Prediction of Protein Cleavage Site with Feature Selection by Random Forest

Proteinases play critical roles in both intra and extracellular processes by binding and cleaving their protein substrates. The cleavage can either be non-specific as part of degradation during protein catabolism or highly specific as part of proteolytic cascades and signal transduction events. Identification of these targets is extremely challenging. Current computational approaches for predicting cleavage sites are very limited since they mainly represent the amino acid sequences as patterns or frequency matrices. In this work, we developed a novel predictor based on Random Forest algorithm (RF) using maximum relevance minimum redundancy (mRMR) method followed by incremental feature selection (IFS). The features of physicochemical/biochemical properties, sequence conservation, residual disorder, amino acid occurrence frequency, secondary structure and solvent accessibility were utilized to represent the peptides concerned. Here, we compared existing prediction tools which are available for predicting possible cleavage sites in candidate substrates with ours. It is shown that our method makes much more reliable predictions in terms of the overall prediction accuracy. In addition, this predictor allows the use of a wide range of proteinases.     

Class Prediction and Feature Selection with Linear Optimization for Metagenomic Count Data

The amount of metagenomic data is growing rapidly while the computational methods for metagenome analysis are still in their infancy. It is important to develop novel statistical learning tools for the prediction of associations between bacterial communities and disease phenotypes and for the detection of differentially abundant features. In this study, we presented a novel statistical learning method for simultaneous association prediction and feature selection with metagenomic samples from two or multiple treatment populations on the basis of count data. We developed a linear programming based support vector machine with and joint penalties for binary and multiclass classifications with metagenomic count data (metalinprog). We evaluated the performance of our method on several real and simulation datasets. The proposed method can simultaneously identify features and predict classes with the metagenomic count data.     

Protease Cleavage Leads to Formation of Mature Trimer Interface in HIV-1 Capsid

During retrovirus particle maturation, the assembled Gag polyprotein is cleaved by the viral protease into matrix (MA), capsid (CA), and nucleocapsid (NC) proteins. To form the mature viral capsid, CA rearranges, resulting in a lattice composed of hexameric and pentameric CA units. Recent structural studies of assembled HIV-1 CA revealed several inter-subunit interfaces in the capsid lattice, including a three-fold interhexamer interface that is critical for proper capsid stability. Although a general architecture of immature particles has been provided by cryo-electron tomographic studies, the structural details of the immature particle and the maturation pathway remain unknown. Here, we used cryo-electron microscopy (cryoEM) to determine the structure of tubular assemblies of the HIV-1 CA-SP1-NC protein. Relative to the mature assembled CA structure, we observed a marked conformational difference in the position of the CA-CTD relative to the NTD in the CA-SP1-NC assembly, involving the flexible hinge connecting the two domains. This difference was verified via engineered disulfide crosslinking, revealing that inter-hexamer contacts, in particular those at the pseudo three-fold axis, are altered in the CA-SP1-NC assemblies compared to the CA assemblies. Results from crosslinking analyses of mature and immature HIV-1 particles containing the same Cys substitutions in the Gag protein are consistent with these findings. We further show that cleavage of preassembled CA-SP1-NC by HIV-1 protease in vitro leads to release of SP1 and NC without disassembly of the lattice. Collectively, our results indicate that the proteolytic cleavage of Gag leads to a structural reorganization of the polypeptide and creates the three-fold interhexamer interface, important for the formation of infectious HIV-1 particles.     

Studies on Substrate Specificity at PR/p3 Cleavage Site of HTLV-1 Protease

Substrate specificities for recognition at the PR/p3 site of HTLV-1 protease were clarified using small libraries of substrate peptides. Specificities at P₁ and P₁′ positions were examined by parallel synthesis/digestion of synthetic peptides covering the PR/p3 site (KGPPVILPIQA). Specificities at P₂ to P₄ positions were examined by split and mix syntheses of olefin-peptide libraries containing the substrate sequence (PPVILPIQ). The solid-phase Horner-Emmons reaction was successfully applied to syntheses of multi-component substrates for library preparation. From the digestion of substrate peptides by a chemically synthesized mutant of HTLV-1 protease (C2A HTLV-1 PR), it was found for the first time that the preference for Pro at the P₁′ position and for Ile at the P₂ position is unique for this enzyme. We dedicate this article to Prof. Bruce Merrifield for his great role and impact on solid-phase chemistry.     

HIV-1蛋白酶的表达、纯化及其抑制剂体外筛选方法的建立

从HIV-1IIIB病毒RNA经RT-PCR得到HIV-1蛋白酶编码序列,克隆到pet28a质粒中构建HIV-1蛋白酶表达载体。阳性克隆转染E.coliBL21DE3,经IPTG诱导,蛋白酶以包涵体的形式表达,表达量占菌体总蛋白量的40%。包涵体经TritonX-100洗涤后溶解于8M尿素,溶解后的蛋白溶液经sephacyls-200H.R分子筛柱纯化后纯度达到90%以上,收集蛋白酶峰稀释复性并通过超滤进行浓缩。经检测,纯化的蛋白酶具有较高的活性。用荧光标记的蛋白酶底物检测不同浓度indinavir对蛋白酶活性的影响,表明该方法可以用于蛋白酶抑制剂的筛选。     

HIV-1蛋白酶的表达、纯化及其抑制剂体外筛选方法的建立

从HIV-1 ⅢB病毒RNA经RT-PCR得到HIV-1蛋白酶编码序列,克隆到pet28a质粒中构建HIV-1蛋白酶表达载体.阳性克隆转染E.coli BL21 DE3,经IPTG诱导,蛋白酶以包涵体的形式表达,表达量占菌体总蛋白量的40%.包涵体经Triton X-100洗涤后溶解于8M尿素,溶解后的蛋白溶液经sephacyl s-200 H.R分子筛柱纯化后纯度达到90%以上,收集蛋白酶峰稀释复性并通过超滤进行浓缩.经检测,纯化的蛋白酶具有较高的活性.用荧光标记的蛋白酶底物检测不同浓度indinavir对蛋白酶活性的影响,表明该方法可以用于蛋白酶抑制剂的筛选.     

HIV- 1 Protease Inhibits Cap- and Poly(A)-Dependent Translation upon eIF4GI and PABP Cleavage

A number of viral proteases are able to cleave translation initiation factors leading to the inhibition of cellular translation. This is the case of human immunodeficiency virus type 1 protease (HIV-1 PR), which hydrolyzes eIF4GI and poly(A)-binding protein (PABP). Here, the effect of HIV-1 PR on cellular and viral protein synthesis has been examined using cell-free systems. HIV-1 PR strongly hampers translation of pre-existing capped luc mRNAs, particularly when these mRNAs contain a poly(A) tail. In fact, HIV-1 PR efficiently blocks cap- and poly(A)-dependent translation initiation in HeLa extracts. Addition of exogenous PABP to HIV-1 PR treated extracts partially restores the translation of polyadenylated luc mRNAs, suggesting that PABP cleavage is directly involved in the inhibition of poly(A)-dependent translation. In contrast to these data, PABP cleavage induced by HIV-1 PR has little impact on the translation of polyadenylated encephalomyocarditis virus internal ribosome entry site (IRES)-containing mRNAs. In this case, the loss of poly(A)-dependent translation is compensated by the IRES transactivation provided by eIF4G cleavage. Finally, translation of capped and polyadenylated HIV-1 genomic mRNA takes place in HeLa extracts when eIF4GI and PABP have been cleaved by HIV-1 PR. Together these results suggest that proteolytic cleavage of eIF4GI and PABP by HIV-1 PR blocks cap- and poly(A)-dependent initiation of translation, leading to the inhibition of cellular protein synthesis. However, HIV-1 genomic mRNA can be translated under these conditions, giving rise to the production of Gag polyprotein.     

A Kernel-Based Multivariate Feature Selection Method for Microarray Data Classification

High dimensionality and small sample sizes, and their inherent risk of overfitting, pose great challenges for constructing efficient classifiers in microarray data classification. Therefore a feature selection technique should be conducted prior to data classification to enhance prediction performance. In general, filter methods can be considered as principal or auxiliary selection mechanism because of their simplicity, scalability, and low computational complexity. However, a series of trivial examples show that filter methods result in less accurate performance because they ignore the dependencies of features. Although few publications have devoted their attention to reveal the relationship of features by multivariate-based methods, these methods describe relationships among features only by linear methods. While simple linear combination relationship restrict the improvement in performance. In this paper, we used kernel method to discover inherent nonlinear correlations among features as well as between feature and target. Moreover, the number of orthogonal components was determined by kernel Fishers linear discriminant analysis (FLDA) in a self-adaptive manner rather than by manual parameter settings. In order to reveal the effectiveness of our method we performed several experiments and compared the results between our method and other competitive multivariate-based features selectors. In our comparison, we used two classifiers (support vector machine, -nearest neighbor) on two group datasets, namely two-class and multi-class datasets. Experimental results demonstrate that the performance of our method is better than others, especially on three hard-classify datasets, namely Wang''s Breast Cancer, Gordon''s Lung Adenocarcinoma and Pomeroy''s Medulloblastoma.     

A Quantitative Proteomics Design for Systematic Identification of Protease Cleavage Events

We present here a novel proteomics design for systematic identification of protease cleavage events by quantitative N-terminal proteomics, circumventing the need for time-consuming manual validation. We bypass the singleton detection problem of protease-generated neo-N-terminal peptides by introducing differential isotopic proteome labeling such that these substrate reporter peptides are readily distinguished from all other N-terminal peptides. Our approach was validated using the canonical human caspase-3 protease and further applied to mouse cathepsin D and E substrate processing in a mouse dendritic cell proteome, identifying the largest set of protein protease substrates ever reported and gaining novel insight into substrate specificity differences of these cathepsins.Several protocols for proteome-wide identification of protease processing events were recently published. They all follow strategies in which N-terminal peptides, including neo-N-terminal peptides generated by protease action, are enriched from whole proteome digests before identification (e.g. Refs. 1–4). LC-MS/MS analyses of these peptides often yield hundreds of processing events identified in a single experiment (e.g. Refs. 3–5). The N-terminal COFRADIC1 technology developed in our laboratory (6) has been successful in identifying cleavage events of both canonical (e.g. caspases-3 and -7 (7)) and non-canonical proteases (e.g. HtrA2/Omi (8)). Differential stable isotopic labeling in particular, necessary to univocally distinguish genuine neo-N-terminal peptides, allows analyzing control and protease-treated proteomes in a single run. However, this also introduces the most important bottleneck of the technology: verifying whether the peptide envelope of a neo-N-terminal peptide only carries the isotopic label of the protease-treated sample (see Fig. 1A) often had to be done manually for each identified peptide. This “singleton detection problem” can to some extent be automated by software routines such as ProteinProspector (http://prospector.ucsf.edu/prospector/mshome.htm), the MASCOT Distiller Quantitation Toolbox (www.matrixscience.com/distiller.html), and ICPLQuant (9), although these often need specific or proprietary data formats or can only handle MALDI-MS data (9), and researchers still need to individually check correct calling of a neo-N-terminal peptide (10).Open in a separate windowFig. 1.Manual versus automated annotation of protease cleavage events. A, in a typical setup, a heavy (H) labeled proteome is used for protease treatment, and the light (L) labeled proteome serves as a control. Following mixing and N-terminal COFRADIC sorting, neo-N-terminal peptides generated by the added protease are present as singletons, whereas all other N-terminal peptides are present as couples with (light/heavy) ratios around 1 (0 in log₂ scale). B, a mixture of light and heavy labeled proteins (mixed in a 1:1 ratio) is treated with a protease, and as a result, neo-N-terminal peptides generated by the action of the added protease are now present in light/heavy ratios distributed around 1 (0 in log₂ scale) and are clearly distinct from all other N-terminal peptides that come in ratios around 3 (1.58 in log₂ scale). Both types of peptides are readily quantified, circumventing the need for manual validation.To fully overcome this singleton detection problem, here we present and validate a method for highly automated, software-based quantification and annotation of protein processing events on a proteomics scale based on stable isotopic labeling and positional proteomics. We illustrate its strength by generating the largest set of cathepsin D and E substrates hitherto reported. Furthermore, differences in the specificity profiles of these non-canonical proteases are illustrated by the validation of a cleavage event specific for cathepsin E in filamin-A.     

Rationale for More Diverse Inhibitors in Competition with Substrates in HIV-1 Protease

The structural fluctuations of HIV-1 protease in interaction with its substrates versus inhibitors were analyzed using the anisotropic network model. The directions of fluctuations in the most cooperative functional modes differ mainly around the dynamically key regions, i.e., the hinge axes, which appear to be more flexible in substrate complexes. The flexibility of HIV-1 protease is likely optimized for the substrates' turnover, resulting in substrate complexes being dynamic. In contrast, in an inhibitor complex, the inhibitor should bind and lock down to inactivate the active site. Protease and ligands are not independent. Substrates are also more flexible than inhibitors and have the potential to meet the dynamic distributions that are inherent in the protease. This may suggest a rationale and guidelines for designing inhibitors that can better fit the ensemble of binding sites that are dynamically accessible to the protease.     

Catalytic Water Co-Existing with a Product Peptide in the Active Site of HIV-1 Protease Revealed by X-Ray Structure Analysis

Background

It is known that HIV-1 protease is an important target for design of antiviral compounds in the treatment of Acquired Immuno Deficiency Syndrome (AIDS). In this context, understanding the catalytic mechanism of the enzyme is of crucial importance as transition state structure directs inhibitor design. Most mechanistic proposals invoke nucleophilic attack on the scissile peptide bond by a water molecule. But such a water molecule coexisting with any ligand in the active site has not been found so far in the crystal structures.

Principal Findings

We report here the first observation of the coexistence in the active site, of a water molecule WAT1, along with the carboxyl terminal product (Q product) peptide. The product peptide has been generated in situ through cleavage of the full-length substrate. The N-terminal product (P product) has diffused out and is replaced by a set of water molecules while the Q product is still held in the active site through hydrogen bonds. The position of WAT1, which hydrogen bonds to both the catalytic aspartates, is different from when there is no substrate bound in the active site. We propose WAT1 to be the position from where catalytic water attacks the scissile peptide bond. Comparison of structures of HIV-1 protease complexed with the same oligopeptide substrate, but at pH 2.0 and at pH 7.0 shows interesting changes in the conformation and hydrogen bonding interactions from the catalytic aspartates.

Conclusions/Significance

The structure is suggestive of the repositioning, during substrate binding, of the catalytic water for activation and subsequent nucleophilic attack. The structure could be a snap shot of the enzyme active site primed for the next round of catalysis. This structure further suggests that to achieve the goal of designing inhibitors mimicking the transition-state, the hydrogen-bonding pattern between WAT1 and the enzyme should be replicated.     

基于马尔科夫链的胰蛋白酶肽段蛋白酶切位点概率预测

在基于质谱技术的蛋白质鉴定过程中,数据库搜索是主要的方法。漏切位点和酶切规则决定了图谱候选肽段的范围,是数据库搜索算法的重要参数。对于常用的胰蛋白酶切来说,除了局部构象、三维结构、实验条件,以及其它偶然因素会影响赖氨酸K或者精氨酸R后的位点能否被酶切外,该位点附近的其它氨基酸也会影响蛋白水解酶的酶切效果。从质谱图谱中时常会鉴定出包含漏切位点的肽段,因此,预测蛋白质的酶切位点能够为数据库搜索算法提供更为可靠的模型,也能够为了解和分析蛋白质的酶切规律提供依据。本文提出了一种基于马尔科夫(Markov)链的预测方法,能够利用蛋白质的序列信息来预测候选酶切位点的酶切概率,在蛋白酶切过程中,预测肽段的覆盖率可以达到85%以上。     

Quantized Water Access to the HIV-1 Protease Active Site as a Proposed Mechanism for Cooperative Mutations in Drug Affinity

The development of resistance to different drugs remains a major problem for a wide range of infections. In particular, combinations of specific mutations, which individually demonstrate no effect, exhibit significant cooperativity. Here we show that changes to the energy of ligand binding in different resistant HIV-1 proteases are correlated with the creation of water binding sites in the active site. This correlation is conserved across two drugs (ritonavir and lopinavir). We propose that individual mutations induce changes in flap packing that are insufficient to allow water binding but in combination allow access, leading to the observed cooperative resistance.     

一种基于特征筛选的原核生物启动子判别分析方法

启动子识别是研究基因转录调控的重要环节,但目前方法的识别正确率偏低。在深入分析原核启动子特征的基础上,提出了一种基于特征筛选的原核启动子判别分析方法,首先在启动子序列的组成特征、信号特征和结构特征中选取备选特征,为每个特征建立适当的描述模型,并对主要的保守模式采用复合模式模型;再通过模型计算对备选特征进行逐步筛选,优化特征集,将序列表示为组合特征向量;最终利用二次判别分析实现识别。对大肠杆菌和枯草杆菌实际启动子数据进行的刀切法测试验证了方法的有效性和通用性。对于大肠杆菌非编码区(70启动子,识别的平均正确率达到了85.8%,优于其它几种典型识别方法;对于大肠杆菌编码区内部)70启动子和其它几种原核启动子,平均正确率也都超过了80%。方法框架还具有良好的可扩展性,能够方便地容纳新特征,使识别性能不断提高。     

SlimPLS: A Method for Feature Selection in Gene Expression-Based Disease Classification

A major challenge in biomedical studies in recent years has been the classification of gene expression profiles into categories, such as cases and controls. This is done by first training a classifier by using a labeled training set containing labeled samples from the two populations, and then using that classifier to predict the labels of new samples. Such predictions have recently been shown to improve the diagnosis and treatment selection practices for several diseases. This procedure is complicated, however, by the high dimensionality if the data. While microarrays can measure the levels of thousands of genes per sample, case-control microarray studies usually involve no more than several dozen samples. Standard classifiers do not work well in these situations where the number of features (gene expression levels measured in these microarrays) far exceeds the number of samples. Selecting only the features that are most relevant for discriminating between the two categories can help construct better classifiers, in terms of both accuracy and efficiency. In this work we developed a novel method for multivariate feature selection based on the Partial Least Squares algorithm. We compared the method''s variants with common feature selection techniques across a large number of real case-control datasets, using several classifiers. We demonstrate the advantages of the method and the preferable combinations of classifier and feature selection technique.     

Protease Cleavage Sites in HIV-1 gp120 Recognized by Antigen Processing Enzymes Are Conserved and Located at Receptor Binding Sites

The identification of vaccine immunogens able to elicit broadly neutralizing antibodies (bNAbs) is a major goal in HIV vaccine research. Although it has been possible to produce recombinant envelope glycoproteins able to adsorb bNAbs from HIV-positive sera, immunization with these proteins has failed to elicit antibody responses effective against clinical isolates of HIV-1. Thus, the epitopes recognized by bNAbs are present on recombinant proteins, but they are not immunogenic. These results led us to consider the possibility that changes in the pattern of antigen processing might alter the immune response to the envelope glycoprotein to better elicit protective immunity. In these studies, we have defined protease cleavage sites on HIV gp120 recognized by three major human proteases (cathepsins L, S, and D) important for antigen processing and presentation. Remarkably, six of the eight sites identified in gp120 were highly conserved and clustered in regions of the molecule associated with receptor binding and/or the binding of neutralizing antibodies. These results suggested that HIV may have evolved to take advantage of major histocompatibility complex (MHC) class II antigen processing enzymes in order to evade or direct the antiviral immune response.A major goal of HIV vaccine development is the development of immunogens that elicit protective antiviral antibody and cellular immune responses. However, after more than 25 years of research, vaccine immunogens able to elicit protective immunity in humans have yet to be described (11, 31). Although it has been possible to produce recombinant envelope proteins (gp120 and gp140) with many of the features of native virus proteins (e.g., complex glycosylation and the ability to bind CD4, chemokine receptors, and neutralizing antibodies), these antigens have not been able to elicit broadly neutralizing antibodies (bNAbs) or protective immune responses when used as immunogens (11, 32, 43, 50, 56, 74, 79). The fact that recombinant proteins can adsorb virus bNAbs from HIV-1-positive sera (59, 91) indicates that many recombinant envelope proteins are correctly folded but that the epitopes recognized by bNAbs are simply not immunogenic. Over the last decade, several different approaches have been employed to create immunogens able to elicit broadly neutralizing antibodies. These strategies have included efforts to duplicate and/or stabilize the oligomeric structure of HIV envelope proteins (5, 26, 87), the creation of minimal antigenic structures lacking epitopes that conceal important neutralizing sites (27, 46, 70, 89), and prime/boost strategies combining protein immunization with DNA immunization or infection with recombinant viruses in order to stimulate the endogenous synthesis and presentation of HIV immunogens (15, 29, 30, 83). However, none of these approaches has resulted in a clinically significant improvement in antiviral immunity or HIV vaccine efficacy. Efforts to elicit protective cellular immune responses (e.g., cytotoxic lymphocytes) by use of recombinant virus vaccines have likewise been disappointing (10, 61). In fact, such vaccines may have promoted HIV infection rather than inhibiting it (22, 23).In the present study, we describe the first steps in a new approach to reengineering the immunogenicity of HIV envelope proteins in order to improve the potency and specificity of humoral and cellular immune responses. The approach is based on defining the determinants of antigen processing and presentation of HIV envelope glycoproteins. Both humoral and cellular immune responses depend on proteolytic degradation of protein antigens prior to antigen presentation, mediated by professional antigen-presenting cells (APCs) such as macrophages, dendritic cells, and B cells (97). Normally, proteins of intracellular origin are processed by the proteasome, a 14- to 17-subunit protein complex located in the cytosol. Proteins of extracellular origin are processed in lysosomes or late endosomes of APCs. The resulting peptide epitopes are then loaded into major histocompatibility complex (MHC) class I or class II molecules and presented on the surfaces of APCs to CD8 or CD4 T cells. Within the endosomes and lysosomes of APCs, there are cathepsins, acid thiol reductase, and aspartyl endopeptidase. The enzymes perform two activities: degrading endocytosed protein antigens to liberate peptides for MHC class II binding (99) and removing the invariant chain chaperone (6, 94). Although all cathepsins can liberate epitopes from a diverse range of antigens (16), only cathepsins S and L have nonredundant roles in antigen processing in vivo (reviewed by Hsing and Rudensky [45]). Cathepsin L is expressed in thymic cortical epithelial cells but not in B cells or dendritic cells, while cathepsin S is found in all three types of APCs. Unlike cathepsins L and S, which are cysteine proteases and active at neutral pH, cathepsin D is an aspartic protease, is active at acidic pH, and participates in proteolysis and antigen presentation in connection with MHC class I and class II antigen presentation pathways established for CD4 and CD8 T cells. In considering the use of envelope proteins as potential vaccines, the route of immunization, formulation (e.g., adjuvants), protein folding, disulfide bonding, and glycosylation pattern all determine which peptides are available for MHC-restricted presentation.Previous studies provided evidence that gp120 was sensitive to digestion by cathepsins B, D, and L, but the specific cleavage sites were not defined (18). In the present study, we (i) describe the locations of eight protease cleavage sites on HIV-1 gp120 recognized by cathepsins L, S, and D, involved in antigen processing; (ii) determine the extent to which they are conserved; and (iii) evaluate the effect of cathepsin cleavage on the binding of gp120 to CD4-IgG and neutralizing antibodies. The results obtained provide new insights into the basis of envelope immunogenicity that may prove to be useful in the development of HIV vaccine antigens.