Essentiality, protein-protein interactions and evolutionary properties are key predictors for identifying cancer-associated genes using machine learning

The distinctive nature of cancer as a disease prompts an exploration of the special characteristics the genes implicated in cancer exhibit. The identification of cancer-associated genes and their characteristics is crucial to further our understanding of this disease and enhanced likelihood of therapeutic drug targets success. However, the rate at which cancer genes are being identified experimentally is slow. Applying predictive analysis techniques, through the building of accurate machine learning models, is potentially a useful approach in enhancing the identification rate of these genes and their characteristics. Here, we investigated gene essentiality scores and found that they tend to be higher for cancer-associated genes compared to other protein-coding human genes. We built a dataset of extended gene properties linked to essentiality and used it to train a machine-learning model; this model reached 89% accuracy and > 0.85 for the Area Under Curve (AUC). The model showed that essentiality, evolutionary-related properties, and properties arising from protein-protein interaction networks are particularly effective in predicting cancer-associated genes. We were able to use the model to identify potential candidate genes that have not been previously linked to cancer. Prioritising genes that score highly by our methods could aid scientists in their cancer genes research.

Predicting congenital renal tract malformation genes using machine learning

Congenital renal tract malformations (RTMs) are the major cause of severe kidney failure in children. Studies to date have identified defined genetic causes for only a minority of human RTMs. While some RTMs may be caused by poorly defined environmental perturbations affecting organogenesis, it is likely that numerous causative genetic variants have yet to be identified. Unfortunately, the speed of discovering further genetic causes for RTMs is limited by challenges in prioritising candidate genes harbouring sequence variants. Here, we exploited the computer-based artificial intelligence methodology of supervised machine learning to identify genes with a high probability of being involved in renal development. These genes, when mutated, are promising candidates for causing RTMs. With this methodology, the machine learning classifier determines which attributes are common to renal development genes and identifies genes possessing these attributes. Here we report the validation of an RTM gene classifier and provide predictions of the RTM association status for all protein-coding genes in the mouse genome. Overall, our predictions, whilst not definitive, can inform the prioritisation of genes when evaluating patient sequence data for genetic diagnosis. This knowledge of renal developmental genes will accelerate the processes of reaching a genetic diagnosis for patients born with RTMs.

Viral Involvement in Alzheimer's Disease

Alzheimer's disease (AD) is characterized by the presence of β-amyloid plaques (Aβ) and neurofibrillary tangles (NFTs) in the brain. The prevalence of the disease is increasing and is expected to reach 141 million cases by 2050. Despite the risk factors associated with the disease, there is no known causative agent for AD. Clinical trials with many drugs have failed over the years, and no therapeutic has been approved for AD. There is increasing evidence that pathogens are found in the brains of AD patients and controls, such as human herpes simplex virus-1 (HSV-1). Given the lack of a human model, the route for pathogen entry into the brain remains open for scrutiny and may include entry via a disturbed blood-brain barrier or the olfactory nasal route. Many factors can contribute to the pathogenicity of HSV-1, such as the ability of HSV-1 to remain latent, tau protein phosphorylation, increased accumulation of Aβ invivo and in vitro, and repeated cycle of reactivation if immunocompromised. Intriguingly, valacyclovir, a widely used drug for the treatment of HSV-1 and HSV-2 infection, has shown patient improvement in cognition compared to controls in AD clinical studies. We discuss the potential role of HSV-1 in AD pathogenesis and argue for further studies to investigate this relationship.

Amyloid Oligomers: A Joint Experimental/Computational Perspective on Alzheimer's Disease, Parkinson's Disease, Type II Diabetes, and Amyotrophic Lateral Sclerosis

Protein misfolding and aggregation is observed in many amyloidogenic diseases affecting either the central nervous system or a variety of peripheral tissues. Structural and dynamic characterization of all species along the pathways from monomers to fibrils is challenging by experimental and computational means because they involve intrinsically disordered proteins in most diseases. Yet understanding how amyloid species become toxic is the challenge in developing a treatment for these diseases. Here we review what computer, in vitro, in vivo, and pharmacological experiments tell us about the accumulation and deposition of the oligomers of the (Aβ, tau), α-synuclein, IAPP, and superoxide dismutase 1 proteins, which have been the mainstream concept underlying Alzheimer's disease (AD), Parkinson's disease (PD), type II diabetes (T2D), and amyotrophic lateral sclerosis (ALS) research, respectively, for many years.

Raman Spectroscopy to Monitor Post-Translational Modifications and Degradation in Monoclonal Antibody Therapeutics

Monoclonal antibodies (mAbs) represent a rapidly expanding market for biotherapeutics. Structural changes in the mAb can lead to unwanted immunogenicity, reduced efficacy, and loss of material during production. The pharmaceutical sector requires new protein characterization tools that are fast, applicable in situ and to the manufacturing process. Raman has been highlighted as a technique to suit this application as it is information-rich, minimally invasive, insensitive to water background and requires little to no sample preparation. This study investigates the applicability of Raman to detect Post-Translational Modifications (PTMs) and degradation seen in mAbs. IgG4 molecules have been incubated under a range of conditions known to result in degradation of the therapeutic including varied pH, temperature, agitation, photo, and chemical stresses. Aggregation was measured using size-exclusion chromatography, and PTM levels were calculated using peptide mapping. By combining principal component analysis (PCA) with Raman spectroscopy and circular dichroism (CD) spectroscopy structural analysis we were able to separate proteins based on PTMs and degradation. Furthermore, by identifying key bands that lead to the PCA separation we could correlate spectral peaks to specific PTMs. In particular, we have identified a peak which exhibits a shift in samples with higher levels of Trp oxidation. Through separation of IgG4 aggregates, by size, we have shown a linear correlation between peak wavenumbers of specific functional groups and the amount of aggregate present. We therefore demonstrate the capability for Raman spectroscopy to be used as an analytical tool to measure degradation and PTMs in-line with therapeutic production.

Deconvolution of conformational exchange from Raman spectra of aqueous RNA nucleosides

Ribonucleic acids (RNAs) are key to the central dogma of molecular biology. While Raman spectroscopy holds great potential for studying RNA conformational dynamics, current computational Raman prediction and assignment methods are limited in terms of system size and inclusion of conformational exchange. Here, a framework is presented that predicts Raman spectra using mixtures of sub-spectra corresponding to major conformers calculated using classical and ab initio molecular dynamics. Experimental optimization allowed purines and pyrimidines to be characterized as predominantly syn and anti, respectively, and ribose into exchange between equivalent south and north populations. These measurements are in excellent agreement with Raman spectroscopy of ribonucleosides, and previous experimental and computational results. This framework provides a measure of ribonucleoside solution populations and conformational exchange in RNA subunits. It complements other experimental techniques and could be extended to other molecules, such as proteins and carbohydrates, enabling biological insights and providing a new analytical tool.

Quantification of protein glycation using vibrational spectroscopy

FTIR-ATR and Raman spectroscopy can distinguish between glycated and non-glycated proteins.

The Essentiality Status of Mouse Duplicate Gene Pairs Correlates with Developmental Co-Expression Patterns

During the evolution of multicellular eukaryotes, gene duplication occurs frequently to generate new genes and/or functions. A duplicated gene may have a similar function to its ancestral gene. Therefore, it may be expected that duplicated genes are less likely to be critical for the survival of an organism, since there are multiple copies of the gene rendering each individual copy redundant. In this study, we explored the developmental expression patterns of duplicate gene pairs and the relationship between development co-expression and phenotypes resulting from the knockout of duplicate genes in the mouse. We define genes that generate lethal phenotypes in single gene knockout experiments as essential genes. We found that duplicate gene pairs comprised of two essential genes tend to be expressed at different stages of development, compared to duplicate gene pairs with at least one non-essential member, showing that the timing of developmental expression affects the ability of one paralogue to compensate for the loss of the other. Gene essentiality, developmental expression and gene duplication are thus closely linked.

Identifying mouse developmental essential genes using machine learning

The genes that are required for organismal survival are annotated as ‘essential genes’. Identifying all the essential genes of an animal species can reveal critical functions that are needed during the development of the organism. To inform studies on mouse development, we developed a supervised machine learning classifier based on phenotype data from mouse knockout experiments. We used this classifier to predict the essentiality of mouse genes lacking experimental data. Validation of our predictions against a blind test set of recent mouse knockout experimental data indicated a high level of accuracy (>80%). We also validated our predictions for other mouse mutagenesis methodologies, demonstrating that the predictions are accurate for lethal phenotypes isolated in random chemical mutagenesis screens and embryonic stem cell screens. The biological functions that are enriched in essential and non-essential genes have been identified, showing that essential genes tend to encode intracellular proteins that interact with nucleic acids. The genome distribution of predicted essential and non-essential genes was analysed, demonstrating that the density of essential genes varies throughout the genome. A comparison with human essential and non-essential genes was performed, revealing conservation between human and mouse gene essentiality status. Our genome-wide predictions of mouse essential genes will be of value for the planning of mouse knockout experiments and phenotyping assays, for understanding the functional processes required during mouse development, and for the prioritisation of disease candidate genes identified in human genome and exome sequence datasets.

Summary: Here, we used computer-based machine learning methodology to predict which genes in the mouse genome are essential for development, and present a database of mouse essential and non-essential genes.

Effect of IAPP on the proteome of cultured Rin-5F cells

Background

Islet amyloid polypeptide (IAPP) or amylin deposits can be found in the islets of type 2 diabetes patients. The peptide is suggested to be involved in the etiology of the disease through formation of amyloid deposits and destruction of β islet cells, though the underlying molecular events leading from IAPP deposition to β cell death are still largely unknown.

Results

We used OFFGEL™ proteomics to study how IAPP exposure affects the proteome of rat pancreatic insulinoma Rin-5F cells. The OFFGEL™ methodology is highly effective at generating quantitative data on hundreds of proteins affected by IAPP, with its accuracy confirmed by In Cell Western and Quantitative Real Time PCR results. Combining data on individual proteins identifies pathways and protein complexes affected by IAPP. IAPP disrupts protein synthesis and degradation, and induces oxidative stress. It causes decreases in protein transport and localization. IAPP disrupts the regulation of ubiquitin-dependent protein degradation and increases catabolic processes. IAPP causes decreases in protein transport and localization, and affects the cytoskeleton, DNA repair and oxidative stress.

Conclusions

Results are consistent with a model where IAPP aggregates overwhelm the ability of a cell to degrade proteins via the ubiquitin system. Ultimately this leads to apoptosis. IAPP aggregates may be also toxic to the cell by causing oxidative stress, leading to DNA damage or by decreasing protein transport. The reversal of any of these effects, perhaps by targeting proteins which alter in response to IAPP, may be beneficial for type II diabetes.

Electronic supplementary material

The online version of this article (10.1186/s12858-018-0099-3) contains supplementary material, which is available to authorized users.

Positive Feedback Loops in Alzheimer's Disease: The Alzheimer's Feedback Hypothesis

The dominant model for Alzheimer's disease (AD) is the amyloid cascade hypothesis, in which the accumulation of excess amyloid-β (Aβ) leads to inflammation, excess glutamate and intracellular calcium, oxidative stress, tau hyperphosphorylation and tangle formation, neuronal loss, and ultimately dementia. In a cascade, AD proceeds in a unidirectional fashion, with events only affecting downstream processes. Compelling evidence now exists for the presence of positive feedback loops in AD, however, involving oxidative stress, inflammation, glutamate, calcium, and tau. The pathological state of AD is thus a system of positive feedback loops, leading to amplification of the initial perturbation, rather than a linear cascade. Drugs may therefore be effective by targeting numerous points within the loops, rather than concentrating on upstream processes. Anti-inflammatories and anti-oxidants may be especially valuable, since these processes are involved in many loops and hence would affect numerous processes in AD.

Why Is Research on Amyloid-β Failing to Give New Drugs for Alzheimer's Disease?

The two hallmarks of Alzheimer's disease (AD) are the presence of neurofibrillary tangles (NFT) made of aggregates of the hyperphosphorylated tau protein and of amyloid plaques composed of amyloid-β (Aβ) peptides, primarily Aβ1-40 and Aβ1-42. Targeting the production, aggregation, and toxicity of Aβ with small molecule drugs or antibodies is an active area of AD research due to the general acceptance of the amyloid cascade hypothesis, but thus far all drugs targeting Aβ have failed. From a review of the recent literature and our own experience based on in vitro, in silico, and in vivo studies, we present some reasons to explain this repetitive failure.

Properties of genes essential for mouse development

Essential genes are those that are critical for life. In the specific case of the mouse, they are the set of genes whose deletion means that a mouse is unable to survive after birth. As such, they are the key minimal set of genes needed for all the steps of development to produce an organism capable of life ex utero. We explored a wide range of sequence and functional features to characterise essential (lethal) and non-essential (viable) genes in mice. Experimental data curated manually identified 1301 essential genes and 3451 viable genes. Very many sequence features show highly significant differences between essential and viable mouse genes. Essential genes generally encode complex proteins, with multiple domains and many introns. These genes tend to be: long, highly expressed, old and evolutionarily conserved. These genes tend to encode ligases, transferases, phosphorylated proteins, intracellular proteins, nuclear proteins, and hubs in protein-protein interaction networks. They are involved with regulating protein-protein interactions, gene expression and metabolic processes, cell morphogenesis, cell division, cell proliferation, DNA replication, cell differentiation, DNA repair and transcription, cell differentiation and embryonic development. Viable genes tend to encode: membrane proteins or secreted proteins, and are associated with functions such as cellular communication, apoptosis, behaviour and immune response, as well as housekeeping and tissue specific functions. Viable genes are linked to transport, ion channels, signal transduction, calcium binding and lipid binding, consistent with their location in membranes and involvement with cell-cell communication. From the analysis of the composite features of essential and viable genes, we conclude that essential genes tend to be required for intracellular functions, and viable genes tend to be involved with extracellular functions and cell-cell communication. Knowledge of the features that are over-represented in essential genes allows for a deeper understanding of the functions and processes implemented during mammalian development.

Secondary Structure and Glycosylation of Mucus Glycoproteins by Raman Spectroscopies

The major structural components of protective mucus hydrogels on mucosal surfaces are the secreted polymeric gel-forming mucins. The very high molecular weight and extensive O-glycosylation of gel-forming mucins, which are key to their viscoelastic properties, create problems when studying mucins using conventional biochemical/structural techniques. Thus, key structural information, such as the secondary structure of the various mucin subdomains, and glycosylation patterns along individual molecules, remains to be elucidated. Here, we utilized Raman spectroscopy, Raman optical activity (ROA), circular dichroism (CD), and tip-enhanced Raman spectroscopy (TERS) to study the structure of the secreted polymeric gel-forming mucin MUC5B. ROA indicated that the protein backbone of MUC5B is dominated by unordered conformation, which was found to originate from the heavily glycosylated central mucin domain by isolation of MUC5B O-glycan-rich regions. In sharp contrast, recombinant proteins of the N-terminal region of MUC5B (D1-D2-D′-D3 domains, NT5B), C-terminal region of MUC5B (D4-B-C-CK domains, CT5B) and the Cys-domain (within the central mucin domain of MUC5B) were found to be dominated by the β-sheet. Using these findings, we employed TERS, which combines the chemical specificity of Raman spectroscopy with the spatial resolution of atomic force microscopy to study the secondary structure along 90 nm of an individual MUC5B molecule. Interestingly, the molecule was found to contain a large amount of α-helix/unordered structures and many signatures of glycosylation, pointing to a highly O-glycosylated region on the mucin.

Combined experimental and simulation studies suggest a revised mode of action of the anti-Alzheimer disease drug NQ-Trp

Inhibition of the aggregation of the monomeric peptide β-amyloid (Aβ) into oligomers is a widely studied therapeutic approach in Alzheimer's disease (AD). Many small molecules have been reported to work in this way, including 1,4-naphthoquinon-2-yl-L-tryptophan (NQ-Trp). NQ-Trp has been reported to inhibit aggregation, to rescue cells from Aβ toxicity, and showed complete phenotypic recovery in an in vivo AD model. In this work we investigated its molecular mechanism by using a combined approach of experimental and theoretical studies, and obtained converging results. NQ-Trp is a relatively weak inhibitor and the fluorescence data obtained by employing the fluorophore widely used to monitor aggregation into fibrils can be misinterpreted due to the inner filter effect. Simulations and NMR experiments showed that NQ-Trp has no specific "binding site"-type interaction with mono- and dimeric Aβ, which could explain its low inhibitory efficiency. This suggests that the reported anti-AD activity of NQ-Trp-type molecules in in vivo models has to involve another mechanism. This study has revealed the potential pitfalls in the development of aggregation inhibitors for amyloidogenic peptides, which are of general interest for all the molecules studied in the context of inhibiting the formation of toxic aggregates.

Properties of protein drug target classes

Accurate identification of drug targets is a crucial part of any drug development program. We mined the human proteome to discover properties of proteins that may be important in determining their suitability for pharmaceutical modulation. Data was gathered concerning each protein’s sequence, post-translational modifications, secondary structure, germline variants, expression profile and drug target status. The data was then analysed to determine features for which the target and non-target proteins had significantly different values. This analysis was repeated for subsets of the proteome consisting of all G-protein coupled receptors, ion channels, kinases and proteases, as well as proteins that are implicated in cancer. Machine learning was used to quantify the proteins in each dataset in terms of their potential to serve as a drug target. This was accomplished by first inducing a random forest that could distinguish between its targets and non-targets, and then using the random forest to quantify the drug target likeness of the non-targets. The properties that can best differentiate targets from non-targets were primarily those that are directly related to a protein’s sequence (e.g. secondary structure). Germline variants, expression levels and interactions between proteins had minimal discriminative power. Overall, the best indicators of drug target likeness were found to be the proteins’ hydrophobicities, in vivo half-lives, propensity for being membrane bound and the fraction of non-polar amino acids in their sequences. In terms of predicting potential targets, datasets of proteases, ion channels and cancer proteins were able to induce random forests that were highly capable of distinguishing between targets and non-targets. The non-target proteins predicted to be targets by these random forests comprise the set of the most suitable potential future drug targets, and should therefore be prioritised when building a drug development programme.

Molecular structure of the NQTrp inhibitor with the Alzheimer Aβ1-28 monomer

The self-assembly of the amyloid-β (Aβ) peptide of various amino acid lengths into senile plaques is one hallmark of Alzheimer's disease pathology. In the past decade, many small molecules, including NQTrp, have been identified to reduce aggregation and toxicity. However, due to the heterogeneity of the conformational ensemble of Aβ with drugs, we lack detailed structures of the transient complexes. Following our previous simulation of the monomer of Aβ1-28, here we characterize the equilibrium ensemble of the Aβ1-28 monomer with NQTrp by means of extensive atomistic replica exchange molecular dynamics simulations using a force field known to fold diverse proteins correctly. While the secondary structure content and the intrinsic disorder of the whole peptides are very similar and the lifetimes of the salt-bridges remain constant, the population of β-hairpin is reduced by a factor of 1.5 and the population of α-helix in the region 17-24 is increased by a factor of two upon NQTrp binding. These two factors, which impact the free energy barrier for nucleation, provide a first explanation for the reported reduced Aβ1-40/1-42 aggregation kinetics in the presence of NQTrp. Backbone and side-chain interactions of Aβ with NQTrp may also inhibit Aβ-Aβ contacts. The fraction of free Aβ1-28 monomer is, however, on the order of 20-25% at 17.5 mM, and this shows that the affinity of NQTrp is low and hence its inhibitory activity is not very strong. This inhibitor can be improved to reduce the formation of dimer, a critical step in aggregation and toxicity.

Maximising the size of non-redundant protein datasets using graph theory

Analysis of protein data sets often requires prior removal of redundancy, so that data is not biased by containing similar proteins. This is usually achieved by pairwise comparison of sequences, followed by purging so that no two pairs have similarities above a chosen threshold. From a starting set, such as the PDB or a genome, one should remove as few sequences as possible, to give the largest possible non-redundant set for subsequent analysis. Protein redundancy can be represented as a graph, with proteins as nodes connected by undirected edges, if they have a pairwise similarity above the chosen threshold. The problem is then equivalent to finding the maximum independent set (MIS), where as few nodes are removed as possible to remove all edges. We tested seven MIS algorithms, three of which are new. We applied the methods to the PDB, subsets of the PDB, various genomes and the BHOLSIB benchmark datasets. For PDB subsets of up to 1000 proteins, we could compare to the exact MIS, found by the Cliquer algorithm. The best algorithm was the new method, Leaf. This works by adding clique members that have no edges to nodes outside the clique to the MIS, starting with the smallest cliques. For PDB subsets of up to 1000 members, it usually finds the MIS and is fast enough to apply to data sets of tens of thousands of proteins. Leaf gives sets that are around 10% larger than the commonly used PISCES algorithm, that are of identical quality. We therefore suggest that Leaf should be the method of choice for generating non-redundant protein data sets, though it is ineffective on dense graphs, such as the BHOLSIB benchmarks. The Leaf algorithm is available at: https://github.com/SimonCB765/Leaf, and sets from genomes and the PDB are available at: http://www.bioinf.manchester.ac.uk/leaf/.

The N-methylated peptide SEN304 powerfully inhibits Aβ(1-42) toxicity by perturbing oligomer formation

Oligomeric forms of β-amyloid (Aβ) have potent neurotoxic activity and are the primary cause of neuronal injury and cell death in Alzheimer's disease (AD). Compounds that perturb oligomer formation or structure may therefore be therapeutic for AD. We previously reported that d-[(chGly)-(Tyr)-(chGly)-(chGly)-(mLeu)]-NH(2) (SEN304) is able to inhibit Aβ aggregation and toxicity, shown primarily by thioflavin T fluorescence and MTT (Kokkoni, N. et al. (2006) N-Methylated peptide inhibitors of β-amyloid aggregation and toxicity. Optimisation of inhibitor structure. Biochemistry 45, 9906-9918). Here we extensively characterize how SEN304 affects Aβ(1-42) aggregation and toxicity, using biophysical assays (thioflavin T, circular dichroism, SDS-PAGE, size exclusion chromatography, surface plasmon resonance, traveling wave ion mobility mass spectrometry, electron microscopy, ELISA), toxicity assays in cell culture (MTT and lactate dehydrogenase in human SH-SHY5Y cells, mouse neuronal cell death and synaptophysin) and long-term potentiation in a rat hippocampal brain slice. These data, with dose response curves, show that SEN304 is a powerful inhibitor of Aβ(1-42) toxicity, particularly effective at preventing Aβ inhibition of long-term potentiation. It can bind directly to Aβ(1-42), delay β-sheet formation and promote aggregation of toxic oligomers into a nontoxic form, with a different morphology that cannot bind thioflavin T. SEN304 appears to work by inducing aggregation, and hence removal, of Aβ oligomers. It is therefore a promising lead compound for Alzheimer's disease.

Determination of protein fold class from Raman or Raman optical activity spectra using random forests

Knowledge of the fold class of a protein is valuable because fold class gives an indication of protein function and evolution. Fold class can be accurately determined from a crystal structure or NMR structure, though these methods are expensive, time-consuming, and inapplicable to all proteins. In contrast, vibrational spectra [infra-red, Raman, or Raman optical activity (ROA)] are rapidly obtained for proteins under wide range of biological molecules under diverse experimental and physiological conditions. Here, we show that the fold class of a protein can be determined from Raman or ROA spectra by converting a spectrum into data of 10 cm(-1) bin widths and applying the random forest machine learning algorithm. Spectral data from 605 and 1785 cm(-1) were analyzed, as well as the amide I, II, and III regions in isolation and in combination. ROA amide II and III data gave the best performance, with 33 of 44 proteins assigned to one of the correct four top-level structural classification of proteins (SCOP) fold class (all α, all β, α and β, and disordered). The method also shows which spectral regions are most valuable in assigning fold class.

Properties and identification of antibiotic drug targets

BACKGROUND

We analysed 48 non-redundant antibiotic target proteins from all bacteria, 22 antibiotic target proteins from E. coli only and 4243 non-drug targets from E. coli to identify differences in their properties and to predict new potential drug targets.

RESULTS

When compared to non-targets, bacterial antibiotic targets tend to be long, have high beta-sheet and low alpha-helix contents, are polar, are found in the cytoplasm rather than in membranes, and are usually enzymes, with ligases particularly favoured. Sequence features were used to build a support vector machine model for E. coli proteins, allowing the assignment of any sequence to the drug target or non-target classes, with an accuracy in the training set of 94%. We identified 319 proteins (7%) in the non-target set that have target-like properties, many of which have unknown function. 63 of these proteins have significant and undesirable similarity to a human protein, leaving 256 target like proteins that are not present in humans.

CONCLUSIONS

We suggest that antibiotic discovery programs would be more likely to succeed if new targets are chosen from this set of target like proteins or their homologues. In particular, 64 are essential genes where the cell is not able to recover from a random insertion disruption.

Amyloidogenic sequences in native protein structures

Numerous short peptides have been shown to form beta-sheet amyloid aggregates in vitro. Proteins that contain such sequences are likely to be problematic for a cell, due to their potential to aggregate into toxic structures. We investigated the structures of 30 proteins containing 45 sequences known to form amyloid, to see how the proteins cope with the presence of these potentially toxic sequences, studying secondary structure, hydrogen-bonding, solvent accessible surface area and hydrophobicity. We identified two mechanisms by which proteins avoid aggregation: Firstly, amyloidogenic sequences are often found within helices, despite their inherent preference to form beta structure. Helices may offer a selective advantage, since in order to form amyloid the sequence will presumably have to first unfold and then refold into a beta structure. Secondly, amyloidogenic sequences that are found in beta structure are usually buried within the protein. Surface exposed amyloidogenic sequences are not tolerated in strands, presumably because they lead to protein aggregation via assembly of the amyloidogenic regions. The use of alpha-helices, where amyloidogenic sequences are forced into helix, despite their intrinsic preference for beta structure, is thus a widespread mechanism to avoid protein aggregation.

SitesIdentify: a protein functional site prediction tool

Background

The rate of protein structures being deposited in the Protein Data Bank surpasses the capacity to experimentally characterise them and therefore computational methods to analyse these structures have become increasingly important. Identifying the region of the protein most likely to be involved in function is useful in order to gain information about its potential role. There are many available approaches to predict functional site, but many are not made available via a publicly-accessible application.

Results

Here we present a functional site prediction tool (SitesIdentify), based on combining sequence conservation information with geometry-based cleft identification, that is freely available via a web-server. We have shown that SitesIdentify compares favourably to other functional site prediction tools in a comparison of seven methods on a non-redundant set of 237 enzymes with annotated active sites.

Conclusion

SitesIdentify is able to produce comparable accuracy in predicting functional sites to its closest available counterpart, but in addition achieves improved accuracy for proteins with few characterised homologues. SitesIdentify is available via a webserver at http://www.manchester.ac.uk/bioinformatics/sitesidentify/

Statistical thermodynamics of the collagen triple-helix/coil transition. Free energies for amino acid substitutions within the triple-helix

Collagen sequences frequently deviate from the most thermally stable (Gly-Pro-Hyp)(n) pattern, with many mutations causing osteogenesis imperfecta (or "brittle bone disease"). The effects of collagen mutations have been studied in short peptides. The analysis of this work is problematic, however, as triple-helices fray from their ends, making the coil/triple-helix equilibrium non-two-state. Here, I develop a statistical thermodynamic model to handle this equilibrium that is applicable to peptides that follow the (G-X-Y)(n) pattern, where Gly is present at every third position and where all three chains are identical. Parameters for substitutions at each position are included, as well as a penalty for initiating triple-helix formation. The model is applied to equilibrium experimental data at 37 degrees C to show that the extension of a triple-helix by a three residue unit stabilizes the triple-helix by 0.76 kcal/mol for Gly-Pro-Hyp and 0.33 kcal/mol for Gly-Pro-Pro. The replacement of Hyp by Arg, Asp, or Trp destabilizes the triple-helix by 1.5, 2.4, and 2.9 kcal/mol, respectively, where the substitution is present once in each chain. The model can thus be used to quantitatively interpret data on collagen peptides, giving free energies that can help rationalize mutations that affect collagen stability, and to design new collagen sequences.

Sequence and structural features of enzymes and their active sites by EC class

We have analysed a non-redundant set of 294 enzymes for differences in sequence and structural features between the six main Enzyme Commission (EC) classification groups. This systematic study of enzymes, and their active sites in particular, aims to increase understanding of how the structure of an enzyme relates to its functional role. Many features showed significant differences between the EC classes, including active-site polarity, enzyme size and active-site amino acid propensities. Many attributes correlate with each other to form clusters of related features from which we chose representative features for further analysis. Oxidoreductases have more non-polar active sites, which can be attributed to cofactor binding and a preference for Glu over Asp in active sites in comparison to the other classes. Lyases form a significantly higher proportion of oligomers than any other class, whilst the hydrolases form the largest proportion of monomers. These features were then used in a prediction model that classified each enzyme into its top EC class with an accuracy of 33.1%, which is an increase of 16.4% over random classification. Understanding the link between structure and function is critical to improving enzyme design and the prediction of protein function from structure without transfer of annotation from alignments.

Design of an N-methylated peptide inhibitor of alpha-synuclein aggregation guided by solid-state NMR

Many neurodegenerative diseases are associated with the aggregation of misfolded proteins into amyloid oligomers or fibrils that are deposited as pathological lesions within areas of the brain. An attractive therapeutic strategy for preventing or ameliorating amyloid formation is to identify agents that inhibit the onset or propagation of protein aggregation. Here we demonstrate how solid-state nuclear magnetic resonance (ssNMR) may be used to identify key residues within amyloidogenic protein sequences that may be targeted to inhibit the aggregation of the host protein. For alpha-synuclein, the major protein component of Lewy bodies associated with Parkinson's disease, we have used a combination of ssNMR and biochemical data to identify the key region for self-aggregation of the protein as residues 77-82 (VAQKTV). We used our new structural information to design a peptide derived from residues 77 to 82 of alpha-synuclein with an N-methyl group at the C-terminal residue, which was able to disrupt the aggregation of alpha-synuclein. Thus, we have shown how structural data obtained from ssNMR can guide the design of modified peptides for use as amyloid inhibitors, as a primary step toward developing therapeutic compounds for prevention and/or treatment of amyloid diseases.

Peptide inhibitors of beta-amyloid aggregation

The aggregation of the beta-amyloid peptide (Abeta) into toxic oligomers is probably the key pathogenic event in the onset of Alzheimer's disease, and is therefore a target for the treatment of the disease. A plethora of organic molecules have been shown to inhibit Abeta aggregation and toxicity. Many inhibitors are peptides or peptidomimetics, typically based on the amino acids 16 to 20 of Abeta (KLVFF), which binds to Abeta in isolation. Peptides have been modified by adding bulky groups, charged sequences or polyethylene glycol to their termini, or by N-methylation, replacing a backbone amide by an ester, and by replacing a residue with proline.

Structure and stability of the alpha-helix: lessons for design

The alpha-helix is the most abundant secondary structure in proteins. We now have an excellent understanding of the rules for helix formation because of experimental studies of helices in isolated peptides and within proteins, examination of helices in crystal structures, computer modeling and simulations, and theoretical work. Here we discuss structural features that are important for designing peptide helices, including amino acid preferences for interior and terminal positions, side chain interactions, disulfide bonding, metal binding, and phosphorylation. The solubility and stability of a potential design can be predicted with helical wheels and helix/coil theory, respectively. The helical content of a peptide is most often quantified by circular dichroism, so its use is discussed in detail.

The CXXC motif at the N terminus of an alpha-helical peptide

An active site containing a CXXC motif is always found in the thiol-disulphide oxidoreductase superfamily. A survey of crystal structures revealed that the CXXC motif had a very high local propensity (26.3 +/- 6.2) for the N termini of alpha-helices. A helical peptide with the sequence CAAC at the N terminus was synthesized to examine the helix-stabilizing capacity of the CXXC motif. Circular dichroism was used to confirm the helical nature of the peptide and study behavior under titration with various species. With DTT, a redox potential of E(o) = -230 mV was measured, indicating that the isolated peptide is reducing in nature and similar to native human thioredoxin. The pK(a) values of the individual Cys residues could not be separated in the titration of the reduced state, giving a single transition with an apparent pK(a) of 6.74 (+/-0.06). In the oxidized state, the N-terminal pK(a) is 5.96 (+/-0.05). Analysis of results with the modified helix-coil theory indicated that the disulfide bond stabilized the alpha-helical structure by 0.5 kcal/mol. Reducing the disulfide destabilizes the helix by 0.9 kcal/mol.

N-Methylated peptide inhibitors of beta-amyloid aggregation and toxicity. Optimization of the inhibitor structure

The key pathogenic event in the onset of Alzheimer's disease (AD) is believed to be the aggregation of the beta-amyloid (Abeta) peptide into toxic oligomers. Molecules that interfere with this process may therefore act as therapeutic agents for the treatment of AD. N-Methylated peptides (meptides) are a general class of peptide aggregation inhibitors that act by binding to one face of the aggregating peptide but are unable to hydrogen bond on the other face, because of the N-methyl group replacing a backbone NH group. Here, we optimize the structure of meptide inhibitors of Abeta aggregation, starting with the KLVFF sequence that is known to bind to Abeta. We varied the meptide length, N-methylation sites, acetylation, and amidation of the N and C termini, side-chain identity, and chirality, via five compound libraries. Inhibitor activity was tested by thioflavin T binding, affinity chromatography, electron microscopy, and an 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide toxicity assay. We found that inhibitors should have all d chirality, have a free N terminus but an amidated C terminus, and have large, branched hydrophobic side chains at positions 1-4, while the side chain at position 5 was less important. A single N-methyl group was necessary and sufficient. The most active compound, d-[(chGly)-(Tyr)-(chGly)-(chGly)-(mLeu)]-NH(2), was more active than all previously reported peptide inhibitors. Its related non-N-methylated analogues were insoluble and toxic.

A study of the regional effects of alpha-synuclein on the organization and stability of phospholipid bilayers

Associations between the protein alpha-synuclein (alpha-syn) and presynaptic vesicles have been implicated in synaptic plasticity and neurotransmitter release and may also affect how the protein aggregates into fibrils found in Lewy bodies, the cellular inclusions associated with neurodegenerative diseases. This work investigated how alpha-syn interacts with model phospholipid membranes and examined what effect protein binding has upon the physical properties of lipid bilayers. Wide line 2H and 31P NMR spectra of phospholipid vesicles revealed that alpha-syn associates with membranes containing lipids with anionic headgroups and can disrupt the integrity of the lipid bilayer, but the protein has little effect on membranes of zwitterionic phosphatidylcholine. A peptide, alpha-syn(10-48), which corresponds to the lysine-rich N-terminal region of alpha-syn, was found to associate with lipid headgroups with a preference for a negative membrane surface charge. Another peptide, alpha-syn(120-140), which corresponds to the glutamate-rich C-terminal region, also associates weakly with lipid headgroups but with a slightly higher affinity for membranes with no net surface charge than for negatively charged membrane surfaces. Binding of alpha-syn(10-48) and alpha-syn(120-140) to the lipid vesicles did not disrupt the lamellar structure of the membranes, but both peptides appeared to induce the lateral segregation of the lipids into clusters of acidic lipid-enriched and acidic lipid-deficient domains. From these findings, it is speculated that the N-terminal and C-terminal domains of full-length alpha-syn might act in concert to organize the membrane components during normal protein function and perhaps play a role in presynaptic vesicle synthesis, maintenance, and fusion.

Studies of the aggregation of an amyloidogenic alpha-synuclein peptide fragment

The deposition of alpha-syn (alpha-synuclein) fibrils in Lewy bodies is a characteristic feature of individuals with neurodegenerative disorders. A peptide comprising the central residues 71-82 of alpha-syn [alpha-syn(71-82)] is capable of forming beta-sheet-rich, amyloid-like fibrils with similar morphologies to fibrils of the full-length protein, providing a useful model of pathogenic alpha-syn fibrils that is suitable for detailed structural analysis. We have studied the morphology and gross structural features of alpha-syn(71-82) fibrils formed under different conditions in order to obtain reliable conditions for producing fibrils for further structural investigations. The results indicate that the rate of aggregation and the morphology of the fibrils formed are sensitive to pH and temperature.

Anticooperativity in a Glu-Lys-Glu salt bridge triplet in an isolated alpha-helical peptide

Salt bridges between oppositely charged side chains are well-known to stabilize protein structure, though their contributions vary considerably. Here we study Glu-Lys and Lys-Glu salt bridges, formed when the residues are spaced i, i + 4 surface of an isolated alpha-helix in aqueous solution. Both are stabilizing by -0.60 and -1.02 kcal/mol, respectively, when the interacting residues are fully charged. When the side chains are spaced i, i + 4, i + 8, forming a Glu-Lys-Glu triplet, the second salt bridge provides no additional stabilization to the helix. We attribute this to the inability of the central Lys to form two salt bridges simultaneously. Analysis of these salt bridges in protein structures shows that the Lys-Glu interaction is dominant, with the side chains of the Glu-Lys pair far apart.

Improved prediction for N-termini of alpha-helices using empirical information

The prediction of the secondary structure of proteins from their amino acid sequences remains a key component of many approaches to the protein folding problem. The most abundant form of regular secondary structure in proteins is the alpha-helix, in which specific residue preferences exist at the N-terminal locations. Propensities derived from these observed amino acid frequencies in the Protein Data Bank (PDB) database correlate well with experimental free energies measured for residues at different N-terminal positions in alanine-based peptides. We report a novel method to exploit this data to improve protein secondary structure prediction through identification of the correct N-terminal sequences in alpha-helices, based on existing popular methods for secondary structure prediction. With this algorithm, the number of correctly predicted alpha-helix start positions was improved from 30% to 38%, while the overall prediction accuracy (Q3) remained the same, using cross-validated testing. Although the algorithm was developed and tested on multiple sequence alignment-based secondary structure predictions, it was also able to improve the predictions of start locations by methods that use single sequences to make their predictions. Furthermore, the residue frequencies at N-terminal positions of the improved predictions better reflect those seen at the N-terminal positions of alpha-helices in proteins. This has implications for areas such as comparative modeling, where a more accurate prediction of the N-terminal regions of alpha-helices should benefit attempts to model adjacent loop regions. The algorithm is available as a Web tool, located at http://rocky.bms.umist.ac.uk/elephant.

A Phosphoserine−Lysine Salt Bridge within an α-Helical Peptide, the Strongest α-Helix Side-Chain Interaction Measured to Date†

Phosphorylation is ubiquitous in control of protein activity, yet its effects on protein structure are poorly understood. Here we investigate the effect of serine phosphorylation in the interior of an alpha-helix when a salt bridge is present between the phosphate group and a positively charged side chain (in this case lysine) at i,i + 4 spacing. The stabilization of the helix is considerable and can overcome the intrinsically low preference of phosphoserine for the interior of the helix. The effect is pH dependent, as both the lysine and phosphate groups are titratable, and so calculations are given for several charge combinations. These results, with our previous work, highlight the different, context-dependent effects of phosphorylation in the alpha-helix. The interaction between the phosphate(2)(-) group and the lysine side chain is the strongest yet recorded in helix-coil studies. The results are of interest both in de novo design of peptides and in understanding the structural modes of control by phosphorylation.