models performed fairly well (BOADICEA 0.79, BRCAPRO 0.79, Penn II 0.75, Myriad 0.70). For women with personal history of breast or ovarian cancer but no family history (n = 262), most models did poorly. Between the two well-performed models, BOADICEA underestimated mutation risks while BRCAPRO overestimated mutation risks (expected/observed ratio 0.67 and 2.34, respectively). Among 424 women with personal history of breast cancer and available tumor ER/PR/HER2 data, the predictive models performed better for women with triple negative breast cancer (AUC 0.74 to 0.80) than for women with luminal or HER2 overexpressed breast cancer (AUC 0.63 to 0.69). However, incorporating ER/PR/HER2 status into the BOADICEA model calculation did not improve its predictive accuracy.

Effective binding to protein antigens by antibodies from antibody libraries designed with enhanced protein recognition propensities.

Antibodies provide immune protection by recognizing antigens of diverse chemical properties, but elucidating the amino acid sequence-function relationships underlying the specificity and affinity of antibody-antigen interactions remains challenging. We designed and constructed phage-displayed synthetic antibody libraries with enriched protein antigen-recognition propensities calculated with machine learning predictors, which indicated that the designed single-chain variable fragment variants were encoded with enhanced distributions of complementarity-determining region (CDR) hot spot residues with high protein antigen recognition propensities in comparison with those in the human antibody germline sequences. Antibodies derived directly from the synthetic antibody libraries, without affinity maturation cycles comparable to those in in vivo immune systems, bound to the corresponding protein antigen through diverse conformational or linear epitopes with specificity and affinity comparable to those of the affinity-matured antibodies from in vivo immune systems. The results indicated that more densely populated CDR hot spot residues were sustainable by the antibody structural frameworks and could be accompanied by enhanced functionalities in recognizing protein antigens. Our study results suggest that synthetic antibody libraries, which are not limited by the sequences found in antibodies in nature, could be designed with the guidance of the computational machine learning algorithms that are programmed to predict interaction propensities to molecules of diverse chemical properties, leading to antibodies with optimal characteristics pertinent to their medical applications.

Antibody-drug conjugates with HER2-targeting antibodies from synthetic antibody libraries are highly potent against HER2-positive human gastric tumor in xenograft models.

HER2-ECD (human epidermal growth factor receptor 2 - extracellular domain) is a prominent therapeutic target validated for treating HER2-positive breast and gastric cancer, but HER2-specific therapeutic options for treating advanced gastric cancer remain limited. We have developed antibody-drug conjugates (ADCs), comprising IgG1 linked via valine-citrulline to monomethyl auristatin E, with potential to treat HER2-positive gastric cancer in humans. The antibodies optimally selected from the ADC discovery platform, which was developed to discover antibody candidates suitable for immunoconjugates from synthetic antibody libraries designed using antibody-antigen interaction principles, were demonstrated to be superior immunoconjugate targeting modules in terms of efficacy and off-target toxicity. In comparison with the two control humanized antibodies (trastuzumab and H32) derived from murine antibody repertoires, the antibodies derived from the synthetic antibody libraries had enhanced receptor-mediated internalization rate, which could result in ADCs with optimal efficacies. Along with the ADCs, two other forms of immunoconjugates (scFv-PE38KDEL and IgG1-AL1-PE38KDEL) were used to test the antibodies for delivering cytotoxic payloads to xenograft tumor models in vivo and to cultured cells in vitro. The in vivo experiments with the three forms of immunoconjugates revealed minimal off-target toxicities of the selected antibodies from the synthetic antibody libraries; the off-target toxicities of the control antibodies could have resulted from the antibodies' propensity to target the liver in the animal models. Our ADC discovery platform and the knowledge gained from our in vivo tests on xenograft models with the three forms of immunoconjugates could be useful to anyone developing optimal ADC cancer therapeutics.

Germline breast cancer susceptibility gene mutations and breast cancer outcomes.

BACKGROUND: It is unclear whether germline breast cancer susceptibility gene mutations affect breast cancer related outcomes. We wanted to evaluate mutation patterns in 20 breast cancer susceptibility genes and correlate the mutations with clinical characteristics to determine the effects of these germline mutations on breast cancer prognosis. METHODS: The study cohort included 480 ethnic Chinese individuals in Taiwan with at least one of the six clinical risk factors for hereditary breast cancer: family history of breast or ovarian cancer, young age of onset for breast cancer, bilateral breast cancer, triple negative breast cancer, both breast and ovarian cancer, and male breast cancer. PCR-enriched amplicon-sequencing on a next generation sequencing platform was used to determine the germline DNA sequences of all exons and exon-flanking regions of the 20 genes. Protein-truncating variants were identified as pathogenic. RESULTS: We detected a 13.5% carrier rate of pathogenic germline mutations, with BRCA2 being the most prevalent and the non-BRCA genes accounting for 38.5% of the mutation carriers. BRCA mutation carriers were more likely to be diagnosed of breast cancer with lymph node involvement (66.7% vs 42.6%; P = 0.011), and had significantly worse breast cancer specific outcomes. The 5-year disease-free survival was 73.3% for BRCA mutation carriers and 91.1% for non-carriers (hazard ratio for recurrence or death 2.42, 95% CI 1.29-4.53; P = 0.013). After adjusting for clinical prognostic factors, BRCA mutation remained an independent poor prognostic factor for cancer recurrence or death (adjusted hazard ratio 3.04, 95% CI 1.40-6.58; P = 0.005). Non-BRCA gene mutation carriers did not exhibit any significant difference in cancer characteristics or outcomes compared to those without detected mutations. Among the risk factors for hereditary breast cancer, the odds of detecting a germline mutation increased significantly with having bilateral breast cancer (adjusted odds ratio 3.27, 95% CI 1.64-6.51; P = 0.0008) or having more than one risk factor (odds ratio 2.07, 95% CI 1.22-3.51; P = 0.007). CONCLUSIONS: Without prior knowledge of the mutation status, BRCA mutation carriers had more advanced breast cancer on initial diagnosis and worse cancer-related outcomes. Optimal approach to breast cancer treatment for BRCA mutation carriers warrants further investigation.

Predicting Ligand Binding Sites on Protein Surfaces by 3-Dimensional Probability Density Distributions of Interacting Atoms.

Predicting ligand binding sites (LBSs) on protein structures, which are obtained either from experimental or computational methods, is a useful first step in functional annotation or structure-based drug design for the protein structures. In this work, the structure-based machine learning algorithm ISMBLab-LIG was developed to predict LBSs on protein surfaces with input attributes derived from the three-dimensional probability density maps of interacting atoms, which were reconstructed on the query protein surfaces and were relatively insensitive to local conformational variations of the tentative ligand binding sites. The prediction accuracy of the ISMBLab-LIG predictors is comparable to that of the best LBS predictors benchmarked on several well-established testing datasets. More importantly, the ISMBLab-LIG algorithm has substantial tolerance to the prediction uncertainties of computationally derived protein structure models. As such, the method is particularly useful for predicting LBSs not only on experimental protein structures without known LBS templates in the database but also on computationally predicted model protein structures with structural uncertainties in the tentative ligand binding sites.

Discovering neutralizing antibodies targeting the stem epitope of H1N1 influenza hemagglutinin with synthetic phage-displayed antibody libraries.

Broadly neutralizing antibodies developed from the IGHV1-69 germline gene are known to bind to the stem region of hemagglutinin in diverse influenza viruses but the sequence determinants for the antigen recognition, including neutralization potency and binding affinity, are not clearly understood. Such understanding could inform designs of synthetic antibody libraries targeting the stem epitope on hemagglutinin, leading to artificially designed antibodies that are functionally advantageous over antibodies from natural antibody repertoires. In this work, the sequence space of the complementarity determining regions of a broadly neutralizing antibody (F10) targeting the stem epitope on the hemagglutinin of a strain of H1N1 influenza virus was systematically explored; the elucidated antibody-hemagglutinin recognition principles were used to design a phage-displayed antibody library, which was then used to discover neutralizing antibodies against another strain of H1N1 virus. More than 1000 functional antibody candidates were selected from the antibody library and were shown to neutralize the corresponding strain of influenza virus with up to 7 folds higher potency comparing with the parent F10 antibody. The antibody library could be used to discover functionally effective antibodies against other H1N1 influenza viruses, supporting the notion that target-specific antibody libraries can be designed and constructed with systematic sequence-function information.

Predominant structural configuration of natural antibody repertoires enables potent antibody responses against protein antigens.

Humoral immunity against diverse pathogens is rapidly elicited from natural antibody repertoires of limited complexity. But the organizing principles underlying the antibody repertoires that facilitate this immunity are not well-understood. We used HER2 as a model immunogen and reverse-engineered murine antibody response through constructing an artificial antibody library encoded with rudimentary sequence and structural characteristics learned from high throughput sequencing of antibody variable domains. Antibodies selected in vitro from the phage-displayed synthetic antibody library bound to the model immunogen with high affinity and specificities, which reproduced the specificities of natural antibody responses. We conclude that natural antibody structural repertoires are shaped to allow functional antibodies to be encoded efficiently, within the complexity limit of an individual antibody repertoire, to bind to diverse protein antigens with high specificity and affinity. Phage-displayed synthetic antibody libraries, in conjunction with high-throughput sequencing, can thus be designed to replicate natural antibody responses and to generate novel antibodies against diverse antigens.

Structural Repertoire of HIV-1-Neutralizing Antibodies Targeting the CD4 Supersite in 14 Donors.

The site on the HIV-1 gp120 glycoprotein that binds the CD4 receptor is recognized by broadly reactive antibodies, several of which neutralize over 90% of HIV-1 strains. To understand how antibodies achieve such neutralization, we isolated CD4-binding-site (CD4bs) antibodies and analyzed 16 co-crystal structures -8 determined here- of CD4bs antibodies from 14 donors. The 16 antibodies segregated by recognition mode and developmental ontogeny into two types: CDR H3-dominated and VH-gene-restricted. Both could achieve greater than 80% neutralization breadth, and both could develop in the same donor. Although paratope chemistries differed, all 16 gp120-CD4bs antibody complexes showed geometric similarity, with antibody-neutralization breadth correlating with antibody-angle of approach relative to the most effective antibody of each type. The repertoire for effective recognition of the CD4 supersite thus comprises antibodies with distinct paratopes arrayed about two optimal geometric orientations, one achieved by CDR H3 ontogenies and the other achieved by VH-gene-restricted ontogenies.

Prediction of fatty acid-binding residues on protein surfaces with three-dimensional probability distributions of interacting atoms.

Protein-fatty acid interaction is vital for many cellular processes and understanding this interaction is important for functional annotation as well as drug discovery. In this work, we present a method for predicting the fatty acid (FA)-binding residues by using three-dimensional probability density distributions of interacting atoms of FAs on protein surfaces which are derived from the known protein-FA complex structures. A machine learning algorithm was established to learn the characteristic patterns of the probability density maps specific to the FA-binding sites. The predictor was trained with five-fold cross validation on a non-redundant training set and then evaluated with an independent test set as well as on holo-apo pair's dataset. The results showed good accuracy in predicting the FA-binding residues. Further, the predictor developed in this study is implemented as an online server which is freely accessible at the following website, http://ismblab.genomics.sinica.edu.tw/.

Origins of specificity and affinity in antibody-protein interactions.

Natural antibodies are frequently elicited to recognize diverse protein surfaces, where the sequence features of the epitopes are frequently indistinguishable from those of nonepitope protein surfaces. It is not clearly understood how the paratopes are able to recognize sequence-wise featureless epitopes and how a natural antibody repertoire with limited variants can recognize seemingly unlimited protein antigens foreign to the host immune system. In this work, computational methods were used to predict the functional paratopes with the 3D antibody variable domain structure as input. The predicted functional paratopes were reasonably validated by the hot spot residues known from experimental alanine scanning measurements. The functional paratope (hot spot) predictions on a set of 111 antibody-antigen complex structures indicate that aromatic, mostly tyrosyl, side chains constitute the major part of the predicted functional paratopes, with short-chain hydrophilic residues forming the minor portion of the predicted functional paratopes. These aromatic side chains interact mostly with the epitope main chain atoms and side-chain carbons. The functional paratopes are surrounded by favorable polar atomistic contacts in the structural paratope-epitope interfaces; more that 80% these polar contacts are electrostatically favorable and about 40% of these polar contacts form direct hydrogen bonds across the interfaces. These results indicate that a limited repertoire of antibodies bearing paratopes with diverse structural contours enriched with aromatic side chains among short-chain hydrophilic residues can recognize all sorts of protein surfaces, because the determinants for antibody recognition are common physicochemical features ubiquitously distributed over all protein surfaces.

Prediction of FMN-binding residues with three-dimensional probability distributions of interacting atoms on protein surfaces.

Flavin mono-nucleotide (FMN) is a cofactor which is involved in many biological reactions. The insights on protein-FMN interactions aid the protein functional annotation and also facilitate in drug design. In this study, we have established a new method, making use of an encoding scheme of the three-dimensional probability density maps that describe the distributions of 40 non-covalent interacting atom types around protein surfaces, to predict FMN-binding sites on protein surfaces. One machine learning model was trained for each of the 30 protein atom types to predict tentative FMN-binding sites on protein structures. The method's capability was evaluated by five-fold cross-validation on a dataset containing 81 non-redundant FMN-binding protein structures and further tested on independent datasets of 30 and 15 non-redundant protein structures respectively. These predictions achieved an accuracy of 0.94, 0.94 and 0.96 with the Matthews correlation coefficient (MCC) of 0.53, 0.53 and 0.65 respectively for the three protein structure sets. The prediction capability is superior to the existing method. This is the first structure-based approach that does not rely on evolutionary information for predicting FMN-interacting residues. The webserver for the prediction is available at http://ismblab.genomics.sinica.edu.tw/.

Antibody variable domain interface and framework sequence requirements for stability and function by high-throughput experiments.

Protein structural stability and biological functionality are dictated by the formation of intradomain cores and interdomain interfaces, but the intricate sequence-structure-function interrelationships in the packing of protein cores and interfaces remain difficult to elucidate due to the intractability of enumerating all packing possibilities and assessing the consequences of all the variations. In this work, groups of ß strand residues of model antibody variable domains were randomized with saturated mutagenesis and the functional variants were selected for high-throughput sequencing and high-throughput thermal stability measurements. The results show that the sequence preferences of the intradomain hydrophobic core residues are strikingly flexible among hydrophobic residues, implying that these residues are coupled indirectly with antigen binding through energetic stabilization of the protein structures. By contrast, the interdomain interface residues are directly coupled with antigen binding. The interdomain interface should be treated as an integral part of the antigen-binding site.

Loop-sequence features and stability determinants in antibody variable domains by high-throughput experiments.

Protein loops are frequently considered as critical determinants in protein structure and function. Recent advances in high-throughput methods for DNA sequencing and thermal stability measurement have enabled effective exploration of sequence-structure-function relationships in local protein regions. Using these data-intensive technologies, we investigated the sequence-structure-function relationships of six complementarity-determining regions (CDRs) and ten non-CDR loops in the variable domains of a model vascular endothelial growth factor (VEGF)-binding single-chain antibody variable fragment (scFv) whose sequence had been optimized via a consensus-sequence approach. The results show that only a handful of residues involving long-range tertiary interactions distant from the antigen-binding site are strongly coupled with antigen binding. This implies that the loops are passive regions in protein folding; the essential sequences of these regions are dictated by conserved tertiary interactions and the consensus local loop-sequence features contribute little to protein stability and function.

Amino acid substitutions of MagA in Klebsiella pneumoniae affect the biosynthesis of the capsular polysaccharide.

Mucoviscosity-associated gene A (magA) of Klebsiella pneumoniae contributes to K1 capsular polysaccharide (CPS) biosynthesis. Based on sequence homology and gene alignment, the magA gene has been predicted to encode a Wzy-type CPS polymerase. Sequence alignment with the Wzy_C and RfaL protein families (which catalyze CPS or lipopolysaccharide (LPS) biosynthesis) and topological analysis has suggested that eight highly conserved residues, including G308, G310, G334, G337, R290, P305, H323, and N324, were located in a hypothetical loop region. Therefore, we used site-directed mutagenesis to study the role of these residues in CPS production, and to observe the consequent phenotypes such as mucoviscosity, serum and phagocytosis resistance, and virulence (as assessed in mice) in pyogenic liver abscess strain NTUH-K2044. Alanine substitutions at R290 or H323 abolished all of these properties. The G308A mutant was severely impaired for these functions. The G334A mutant remained mucoid with decreased CPS production, but its virulence was significantly reduced in vivo. No phenotypic change was observed for strains harboring magA G310A, G337A, P305A, or N324A mutations. Therefore, R290, G308, H323, and G334 are functionally important residues of the MagA (Wzy) protein of K. pneumoniae NTUH-K2044, capsular type K1. These amino acids are also likely to be important for the function of Wzy in other capsular types in K. pneumoniae and other species bearing Wzy_C family proteins.

Prediction of carbohydrate binding sites on protein surfaces with 3-dimensional probability density distributions of interacting atoms.

Non-covalent protein-carbohydrate interactions mediate molecular targeting in many biological processes. Prediction of non-covalent carbohydrate binding sites on protein surfaces not only provides insights into the functions of the query proteins; information on key carbohydrate-binding residues could suggest site-directed mutagenesis experiments, design therapeutics targeting carbohydrate-binding proteins, and provide guidance in engineering protein-carbohydrate interactions. In this work, we show that non-covalent carbohydrate binding sites on protein surfaces can be predicted with relatively high accuracy when the query protein structures are known. The prediction capabilities were based on a novel encoding scheme of the three-dimensional probability density maps describing the distributions of 36 non-covalent interacting atom types around protein surfaces. One machine learning model was trained for each of the 30 protein atom types. The machine learning algorithms predicted tentative carbohydrate binding sites on query proteins by recognizing the characteristic interacting atom distribution patterns specific for carbohydrate binding sites from known protein structures. The prediction results for all protein atom types were integrated into surface patches as tentative carbohydrate binding sites based on normalized prediction confidence level. The prediction capabilities of the predictors were benchmarked by a 10-fold cross validation on 497 non-redundant proteins with known carbohydrate binding sites. The predictors were further tested on an independent test set with 108 proteins. The residue-based Matthews correlation coefficient (MCC) for the independent test was 0.45, with prediction precision and sensitivity (or recall) of 0.45 and 0.49 respectively. In addition, 111 unbound carbohydrate-binding protein structures for which the structures were determined in the absence of the carbohydrate ligands were predicted with the trained predictors. The overall prediction MCC was 0.49. Independent tests on anti-carbohydrate antibodies showed that the carbohydrate antigen binding sites were predicted with comparable accuracy. These results demonstrate that the predictors are among the best in carbohydrate binding site predictions to date.