PaleAle 5.0: prediction of protein relative solvent accessibility by deep learning

DL-TCNN: Deep Learning-based Temporal Convolutional Neural Network for prediction of conformational B-cell epitopes

Prediction of conformational B-cell epitopes (CBCE) is an essential phase for vaccine design, drug invention, and accurate disease diagnosis. Many laboratorial and computational approaches have been developed to predict CBCE. However, laboratorial experiments are costly and time consuming, leading to the popularity of Machine Learning (ML)-based computational methods. Although ML methods have succeeded in many domains, achieving higher accuracy in CBCE prediction remains a challenge. To overcome this drawback and consider the limitations of ML methods, this paper proposes a novel DL-based framework for CBCE prediction, leveraging the capabilities of deep learning in the medical domain. The proposed model is named Deep Learning-based Temporal Convolutional Neural Network (DL-TCNN), which hybridizes empirical hyper-tuned 1D-CNN and TCN. TCN is an architecture that employs causal convolutions and dilations, adapting well to sequential input with extensive receptive fields. To train the proposed model, physicochemical features are firstly extracted from antigen sequences. Next, the Synthetic Minority Oversampling Technique (SMOTE) is applied to address the class imbalance problem. Finally, the proposed DL-TCNN is employed for the prediction of CBCE. The model's performance is evaluated and validated on a benchmark antigen-antibody dataset. The DL-TCNN achieves 94.44% accuracy, and 0.989 AUC score for the training dataset, 78.53% accuracy, and 0.661 AUC score for the validation dataset; and 85.10% accuracy, 0.855 AUC score for the testing dataset. The proposed model outperforms all the existing CBCE methods.

IGPRED-MultiTask: A Deep Learning Model to Predict Protein Secondary Structure, Torsion Angles and Solvent Accessibility

Protein secondary structure, solvent accessibility and torsion angle predictions are preliminary steps to predict 3D structure of a protein. Deep learning approaches have achieved significant improvements in predicting various features of protein structure. In this study, IGPRED-Multitask, a deep learning model with multi task learning architecture based on deep inception network, graph convolutional network and a bidirectional long short-term memory is proposed. Moreover, hyper-parameters of the model are fine-tuned using Bayesian optimization, which is faster and more effective than grid search. The same benchmark test data sets as in the OPUS-TASS paper including TEST2016, TEST2018, CASP12, CASP13, CASPFM, HARD68, CAMEO93, CAMEO93_HARD, as well as the train and validation sets, are used for fair comparison with the literature. Statistically significant improvements are observed in secondary structure prediction on 4 datasets, in phi angle prediction on 2 datasets and in psi angel prediction on 3 datasets compared to the state-of-the-art methods. For solvent accessibility prediction, TEST2016 and TEST2018 datasets are used only to assess the performance of the proposed model.

Structural characterisation of amyloidogenic intrinsically disordered zinc finger protein isoforms DPF3b and DPF3a

Double PHD fingers 3 (DPF3) is a zinc finger protein, found in the BAF chromatin remodelling complex, and is involved in the regulation of gene expression. Two DPF3 isoforms have been identified, respectively named DPF3b and DPF3a. Very limited structural information is available for these isoforms, and their specific functionality still remains poorly studied. In a previous work, we have demonstrated the first evidence of DPF3a being a disordered protein sensitive to amyloid fibrillation. Intrinsically disordered proteins (IDPs) lack a defined tertiary structure, existing as a dynamic conformational ensemble, allowing them to act as hubs in protein-protein interaction networks. In the present study, we have more thoroughly characterised DPF3a in vitro behaviour, as well as unravelled and compared the structural properties of the DPF3b isoform, using an array of predictors and biophysical techniques. Predictions, spectroscopy, and dynamic light scattering have revealed a high content in disorder: prevalence of random coil, aromatic residues partially to fully exposed to the solvent, and large hydrodynamic diameters. DPF3a appears to be more disordered than DPF3b, and exhibits more expanded conformations. Furthermore, we have shown that they both time-dependently aggregate into amyloid fibrils, as revealed by typical circular dichroism, deep-blue autofluorescence, and amyloid-dye binding assay fingerprints. Although spectroscopic and microscopic analyses have unveiled that they share a similar aggregation pathway, DPF3a fibrillates at a faster rate, likely through reordering of its C-terminal domain.

Protein Science Meets Artificial Intelligence: A Systematic Review and a Biochemical Meta-Analysis of an Inter-Field

Proteins are some of the most fascinating and challenging molecules in the universe, and they pose a big challenge for artificial intelligence. The implementation of machine learning/AI in protein science gives rise to a world of knowledge adventures in the workhorse of the cell and proteome homeostasis, which are essential for making life possible. This opens up epistemic horizons thanks to a coupling of human tacit–explicit knowledge with machine learning power, the benefits of which are already tangible, such as important advances in protein structure prediction. Moreover, the driving force behind the protein processes of self-organization, adjustment, and fitness requires a space corresponding to gigabytes of life data in its order of magnitude. There are many tasks such as novel protein design, protein folding pathways, and synthetic metabolic routes, as well as protein-aggregation mechanisms, pathogenesis of protein misfolding and disease, and proteostasis networks that are currently unexplored or unrevealed. In this systematic review and biochemical meta-analysis, we aim to contribute to bridging the gap between what we call binomial artificial intelligence (AI) and protein science (PS), a growing research enterprise with exciting and promising biotechnological and biomedical applications. We undertake our task by exploring “the state of the art” in AI and machine learning (ML) applications to protein science in the scientific literature to address some critical research questions in this domain, including What kind of tasks are already explored by ML approaches to protein sciences? What are the most common ML algorithms and databases used? What is the situational diagnostic of the AI–PS inter-field? What do ML processing steps have in common? We also formulate novel questions such as Is it possible to discover what the rules of protein evolution are with the binomial AI–PS? How do protein folding pathways evolve? What are the rules that dictate the folds? What are the minimal nuclear protein structures? How do protein aggregates form and why do they exhibit different toxicities? What are the structural properties of amyloid proteins? How can we design an effective proteostasis network to deal with misfolded proteins? We are a cross-functional group of scientists from several academic disciplines, and we have conducted the systematic review using a variant of the PICO and PRISMA approaches. The search was carried out in four databases (PubMed, Bireme, OVID, and EBSCO Web of Science), resulting in 144 research articles. After three rounds of quality screening, 93 articles were finally selected for further analysis. A summary of our findings is as follows: regarding AI applications, there are mainly four types: 1) genomics, 2) protein structure and function, 3) protein design and evolution, and 4) drug design. In terms of the ML algorithms and databases used, supervised learning was the most common approach (85%). As for the databases used for the ML models, PDB and UniprotKB/Swissprot were the most common ones (21 and 8%, respectively). Moreover, we identified that approximately 63% of the articles organized their results into three steps, which we labeled pre-process, process, and post-process. A few studies combined data from several databases or created their own databases after the pre-process. Our main finding is that, as of today, there are no research road maps serving as guides to address gaps in our knowledge of the AI–PS binomial. All research efforts to collect, integrate multidimensional data features, and then analyze and validate them are, so far, uncoordinated and scattered throughout the scientific literature without a clear epistemic goal or connection between the studies. Therefore, our main contribution to the scientific literature is to offer a road map to help solve problems in drug design, protein structures, design, and function prediction while also presenting the “state of the art” on research in the AI–PS binomial until February 2021. Thus, we pave the way toward future advances in the synthetic redesign of novel proteins and protein networks and artificial metabolic pathways, learning lessons from nature for the welfare of humankind. Many of the novel proteins and metabolic pathways are currently non-existent in nature, nor are they used in the chemical industry or biomedical field.

BIAPSS: A Comprehensive Physicochemical Analyzer of Proteins Undergoing Liquid-Liquid Phase Separation

The liquid-liquid phase separation (LLPS) of biomolecules is a phenomenon which is nowadays recognized as the driving force for the biogenesis of numerous functional membraneless organelles and cellular bodies. The interplay between the protein primary sequence and phase separation remains poorly understood, despite intensive research. To uncover the sequence-encoded signals of protein capable of undergoing LLPS, we developed a novel web platform named BIAPSS (Bioinformatics Analysis of LLPS Sequences). This web server provides on-the-fly analysis, visualization, and interpretation of the physicochemical and structural features for the superset of curated LLPS proteins.

End-to-End Deep Learning Model to Predict and Design Secondary Structure Content of Structural Proteins

Structural proteins are the basis of many biomaterials and key construction and functional components of all life. Further, it is well-known that the diversity of proteins' function relies on their local structures derived from their primary amino acid sequences. Here, we report a deep learning model to predict the secondary structure content of proteins directly from primary sequences, with high computational efficiency. Understanding the secondary structure content of proteins is crucial to designing proteins with targeted material functions, especially mechanical properties. Using convolutional and recurrent architectures and natural language models, our deep learning model predicts the content of two essential types of secondary structures, the α-helix and the β-sheet. The training data are collected from the Protein Data Bank and contain many existing protein geometries. We find that our model can learn the hidden features as patterns of input sequences that can then be directly related to secondary structure content. The α-helix and β-sheet content predictions show excellent agreement with training data and newly deposited protein structures that were recently identified and that were not included in the original training set. We further demonstrate the features of the model by a search for de novo protein sequences that optimize max/min α-helix/β-sheet content and compare the predictions with folded models of these sequences based on AlphaFold2. Excellent agreement is found, underscoring that our model has predictive potential for rapidly designing proteins with specific secondary structures and could be widely applied to biomedical industries, including protein biomaterial designs and regenerative medicine applications.

DeepREx-WS: A web server for characterising protein-solvent interaction starting from sequence

Protein–solvent interaction provides important features for protein surface engineering when the structure is absent or partially solved. Presently, we can integrate the notion of solvent exposed/buried residues with that of their flexibility and intrinsic disorder to highlight regions where mutations may increase or decrease protein stability in order to modify proteins for biotechnological reasons, while preserving their functional integrity. Here we describe a web server, which provides the unique possibility of integrating knowledge of solvent and non-solvent exposure with that of residue conservation, flexibility and disorder of a protein sequence, for a better understanding of which regions are relevant for protein integrity. The core of the webserver is DeepREx, a novel deep learning-based tool that classifies each residue in the sequence as buried or exposed. DeepREx is trained on a high-quality, non-redundant dataset derived from the Protein Data Bank comprising 2332 monomeric protein chains and benchmarked on a blind test set including 200 protein sequences unrelated with the training set. Results show that DeepREx performs at the state-of-the-art in the field. In turn, the Web Server, DeepREx-WS, supplements the predictions of DeepREx with features that allow a better characterisation of exposed and buried regions: i) residue conservation derived from multiple sequence alignment; ii) local sequence hydrophobicity; iii) residue flexibility computed with MEDUSA; iv) a predictor of secondary structure; v) the presence of disordered regions as derived from MobiDB-Lite3.0. The web server allows browsing, selecting and intersecting the different features. We demonstrate a possible application of the DeepREx-WS for assisting the identification of residues to be variated in protein surface engineering processes.

Improved protein relative solvent accessibility prediction using deep multi-view feature learning framework

The accurate prediction of the relative solvent accessibility of a protein is critical to understanding its 3D structure and biological function. In this study, a novel deep multi-view feature learning (DMVFL) framework that integrates three different neural network units, i.e., bidirectional long short-term memory recurrent neural network, squeeze-and-excitation, and fully-connected hidden layer, with four sequence-based single-view features, i.e., position-specific scoring matrix, position-specific frequency matrix, predicted secondary structure, and roughly predicted three-state relative solvent accessibility probability, is developed to accurately predict relative solvent accessibility information of protein. On the basis of this newly developed framework, one new protein relative solvent accessibility predictor was proposed and called DMVFL-RSA, which employs a customized multiple feedback mechanism that helps to extract discriminative information embedded in the four single-view features. In benchmark tests on TEST524 and CASP14-derived (CASP14set) datasets, DMVFL-RSA outperforms other existing state-of-the-art protein relative solvent accessibility predictors when predicting two-state (exposure threshold of 25%), three-state (exposure thresholds of 9% and 36%), and four-state (exposure thresholds of 4%, 25%, and 50%) discrete values. For real-valued prediction on TEST524 and CASP14set, DMVFL-RSA has also gained high Pearson correlation coefficient values, indicating a positive correlation between the predicted and native relative solvent accessibility. Detailed analyses show that the major advantages of DMVFL-RSA lie in the high efficiency of the DMVFL framework, the applied multiple feedback mechanism, and the strong sensitivity of the sequence-based features. The web server of DMVFL-RSA is freely available at https://jun-csbio.github.io/DMVFL-RSA/for academic use. The standalone package of DMVFL-RSA is downloadable at https://github.com/XueQiangFan/DMVFL-RSA.

Accurate Estimation of Solvent Accessible Surface Area for Coarse-Grained Biomolecular Structures with Deep Learning

Coarse-grained (CG) models of biomolecules have been widely used in protein/ribonucleic acid (RNA) three-dimensional structure prediction, docking, drug design, and molecular simulations due to their superiority in computational efficiency. Most of these applications strongly depend on the reasonable estimation of solvation free energy, which requires the accurate calculation of solvent accessible surface area (SASA). Although algorithms for SASA calculations with all-atom protein and RNA structures have been well-established, accurately estimating the SASA based on CG structures is extremely challenging. In this work, we developed a deep learning-based SASA estimator (DeepCGSA), which can provide almost perfect SASA estimation based on CG structures of protein and RNA molecules. Extensive testing analysis showed that for three types of widely used CG protein models, including the Cα-based, Cα-Cβ, and Martini models, the correlation coefficients between the predicted values and the reference values can be as high as 0.95-0.99, which perform dramatically better than available methods. In addition, the new method can be used for CG RNA structures and unfolded protein structures with much improved accuracy. We anticipate that DeepCGSA will be highly useful in the protein/RNA structure prediction, drug design, and other applications, in which accurate estimations of SASA for CG biomolecular structures are critically important.

Recent Applications of Deep Learning Methods on Evolution- and Contact-Based Protein Structure Prediction

The new advances in deep learning methods have influenced many aspects of scientific research, including the study of the protein system. The prediction of proteins' 3D structural components is now heavily dependent on machine learning techniques that interpret how protein sequences and their homology govern the inter-residue contacts and structural organization. Especially, methods employing deep neural networks have had a significant impact on recent CASP13 and CASP14 competition. Here, we explore the recent applications of deep learning methods in the protein structure prediction area. We also look at the potential opportunities for deep learning methods to identify unknown protein structures and functions to be discovered and help guide drug-target interactions. Although significant problems still need to be addressed, we expect these techniques in the near future to play crucial roles in protein structural bioinformatics as well as in drug discovery.

SCLpred-MEM: Subcellular localization prediction of membrane proteins by deep N-to-1 convolutional neural networks

The knowledge of the subcellular location of a protein is a valuable source of information in genomics, drug design, and various other theoretical and analytical perspectives of bioinformatics. Due to the expensive and time-consuming nature of experimental methods of protein subcellular location determination, various computational methods have been developed for subcellular localization prediction. We introduce "SCLpred-MEM," an ab initio protein subcellular localization predictor, powered by an ensemble of Deep N-to-1 Convolutional Neural Networks (N1-NN) trained and tested on strict redundancy reduced datasets. SCLpred-MEM is available as a web-server predicting query proteins into two classes, membrane and non-membrane proteins. SCLpred-MEM achieves a Matthews correlation coefficient of 0.52 on a strictly homology-reduced independent test set and 0.62 on a less strict homology reduced independent test set, surpassing or matching other state-of-the-art subcellular localization predictors.

iEnhancer-GAN: A Deep Learning Framework in Combination with Word Embedding and Sequence Generative Adversarial Net to Identify Enhancers and Their Strength

As critical components of DNA, enhancers can efficiently and specifically manipulate the spatial and temporal regulation of gene transcription. Malfunction or dysregulation of enhancers is implicated in a slew of human pathology. Therefore, identifying enhancers and their strength may provide insights into the molecular mechanisms of gene transcription and facilitate the discovery of candidate drug targets. In this paper, a new enhancer and its strength predictor, iEnhancer-GAN, is proposed based on a deep learning framework in combination with the word embedding and sequence generative adversarial net (Seq-GAN). Considering the relatively small training dataset, the Seq-GAN is designed to generate artificial sequences. Given that each functional element in DNA sequences is analogous to a “word” in linguistics, the word segmentation methods are proposed to divide DNA sequences into “words”, and the skip-gram model is employed to transform the “words” into digital vectors. In view of the powerful ability to extract high-level abstraction features, a convolutional neural network (CNN) architecture is constructed to perform the identification tasks, and the word vectors of DNA sequences are vertically concatenated to form the embedding matrices as the input of the CNN. Experimental results demonstrate the effectiveness of the Seq-GAN to expand the training dataset, the possibility of applying word segmentation methods to extract “words” from DNA sequences, the feasibility of implementing the skip-gram model to encode DNA sequences, and the powerful prediction ability of the CNN. Compared with other state-of-the-art methods on the training dataset and independent test dataset, the proposed method achieves a significantly improved overall performance. It is anticipated that the proposed method has a certain promotion effect on enhancer related fields.

Solvent Accessibility of Residues Undergoing Pathogenic Variations in Humans: From Protein Structures to Protein Sequences

Solvent accessibility (SASA) is a key feature of proteins for determining their folding and stability. SASA is computed from protein structures with different algorithms, and from protein sequences with machine-learning based approaches trained on solved structures. Here we ask the question as to which extent solvent exposure of residues can be associated to the pathogenicity of the variation. By this, SASA of the wild-type residue acquires a role in the context of functional annotation of protein single-residue variations (SRVs). By mapping variations on a curated database of human protein structures, we found that residues targeted by disease related SRVs are less accessible to solvent than residues involved in polymorphisms. The disease association is not evenly distributed among the different residue types: SRVs targeting glycine, tryptophan, tyrosine, and cysteine are more frequently disease associated than others. For all residues, the proportion of disease related SRVs largely increases when the wild-type residue is buried and decreases when it is exposed. The extent of the increase depends on the residue type. With the aid of an in house developed predictor, based on a deep learning procedure and performing at the state-of-the-art, we are able to confirm the above tendency by analyzing a large data set of residues subjected to variations and occurring in some 12,494 human protein sequences still lacking three-dimensional structure (derived from HUMSAVAR). Our data support the notion that surface accessible area is a distinguished property of residues that undergo variation and that pathogenicity is more frequently associated to the buried property than to the exposed one.

Brewery: deep learning and deeper profiles for the prediction of 1D protein structure annotations

MOTIVATION

Protein structural annotations (PSAs) are essential abstractions to deal with the prediction of protein structures. Many increasingly sophisticated PSAs have been devised in the last few decades. However, the need for annotations that are easy to compute, process and predict has not diminished. This is especially true for protein structures that are hardest to predict, such as novel folds.

RESULTS

We propose Brewery, a suite of ab initio predictors of 1D PSAs. Brewery uses multiple sources of evolutionary information to achieve state-of-the-art predictions of secondary structure, structural motifs, relative solvent accessibility and contact density.

AVAILABILITY AND IMPLEMENTATION

The web server, standalone program, Docker image and training sets of Brewery are available at http://distilldeep.ucd.ie/brewery/.

CONTACT

gianluca.pollastri@ucd.ie.

SCLpred-EMS: subcellular localization prediction of endomembrane system and secretory pathway proteins by Deep N-to-1 Convolutional Neural Networks

MOTIVATION

The subcellular location of a protein can provide useful information for protein function prediction and drug design. Experimentally determining the subcellular location of a protein is an expensive and time-consuming task. Therefore, various computer-based tools have been developed, mostly using machine learning algorithms, to predict the subcellular location of proteins.

RESULTS

Here, we present a neural network-based algorithm for protein subcellular location prediction. We introduce SCLpred-EMS a subcellular localization predictor powered by an ensemble of Deep N-to-1 Convolutional Neural Networks. SCLpred-EMS predicts the subcellular location of a protein into two classes, the endomembrane system and secretory pathway versus all others, with a Matthews correlation coefficient of 0.75-0.86 outperforming the other state-of-the-art web servers we tested.

AVAILABILITY AND IMPLEMENTATION

SCLpred-EMS is freely available for academic users at http://distilldeep.ucd.ie/SCLpred2/.

CONTACT

catherine.mooney@ucd.ie.

Impact of protein dynamics on secondary structure prediction

Protein 3D structures support their biological functions. As the number of protein structures is negligible in regards to the number of available protein sequences, prediction methodologies relying only on protein sequences are essential tools. In this field, protein secondary structure prediction (PSSPs) is a mature area, and is considered to have reached a plateau. Nonetheless, proteins are highly dynamical macromolecules, a property that could impact the PSSP methods. Indeed, in a previous study, the stability of local protein conformations was evaluated demonstrating that some regions easily changed to another type of secondary structure. The protein sequences of this dataset were used by PSSPs and their results compared to molecular dynamics to investigate their potential impact on the quality of the secondary structure prediction. Interestingly, a direct link is observed between the quality of the prediction and the stability of the assignment to the secondary structure state. The more stable a local protein conformation is, the better the prediction will be. The secondary structure assignment not taken from the crystallized structures but from the conformations observed during the dynamics slightly increase the quality of the secondary structure prediction. These results show that evaluation of PSSPs can be done differently, but also that the notion of dynamics can be included in development of PSSPs and other approaches such as de novo approaches.

Using physical features of protein core packing to distinguish real proteins from decoys

The ability to consistently distinguish real protein structures from computationally generated model decoys is not yet a solved problem. One route to distinguish real protein structures from decoys is to delineate the important physical features that specify a real protein. For example, it has long been appreciated that the hydrophobic cores of proteins contribute significantly to their stability. We used two sources to obtain datasets of decoys to compare with real protein structures: submissions to the biennial Critical Assessment of protein Structure Prediction competition, in which researchers attempt to predict the structure of a protein only knowing its amino acid sequence, and also decoys generated by 3DRobot, which have user-specified global root-mean-squared deviations from experimentally determined structures. Our analysis revealed that both sets of decoys possess cores that do not recapitulate the key features that define real protein cores. In particular, the model structures appear more densely packed (because of energetically unfavorable atomic overlaps), contain too few residues in the core, and have improper distributions of hydrophobic residues throughout the structure. Based on these observations, we developed a feed-forward neural network, which incorporates key physical features of protein cores, to predict how well a computational model recapitulates the real protein structure without knowledge of the structure of the target sequence. By identifying the important features of protein structure, our method is able to rank decoy structures with similar accuracy to that obtained by state-of-the-art methods that incorporate many additional features. The small number of physical features makes our model interpretable, emphasizing the importance of protein packing and hydrophobicity in protein structure prediction.

Deep learning methods in protein structure prediction

Protein Structure Prediction is a central topic in Structural Bioinformatics. Since the '60s statistical methods, followed by increasingly complex Machine Learning and recently Deep Learning methods, have been employed to predict protein structural information at various levels of detail. In this review, we briefly introduce the problem of protein structure prediction and essential elements of Deep Learning (such as Convolutional Neural Networks, Recurrent Neural Networks and basic feed-forward Neural Networks they are founded on), after which we discuss the evolution of predictive methods for one-dimensional and two-dimensional Protein Structure Annotations, from the simple statistical methods of the early days, to the computationally intensive highly-sophisticated Deep Learning algorithms of the last decade. In the process, we review the growth of the databases these algorithms are based on, and how this has impacted our ability to leverage knowledge about evolution and co-evolution to achieve improved predictions. We conclude this review outlining the current role of Deep Learning techniques within the wider pipelines to predict protein structures and trying to anticipate what challenges and opportunities may arise next.