ExTopoDB: a database of experimentally derived topological models of transmembrane proteins

Semi-supervised learning of Hidden Markov Models for biological sequence analysis

MOTIVATION

Hidden Markov Models (HMMs) are probabilistic models widely used in applications in computational sequence analysis. HMMs are basically unsupervised models. However, in the most important applications, they are trained in a supervised manner. Training examples accompanied by labels corresponding to different classes are given as input and the set of parameters that maximize the joint probability of sequences and labels is estimated. A main problem with this approach is that, in the majority of the cases, labels are hard to find and thus the amount of training data is limited. On the other hand, there are plenty of unclassified (unlabeled) sequences deposited in the public databases that could potentially contribute to the training procedure. This approach is called semi-supervised learning and could be very helpful in many applications.

RESULTS

We propose here, a method for semi-supervised learning of HMMs that can incorporate labeled, unlabeled and partially labeled data in a straightforward manner. The algorithm is based on a variant of the Expectation-Maximization (EM) algorithm, where the missing labels of the unlabeled or partially labeled data are considered as the missing data. We apply the algorithm to several biological problems, namely, for the prediction of transmembrane protein topology for alpha-helical and beta-barrel membrane proteins and for the prediction of archaeal signal peptides. The results are very promising, since the algorithms presented here can significantly improve the prediction performance of even the top-scoring classifiers.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Partial proteolysis improves the identification of the extracellular segments of transmembrane proteins by surface biotinylation

Transmembrane proteins (TMP) play a crucial role in several physiological processes. Despite their importance and diversity, only a few TMP structures have been determined by high-resolution protein structure characterization methods so far. Due to the low number of determined TMP structures, the parallel development of various bioinformatics and experimental methods was necessary for their topological characterization. The combination of these methods is a powerful approach in the determination of TMP topology as in the Constrained Consensus TOPology prediction. To support the prediction, we previously developed a high-throughput topology characterization method based on primary amino group-labelling that is still limited in identifying all TMPs and their extracellular segments on the surface of a particular cell type. In order to generate more topology information, a new step, a partial proteolysis of the cell surface has been introduced to our method. This step results in new primary amino groups in the proteins that can be biotinylated with a membrane-impermeable agent while the cells still remain intact. Pre-digestion also promotes the emergence of modified peptides that are more suitable for MS/MS analysis. The modified sites can be utilized as extracellular constraints in topology predictions and may contribute to the refined topology of these proteins.

Predicting Alpha Helical Transmembrane Proteins Using HMMs

Alpha helical transmembrane (TM) proteins constitute an important structural class of membrane proteins involved in a wide variety of cellular functions. The prediction of their transmembrane topology, as well as their discrimination in newly sequenced genomes, is of great importance for the elucidation of their structure and function. Several methods have been applied for the prediction of the transmembrane segments and the topology of alpha helical transmembrane proteins utilizing different algorithmic techniques. Hidden Markov Models (HMMs) have been efficiently used in the development of several computational methods used for this task. In this chapter we give a brief review of different available prediction methods for alpha helical transmembrane proteins pointing out sequence and structural features that should be incorporated in a prediction method. We then describe the procedure of the design and development of a Hidden Markov Model capable of predicting the transmembrane alpha helices in proteins and discriminating them from globular proteins.

Bioinformatics Analysis of Protein Secretion in Plants

In sessile plants, the dynamic protein secretion pathways orchestrate the cellular responses to internal signals and external environmental changes in almost every aspect of plant developmental events. The cohort of plant proteins, secreted from the plant cells into the extracellular matrix, has been annotated as plant secretome. Therefore, the identification and characterization of secreted proteins will discover novel secretory potentials and establish the functional connection between cellular protein secretion and plant physiological phenomena. Noteworthy, an increasing number of bioinformatics databases and tools have been developed for computational predictions on either secreted proteins or secretory pathways. This chapter summarizes current accessible databases and tools for protein secretion analysis in Arabidopsis thaliana and higher plants, and provides feasible methodologies for bioinformatics analysis of secretome studies for the plant research community.

Key challenges for the creation and maintenance of specialist protein resources

As the volume of data relating to proteins increases, researchers rely more and more on the analysis of published data, thus increasing the importance of good access to these data that vary from the supplemental material of individual articles, all the way to major reference databases with professional staff and long-term funding. Specialist protein resources fill an important middle ground, providing interactive web interfaces to their databases for a focused topic or family of proteins, using specialized approaches that are not feasible in the major reference databases. Many are labors of love, run by a single lab with little or no dedicated funding and there are many challenges to building and maintaining them. This perspective arose from a meeting of several specialist protein resources and major reference databases held at the Wellcome Trust Genome Campus (Cambridge, UK) on August 11 and 12, 2014. During this meeting some common key challenges involved in creating and maintaining such resources were discussed, along with various approaches to address them. In laying out these challenges, we aim to inform users about how these issues impact our resources and illustrate ways in which our working together could enhance their accuracy, currency, and overall value.

Computational modeling of membrane proteins

The determination of membrane protein (MP) structures has always trailed that of soluble proteins due to difficulties in their overexpression, reconstitution into membrane mimetics, and subsequent structure determination. The percentage of MP structures in the protein databank (PDB) has been at a constant 1-2% for the last decade. In contrast, over half of all drugs target MPs, only highlighting how little we understand about drug-specific effects in the human body. To reduce this gap, researchers have attempted to predict structural features of MPs even before the first structure was experimentally elucidated. In this review, we present current computational methods to predict MP structure, starting with secondary structure prediction, prediction of trans-membrane spans, and topology. Even though these methods generate reliable predictions, challenges such as predicting kinks or precise beginnings and ends of secondary structure elements are still waiting to be addressed. We describe recent developments in the prediction of 3D structures of both α-helical MPs as well as β-barrels using comparative modeling techniques, de novo methods, and molecular dynamics (MD) simulations. The increase of MP structures has (1) facilitated comparative modeling due to availability of more and better templates, and (2) improved the statistics for knowledge-based scoring functions. Moreover, de novo methods have benefited from the use of correlated mutations as restraints. Finally, we outline current advances that will likely shape the field in the forthcoming decade.

mRNA-programmed translation pauses in the targeting of E. coli membrane proteins

In all living organisms, ribosomes translating membrane proteins are targeted to membrane translocons early in translation, by the ubiquitous signal recognition particle (SRP) system. In eukaryotes, the SRP Alu domain arrests translation elongation of membrane proteins until targeting is complete. Curiously, however, the Alu domain is lacking in most eubacteria. In this study, by analyzing genome-wide data on translation rates, we identified a potential compensatory mechanism in E. coli that serves to slow down the translation during membrane protein targeting. The underlying mechanism is likely programmed into the coding sequence, where Shine-Dalgarno-like elements trigger elongation pauses at strategic positions during the early stages of translation. We provide experimental evidence that slow translation during targeting and improves membrane protein production fidelity, as it correlates with better folding of overexpressed membrane proteins. Thus, slow elongation is important for membrane protein targeting in E. coli, which utilizes mechanisms different from the eukaryotic one to control the translation speed.

HMMpTM: improving transmembrane protein topology prediction using phosphorylation and glycosylation site prediction

During the last two decades a large number of computational methods have been developed for predicting transmembrane protein topology. Current predictors rely on topogenic signals in the protein sequence, such as the distribution of positively charged residues in extra-membrane loops and the existence of N-terminal signals. However, phosphorylation and glycosylation are post-translational modifications (PTMs) that occur in a compartment-specific manner and therefore the presence of a phosphorylation or glycosylation site in a transmembrane protein provides topological information. We examine the combination of phosphorylation and glycosylation site prediction with transmembrane protein topology prediction. We report the development of a Hidden Markov Model based method, capable of predicting the topology of transmembrane proteins and the existence of kinase specific phosphorylation and N/O-linked glycosylation sites along the protein sequence. Our method integrates a novel feature in transmembrane protein topology prediction, which results in improved performance for topology prediction and reliable prediction of phosphorylation and glycosylation sites. The method is freely available at http://bioinformatics.biol.uoa.gr/HMMpTM.

HMPAS: Human Membrane Protein Analysis System

Background

Membrane proteins perform essential roles in diverse cellular functions and are regarded as major pharmaceutical targets. The significance of membrane proteins has led to the developing dozens of resources related with membrane proteins. However, most of these resources are built for specific well-known membrane protein groups, making it difficult to find common and specific features of various membrane protein groups.

Methods

We collected human membrane proteins from the dispersed resources and predicted novel membrane protein candidates by using ortholog information and our membrane protein classifiers. The membrane proteins were classified according to the type of interaction with the membrane, subcellular localization, and molecular function. We also made new feature dataset to characterize the membrane proteins in various aspects including membrane protein topology, domain, biological process, disease, and drug. Moreover, protein structure and ICD-10-CM based integrated disease and drug information was newly included. To analyze the comprehensive information of membrane proteins, we implemented analysis tools to identify novel sequence and functional features of the classified membrane protein groups and to extract features from protein sequences.

Results

We constructed HMPAS with 28,509 collected known membrane proteins and 8,076 newly predicted candidates. This system provides integrated information of human membrane proteins individually and in groups organized by 45 subcellular locations and 1,401 molecular functions. As a case study, we identified associations between the membrane proteins and diseases and present that membrane proteins are promising targets for diseases related with nervous system and circulatory system. A web-based interface of this system was constructed to facilitate researchers not only to retrieve organized information of individual proteins but also to use the tools to analyze the membrane proteins.

Conclusions

HMPAS provides comprehensive information about human membrane proteins including specific features of certain membrane protein groups. In this system, user can acquire the information of individual proteins and specified groups focused on their conserved sequence features, involved cellular processes, and diseases. HMPAS may contribute as a valuable resource for the inference of novel cellular mechanisms and pharmaceutical targets associated with the human membrane proteins. HMPAS is freely available at http://fcode.kaist.ac.kr/hmpas.

Site-specific epitope tagging of G protein-coupled receptors by bioorthogonal modification of a genetically encoded unnatural amino acid

We developed a general strategy for labeling expressed membrane proteins with a peptide epitope tag and detecting the tagged proteins in native cellular membranes. First, we genetically encoded the unnatural amino acid p-azido-L-phenylalanine (azF) at various specific sites in a G protein-coupled receptor (GPCR), C-C chemokine receptor 5 (CCR5). The reactive azido moiety facilitates Staudinger ligation to a triarylphosphine-conjugated FLAG peptide. We then developed a whole-cell-based enzyme-linked immunosorbent assay approach to detect the modified azF-CCR5 using anti-FLAG mAb. We optimized conditions to achieve labeling and detection of low-abundance GPCRs in live cells. We also performed an accessibility screen to identify azF positions on CCR5 amenable to labeling. Finally, we demonstrate a preparative strategy for obtaining pure bioorthogonally modified GPCRs suitable for single-molecule detection fluorescence experiments. This peptide epitope tagging strategy, which employs genetic encoding and bioorthogonal labeling of azF in live cells, should be useful for studying biogenesis of polytopic membrane proteins and GPCR signaling mechanisms.

Efficient unfolding pattern recognition in single molecule force spectroscopy data

Background

Single-molecule force spectroscopy (SMFS) is a technique that measures the force necessary to unfold a protein. SMFS experiments generate Force-Distance (F-D) curves. A statistical analysis of a set of F-D curves reveals different unfolding pathways. Information on protein structure, conformation, functional states, and inter- and intra-molecular interactions can be derived.

Results

In the present work, we propose a pattern recognition algorithm and apply our algorithm to datasets from SMFS experiments on the membrane protein bacterioRhodopsin (bR). We discuss the unfolding pathways found in bR, which are characterised by main peaks and side peaks. A main peak is the result of the pairwise unfolding of the transmembrane helices. In contrast, a side peak is an unfolding event in the alpha-helix or other secondary structural element. The algorithm is capable of detecting side peaks along with main peaks.

Therefore, we can detect the individual unfolding pathway as the sequence of events labeled with their occurrences and co-occurrences special to bR's unfolding pathway. We find that side peaks do not co-occur with one another in curves as frequently as main peaks do, which may imply a synergistic effect occurring between helices. While main peaks co-occur as pairs in at least 50% of curves, the side peaks co-occur with one another in less than 10% of curves. Moreover, the algorithm runtime scales well as the dataset size increases.

Conclusions

Our algorithm satisfies the requirements of an automated methodology that combines high accuracy with efficiency in analyzing SMFS datasets. The algorithm tackles the force spectroscopy analysis bottleneck leading to more consistent and reproducible results.