Directionality in protein fold prediction

ProtInteract: A deep learning framework for predicting protein-protein interactions

Proteins mainly perform their functions by interacting with other proteins. Protein-protein interactions underpin various biological activities such as metabolic cycles, signal transduction, and immune response. However, due to the sheer number of proteins, experimental methods for finding interacting and non-interacting protein pairs are time-consuming and costly. We therefore developed the ProtInteract framework to predict protein-protein interaction. ProtInteract comprises two components: first, a novel autoencoder architecture that encodes each protein's primary structure to a lower-dimensional vector while preserving its underlying sequence attributes. This leads to faster training of the second network, a deep convolutional neural network (CNN) that receives encoded proteins and predicts their interaction under three different scenarios. In each scenario, the deep CNN predicts the class of a given encoded protein pair. Each class indicates different ranges of confidence scores corresponding to the probability of whether a predicted interaction occurs or not. The proposed framework features significantly low computational complexity and relatively fast response. The contributions of this work are twofold. First, ProtInteract assimilates the protein's primary structure into a pseudo-time series. Therefore, we leverage the nature of the time series of proteins and their physicochemical properties to encode a protein's amino acid sequence into a lower-dimensional vector space. This approach enables extracting highly informative sequence attributes while reducing computational complexity. Second, the ProtInteract framework utilises this information to identify protein interactions with other proteins based on its amino acid configuration. Our results suggest that the proposed framework performs with high accuracy and efficiency in predicting protein-protein interactions.

Robustness by intrinsically disordered C-termini and translational readthrough

During protein synthesis genetic instructions are passed from DNA via mRNA to the ribosome to assemble a protein chain. Occasionally, stop codons in the mRNA are bypassed and translation continues into the untranslated region (3′-UTR). This process, called translational readthrough (TR), yields a protein chain that becomes longer than would be predicted from the DNA sequence alone. Protein sequences vary in propensity for translational errors, which may yield evolutionary constraints by limiting evolutionary paths. Here we investigated TR in Saccharomyces cerevisiae by analysing ribosome profiling data. We clustered proteins as either prone or non-prone to TR, and conducted comparative analyses. We find that a relatively high frequency (5%) of genes undergo TR, including ribosomal subunit proteins. Our main finding is that proteins undergoing TR are highly expressed and have a higher proportion of intrinsically disordered C-termini. We suggest that highly expressed proteins may compensate for the deleterious effects of TR by having intrinsically disordered C-termini, which may provide conformational flexibility but without distorting native function. Moreover, we discuss whether minimizing deleterious effects of TR is also enabling exploration of the phenotypic landscape of protein isoforms.

Enhancing co-translational folding of heterologous protein by deleting non-essential ribosomal proteins in Pichia pastoris

BACKGROUND

Translational regulation played an important role in the correct folding of heterologous proteins to form bioactive conformations during biogenesis. Translational pausing coordinates protein translation and co-translational folding. Decelerating translation elongation speed has been shown to improve the soluble protein yield when expressing heterologous proteins in industrial expression hosts. However, rational redesign of translational pausing via synonymous mutations may not be feasible in many cases. Our goal was to develop a general and convenient strategy to improve heterologous protein synthesis in Pichia pastoris without mutating the expressed genes.

RESULTS

Here, a large-scale deletion library of ribosomal protein (RP) genes was constructed for heterologous protein expression in Pichia pastoris, and 59% (16/27) RP deletants have significantly increased heterologous protein yield. This is due to the delay of 60S subunit assembly by deleting non-essential ribosomal protein genes or 60S subunit processing factors, thus globally decreased the translation elongation speed and improved the co-translational folding, without perturbing the relative transcription level and translation initiation.

CONCLUSION

Global decrease in the translation elongation speed by RP deletion enhanced co-translational folding efficiency of nascent chains and decreased protein aggregates to improve heterologous protein yield. A potential expression platform for efficient pharmaceutical proteins and industrial enzymes production was provided without synonymous mutation.

Sequential search leads to faster, more efficient fragment-based de novo protein structure prediction

Motivation

Most current de novo structure prediction methods randomly sample protein conformations and thus require large amounts of computational resource. Here, we consider a sequential sampling strategy, building on ideas from recent experimental work which shows that many proteins fold cotranslationally.

Results

We have investigated whether a pseudo-greedy search approach, which begins sequentially from one of the termini, can improve the performance and accuracy of de novo protein structure prediction. We observed that our sequential approach converges when fewer than 20 000 decoys have been produced, fewer than commonly expected. Using our software, SAINT2, we also compared the run time and quality of models produced in a sequential fashion against a standard, non-sequential approach. Sequential prediction produces an individual decoy 1.5-2.5 times faster than non-sequential prediction. When considering the quality of the best model, sequential prediction led to a better model being produced for 31 out of 41 soluble protein validation cases and for 18 out of 24 transmembrane protein cases. Correct models (TM-Score > 0.5) were produced for 29 of these cases by the sequential mode and for only 22 by the non-sequential mode. Our comparison reveals that a sequential search strategy can be used to drastically reduce computational time of de novo protein structure prediction and improve accuracy.

Availability and implementation

Data are available for download from: http://opig.stats.ox.ac.uk/resources. SAINT2 is available for download from: https://github.com/sauloho/SAINT2.

Contact

saulo.deoliveira@dtc.ox.ac.uk.

Supplementary information

Supplementary data are available at Bioinformatics online.

Approaches to ab initio molecular replacement of α-helical transmembrane proteins

α-Helical transmembrane proteins are a ubiquitous and important class of proteins, but present difficulties for crystallographic structure solution. Here, the effectiveness of the AMPLE molecular replacement pipeline in solving α-helical transmembrane-protein structures is assessed using a small library of eight ideal helices, as well as search models derived from ab initio models generated both with and without evolutionary contact information. The ideal helices prove to be surprisingly effective at solving higher resolution structures, but ab initio-derived search models are able to solve structures that could not be solved with the ideal helices. The addition of evolutionary contact information results in a marked improvement in the modelling and makes additional solutions possible.

Genetic Code Optimization for Cotranslational Protein Folding: Codon Directional Asymmetry Correlates with Antiparallel Betasheets, tRNA Synthetase Classes

A new codon property, codon directional asymmetry in nucleotide content (CDA), reveals a biologically meaningful genetic code dimension: palindromic codons (first and last nucleotides identical, codon structure XZX) are symmetric (CDA = 0), codons with structures ZXX/XXZ are 5'/3' asymmetric (CDA = - 1/1; CDA = - 0.5/0.5 if Z and X are both purines or both pyrimidines, assigning negative/positive (-/+) signs is an arbitrary convention). Negative/positive CDAs associate with (a) Fujimoto's tetrahedral codon stereo-table; (b) tRNA synthetase class I/II (aminoacylate the 2'/3' hydroxyl group of the tRNA's last ribose, respectively); and (c) high/low antiparallel (not parallel) betasheet conformation parameters. Preliminary results suggest CDA-whole organism associations (body temperature, developmental stability, lifespan). Presumably, CDA impacts spatial kinetics of codon-anticodon interactions, affecting cotranslational protein folding. Some synonymous codons have opposite CDA sign (alanine, leucine, serine, and valine), putatively explaining how synonymous mutations sometimes affect protein function. Correlations between CDA and tRNA synthetase classes are weaker than between CDA and antiparallel betasheet conformation parameters. This effect is stronger for mitochondrial genetic codes, and potentially drives mitochondrial codon-amino acid reassignments. CDA reveals information ruling nucleotide-protein relations embedded in reversed (not reverse-complement) sequences (5'-ZXX-3'/5'-XXZ-3').

Building a better fragment library for de novo protein structure prediction

Fragment-based approaches are the current standard for de novo protein structure prediction. These approaches rely on accurate and reliable fragment libraries to generate good structural models. In this work, we describe a novel method for structure fragment library generation and its application in fragment-based de novo protein structure prediction. The importance of correct testing procedures in assessing the quality of fragment libraries is demonstrated. In particular, the exclusion of homologs to the target from the libraries to correctly simulate a de novo protein structure prediction scenario, something which surprisingly is not always done. We demonstrate that fragments presenting different predominant predicted secondary structures should be treated differently during the fragment library generation step and that exhaustive and random search strategies should both be used. This information was used to develop a novel method, Flib. On a validation set of 41 structurally diverse proteins, Flib libraries presents both a higher precision and coverage than two of the state-of-the-art methods, NNMake and HHFrag. Flib also achieves better precision and coverage on the set of 275 protein domains used in the two previous experiments of the the Critical Assessment of Structure Prediction (CASP9 and CASP10). We compared Flib libraries against NNMake libraries in a structure prediction context. Of the 13 cases in which a correct answer was generated, Flib models were more accurate than NNMake models for 10. “Flib is available for download at: http://www.stats.ox.ac.uk/research/proteins/resources”.

Universally increased mRNA stability downstream of the translation initiation site in eukaryotes and prokaryotes

Local secondary structures in coding sequences have important functions across various translational processes. To date, however, the local structures and their functions in the early stage of translation elongation remain poorly understood. Here, we surveyed the structural stability in the first 180 nucleotides of the coding sequence of 27 species using computational method. We found that the structural stability in the 30-80 nucleotide interval was significantly higher than that in other regions in eukaryotes and most prokaryotes. No significant correlation between local translation efficiency and structural stability was observed, suggesting that this structural region has undergone selection pressure directly to maintain high stability. Furthermore, ribosome was blocked by this region, providing an opportunity for co-translational regulation. Remarkably, in eukaryotes, we found that mRNAs with higher structural stability in the 30-80 nucleotide interval tended to encode the secreted proteins. Overall, our results revealed a previously unappreciated correlation between structural stability and protein localization.

Cotranslational protein folding and terminus hydrophobicity

Peptides fold on a time scale that is much smaller than the time required for synthesis, whence all proteins potentially fold cotranslationally to some degree (followed by additional folding events after release from the ribosome). In this paper, in three different ways, we find that cotranslational folding success is associated with higher hydrophobicity at the N-terminus than at the C-terminus. First, we fold simple HP models on a square lattice and observe that HP sequences that fold better cotranslationally than from a fully extended state exhibit a positive difference (N-C) in terminus hydrophobicity. Second, we examine real proteins using a previously established measure of potential cotranslationality known as ALR (Average Logarithmic Ratio of the extent of previous contacts) and again find a correlation with the difference in terminus hydrophobicity. Finally, we use the cotranslational protein structure prediction program SAINT and again find that such an approach to folding is more successful for proteins with higher N-terminus than C-terminus hydrophobicity. All results indicate that cotranslational folding is promoted in part by a hydrophobic start and a less hydrophobic finish to the sequence.

Ligand-driven vectorial folding of ribosome-bound human CFTR NBD1

The mechanism by which protein folding is coupled to biosynthesis is a critical, but poorly understood, aspect of protein conformational diseases. Here we use fluorescence resonance energy transfer (FRET) to characterize tertiary structural transitions of nascent polypeptides and show that the first nucleotide-binding domain (NBD1) of human CFTR, whose folding is defective in cystic fibrosis, folds via a cotranslational multistep pathway as it is synthesized on the ribosome. Folding begins abruptly as NBD1 residues 389-500 emerge from the ribosome exit tunnel, initiating compaction of a small, N-terminal α/β-subdomain. Real-time kinetics of synchronized nascent chains revealed that subdomain folding is rapid, occurs coincident with synthesis, and is facilitated by direct ATP binding to the nascent polypeptide. These findings localize the major CF defect late in the NBD1 folding pathway and establish a paradigm wherein a cellular ligand promotes vectorial domain folding by facilitating an energetically favored local peptide conformation.

A combinatorial method to enable detailed investigation of protein–protein interactions

Background: Successful structural investigations of protein–protein interactions can be facilitated by studying only the core interacting regions of the constituent proteins. However, attempting the discovery of stable core complexes using informed trial-and-error approaches can prove time and resource intensive. Methods: We describe a valuable extension of combinatorial domain hunting (CDH), a technology for the timely elucidation of soluble protein truncations. The new method, CDH2, enables empirical discovery of stable protein–protein core complexes. CDH2 is demonstrated ab initio using a previously well-characterized Hsp90/Cdc37 complex. Results: Without using a priori information, we demonstrate the isolation of stable protein–protein complexes, suitable for further analyses. Discussion: This resource-efficient process can provide protein complexes for screening of compounds designed to modulate protein–protein interactions, thus facilitating novel drug discovery.

Folding at the birth of the nascent chain: coordinating translation with co-translational folding

In the living cells, the folding of many proteins is largely believed to begin co-translationally, during their biosynthesis at the ribosomes. In the ribosomal tunnel, the nascent peptide may establish local interactions and stabilize α-helical structures. Long-range contacts are more likely outside the ribosomes after release of larger segments of the nascent chain. Examples suggest that domains can attain native-like structure on the ribosome with and without population of folding intermediates. The co-translational folding is limited by the speed of the gradual extrusion of the nascent peptide which imposes conformational restraints on its folding landscape. Recent experimental and in silico modeling studies indicate that translation kinetics fine-tunes co-translational folding by providing a time delay for sequential folding of distinct portions of the nascent chain.