Deep Learning in Prediction of Intrinsic Disorder in Proteins

Taxonomy-specific assessment of intrinsic disorder predictions at residue and region levels in higher eukaryotes, protists, archaea, bacteria and viruses

Intrinsic disorder predictors were evaluated in several studies including the two large CAID experiments. However, these studies are biased towards eukaryotic proteins and focus primarily on the residue-level predictions. We provide first-of-its-kind assessment that comprehensively covers the taxonomy and evaluates predictions at the residue and disordered region levels. We curate a benchmark dataset that uniformly covers eukaryotic, archaeal, bacterial, and viral proteins. We find that predictive performance differs substantially across taxonomy, where viruses are predicted most accurately, followed by protists and higher eukaryotes, while bacterial and archaeal proteins suffer lower levels of accuracy. These trends are consistent across predictors. We also find that current tools, except for flDPnn, struggle with reproducing native distributions of the numbers and sizes of the disordered regions. Moreover, analysis of two variants of disorder predictions derived from the AlphaFold2 predicted structures reveals that they produce accurate residue-level propensities for archaea, bacteria and protists. However, they underperform for higher eukaryotes and generally struggle to accurately identify disordered regions. Our results motivate development of new predictors that target bacteria and archaea and which produce accurate results at both residue and region levels. We also stress the need to include the region-level assessments in future assessments.

Protein subcellular localization prediction tools

Protein subcellular localization prediction is of great significance in bioinformatics and biological research. Most of the proteins do not have experimentally determined localization information, computational prediction methods and tools have been acting as an active research area for more than two decades now. Knowledge of the subcellular location of a protein provides valuable information about its functionalities, the functioning of the cell, and other possible interactions with proteins. Fast, reliable, and accurate predictors provides platforms to harness the abundance of sequence data to predict subcellular locations accordingly. During the last decade, there has been a considerable amount of research effort aimed at developing subcellular localization predictors. This paper reviews recent subcellular localization prediction tools in the Eukaryotic, Prokaryotic, and Virus-based categories followed by a detailed analysis. Each predictor is discussed based on its main features, strengths, weaknesses, algorithms used, prediction techniques, and analysis. This review is supported by prediction tools taxonomies that highlight their rele- vant area and examples for uncomplicated categorization and ease of understandability. These taxonomies help users find suitable tools according to their needs. Furthermore, recent research gaps and challenges are discussed to cover areas that need the utmost attention. This survey provides an in-depth analysis of the most recent prediction tools to facilitate readers and can be considered a quick guide for researchers to identify and explore the recent literature advancements.

Overview of AlphaFold2 and breakthroughs in overcoming its limitations

Predicting three-dimensional (3D) protein structures has been challenging for decades. The emergence of AlphaFold2 (AF2), a deep learning-based machine learning method developed by DeepMind, became a game changer in the protein folding community. AF2 can predict a protein's three-dimensional structure with high confidence based on its amino acid sequence. Accurate prediction of protein structures can dramatically accelerate our understanding of biological mechanisms and provide a solid foundation for reliable drug design. Although AF2 breaks through the barriers in predicting protein structures, many rooms remain to be further studied. This review provides a brief historical overview of the development of protein structure prediction, covering template-based, template-free, and machine learning-based methods. In addition to reviewing the potential benefits (Pros) and considerations (Cons) of using AF2, this review summarizes the diverse applications, including protein structure predictions, dynamic changes, point mutation, integration of language model and experimental data, protein complex, and protein-peptide interaction. It underscores recent advancements in efficiency, reliability, and broad application of AF2. This comprehensive review offers valuable insights into the applications of AF2 and AF2-inspired AI methods in structural biology and its potential for clinically significant drug target discovery.

The structural landscape and diversity of Pyricularia oryzae MAX effectors revisited

Magnaporthe AVRs and ToxB-like (MAX) effectors constitute a family of secreted virulence proteins in the fungus Pyricularia oryzae (syn. Magnaporthe oryzae), which causes blast disease on numerous cereals and grasses. In spite of high sequence divergence, MAX effectors share a common fold characterized by a ß-sandwich core stabilized by a conserved disulfide bond. In this study, we investigated the structural landscape and diversity within the MAX effector repertoire of P. oryzae. Combining experimental protein structure determination and in silico structure modeling we validated the presence of the conserved MAX effector core domain in 77 out of 94 groups of orthologs (OG) identified in a previous population genomic study. Four novel MAX effector structures determined by NMR were in remarkably good agreement with AlphaFold2 (AF2) predictions. Based on the comparison of the AF2-generated 3D models we propose a classification of the MAX effectors superfamily in 20 structural groups that vary in the canonical MAX fold, disulfide bond patterns, and additional secondary structures in N- and C-terminal extensions. About one-third of the MAX family members remain singletons, without strong structural relationship to other MAX effectors. Analysis of the surface properties of the AF2 MAX models also highlights the high variability within the MAX family at the structural level, potentially reflecting the wide diversity of their virulence functions and host targets.

DR-BERT: A protein language model to annotate disordered regions

Despite their lack of a rigid structure, intrinsically disordered regions (IDRs) in proteins play important roles in cellular functions, including mediating protein-protein interactions. Therefore, it is important to computationally annotate IDRs with high accuracy. In this study, we present Disordered Region prediction using Bidirectional Encoder Representations from Transformers (DR-BERT), a compact protein language model. Unlike most popular tools, DR-BERT is pretrained on unannotated proteins and trained to predict IDRs without relying on explicit evolutionary or biophysical data. Despite this, DR-BERT demonstrates significant improvement over existing methods on the Critical Assessment of protein Intrinsic Disorder (CAID) evaluation dataset and outperforms competitors on two out of four test cases in the CAID 2 dataset, while maintaining competitiveness in the others. This performance is due to the information learned during pretraining and DR-BERT's ability to use contextual information.

Accurate and Fast Prediction of Intrinsically Disordered Protein by Multiple Protein Language Models and Ensemble Learning

Intrinsically disordered proteins (IDPs) play a vital role in various biological processes and have attracted increasing attention in the past few decades. Predicting IDPs from the primary structures of proteins offers a rapid and facile means of protein analysis without necessitating crystal structures. In particular, machine learning methods have demonstrated their potential in this field. Recently, protein language models (PLMs) are emerging as a promising approach to extracting essential information from protein sequences and have been employed in protein modeling to utilize their advantages of precision and efficiency. In this article, we developed a novel IDP prediction method named IDP-ELM to predict the intrinsically disordered regions (IDRs) as well as their functions including disordered flexible linkers and disordered protein binding. This method utilizes high-dimensional representations extracted from several state-of-the-art PLMs and predicts IDRs by ensemble learning based on bidirectional recurrent neural networks. The performance of the method was evaluated on two independent test data sets from CAID (critical assessment of protein intrinsic disorder prediction) and CAID2, indicating notable improvements in terms of area under the receiver operating characteristic (AUC), Matthew's correlation coefficient (MCC), and F1 score. Moreover, IDP-ELM requires solely protein sequences as inputs and does not entail a time-consuming process of protein profile generation, which is a prerequisite for most existing state-of-the-art methods, enabling an accurate, fast, and convenient tool for proteome-level analysis. The corresponding reproducible source code and model weights are available at https://github.com/xu-shi-jie/idp-elm.

DescribePROT in 2023: more, higher-quality and experimental annotations and improved data download options

The DescribePROT database of amino acid-level descriptors of protein structures and functions was substantially expanded since its release in 2020. This expansion includes substantial increase in the size, scope, and quality of the underlying data, the addition of experimental structural information, the inclusion of new data download options, and an upgraded graphical interface. DescribePROT currently covers 19 structural and functional descriptors for proteins in 273 reference proteomes generated by 11 accurate and complementary predictive tools. Users can search our resource in multiple ways, interact with the data using the graphical interface, and download data at various scales including individual proteins, entire proteomes, and whole database. The annotations in DescribePROT are useful for a broad spectrum of studies that include investigations of protein structure and function, development and validation of predictive tools, and to support efforts in understanding molecular underpinnings of diseases and development of therapeutics. DescribePROT can be freely accessed at http://biomine.cs.vcu.edu/servers/DESCRIBEPROT/.

Availability of web servers significantly boosts citations rates of bioinformatics methods for protein function and disorder prediction

Motivation

Development of bioinformatics methods is a long, complex and resource-hungry process. Hundreds of these tools were released. While some methods are highly cited and used, many suffer relatively low citation rates. We empirically analyze a large collection of recently released methods in three diverse protein function and disorder prediction areas to identify key factors that contribute to increased citations.

Results

We show that provision of a working web server significantly boosts citation rates. On average, methods with working web servers generate three times as many citations compared to tools that are available as only source code, have no code and no server, or are no longer available. This observation holds consistently across different research areas and publication years. We also find that differences in predictive performance are unlikely to impact citation rates. Overall, our empirical results suggest that a relatively low-cost investment into the provision and long-term support of web servers would substantially increase the impact of bioinformatics tools.

Complex Conformational Space of the RNA Polymerase II C-Terminal Domain upon Phosphorylation

Intrinsically disordered proteins (IDPs) have been closely studied during the past decade due to their importance in many biological processes. The disordered nature of this group of proteins makes it difficult to observe its full span of the conformational space using either experimental or computational studies. In this article, we explored the conformational space of the C-terminal domain (CTD) of RNA polymerase II (Pol II), which is also an intrinsically disordered low complexity domain, using enhanced sampling methods. We provided a detailed conformational analysis of model systems of CTD with different lengths; first with the last 44 residues of the human CTD sequence and finally the CTD model with 2-heptapeptide repeating units. We then investigated the effects of phosphorylation on CTD conformations by performing simulations at different phosphorylated states. We obtained broad conformational spaces in nonphosphorylated CTD models, and phosphorylation has complex effects on the conformations of the CTD. These complex effects depend on the length of the CTD, spacing between the multiple phosphorylation sites, ion coordination, and interactions with the nearby residues.

CoMemMoRFPred: Sequence-based Prediction of MemMoRFs by Combining Predictors of Intrinsic Disorder, MoRFs and Disordered Lipid-binding Regions

Molecular recognition features (MoRFs) are a commonly occurring type of intrinsically disordered regions (IDRs) that undergo disorder-to-order transition upon binding to partner molecules. We focus on recently characterized and functionally important membrane-binding MoRFs (MemMoRFs). Motivated by the lack of computational tools that predict MemMoRFs, we use a dataset of experimentally annotated MemMoRFs to conceptualize, design, evaluate and release an accurate sequence-based predictor. We rely on state-of-the-art tools that predict residues that possess key characteristics of MemMoRFs, such as intrinsic disorder, disorder-to-order transition and lipid-binding. We identify and combine results from three tools that include flDPnn for the disorder prediction, DisoLipPred for the prediction of disordered lipid-binding regions, and MoRFCHiBiLight for the prediction of disorder-to-order transitioning protein binding regions. Our empirical analysis demonstrates that combining results produced by these three methods generates accurate predictions of MemMoRFs. We also show that use of a smoothing operator produces predictions that closely mimic the number and sizes of the native MemMoRF regions. The resulting CoMemMoRFPred method is available as an easy-to-use webserver at http://biomine.cs.vcu.edu/servers/CoMemMoRFPred. This tool will aid future studies of MemMoRFs in the context of exploring their abundance, cellular functions, and roles in pathologic phenomena.

Tutorial: a guide for the selection of fast and accurate computational tools for the prediction of intrinsic disorder in proteins

Intrinsic disorder is instrumental for a wide range of protein functions, and its analysis, using computational predictions from primary structures, complements secondary and tertiary structure-based approaches. In this Tutorial, we provide an overview and comparison of 23 publicly available computational tools with complementary parameters useful for intrinsic disorder prediction, partly relying on results from the Critical Assessment of protein Intrinsic Disorder prediction experiment. We consider factors such as accuracy, runtime, availability and the need for functional insights. The selected tools are available as web servers and downloadable programs, offer state-of-the-art predictions and can be used in a high-throughput manner. We provide examples and instructions for the selected tools to illustrate practical aspects related to the submission, collection and interpretation of predictions, as well as the timing and their limitations. We highlight two predictors for intrinsically disordered proteins, flDPnn as accurate and fast and IUPred as very fast and moderately accurate, while suggesting ANCHOR2 and MoRFchibi as two of the best-performing predictors for intrinsically disordered region binding. We link these tools to additional resources, including databases of predictions and web servers that integrate multiple predictive methods. Altogether, this Tutorial provides a hands-on guide to comparatively evaluating multiple predictors, submitting and collecting their own predictions, and reading and interpreting results. It is suitable for experimentalists and computational biologists interested in accurately and conveniently identifying intrinsic disorder, facilitating the functional characterization of the rapidly growing collections of protein sequences.

The Origin of Discrepancies between Predictions and Annotations in Intrinsically Disordered Proteins

Disorder prediction methods that can discriminate between ordered and disordered regions have contributed fundamentally to our understanding of the properties and prevalence of intrinsically disordered proteins (IDPs) in proteomes as well as their functional roles. However, a recent large-scale assessment of the performance of these methods indicated that there is still room for further improvements, necessitating novel approaches to understand the strengths and weaknesses of individual methods. In this study, we compared two methods, IUPred and disorder prediction, based on the pLDDT scores derived from AlphaFold2 (AF2) models. We evaluated these methods using a dataset from the DisProt database, consisting of experimentally characterized disordered regions and subsets associated with diverse experimental methods and functions. IUPred and AF2 provided consistent predictions in 79% of cases for long disordered regions; however, for 15% of these cases, they both suggested order in disagreement with annotations. These discrepancies arose primarily due to weak experimental support, the presence of intermediate states, or context-dependent behavior, such as binding-induced transitions. Furthermore, AF2 tended to predict helical regions with high pLDDT scores within disordered segments, while IUPred had limitations in identifying linker regions. These results provide valuable insights into the inherent limitations and potential biases of disorder prediction methods.

Comparative evaluation of AlphaFold2 and disorder predictors for prediction of intrinsic disorder, disorder content and fully disordered proteins

We expand studies of AlphaFold2 (AF2) in the context of intrinsic disorder prediction by comparing it against a broad selection of 20 accurate, popular and recently released disorder predictors. We use 25% larger benchmark dataset with 646 proteins and cover protein-level predictions of disorder content and fully disordered proteins. AF2-based disorder predictions secure a relatively high Area Under receiver operating characteristic Curve (AUC) of 0.77 and are statistically outperformed by several modern disorder predictors that secure AUCs around 0.8 with median runtime of about 20 s compared to 1200 s for AF2. Moreover, AF2 provides modestly accurate predictions of fully disordered proteins (F1 = 0.59 vs. 0.91 for the best disorder predictor) and disorder content (mean absolute error of 0.21 vs. 0.15). AF2 also generates statistically more accurate disorder predictions for about 20% of proteins that have relatively short sequences and a few disordered regions that tend to be located at the sequence termini, and which are absent of disordered protein-binding regions. Interestingly, AF2 and the most accurate disorder predictors rely on deep neural networks, suggesting that these models are useful for protein structure and disorder predictions.

Overview Update: Computational Prediction of Intrinsic Disorder in Proteins

There are over 100 computational predictors of intrinsic disorder. These methods predict amino acid-level propensities for disorder directly from protein sequences. The propensities can be used to annotate putative disordered residues and regions. This unit provides a practical and holistic introduction to the sequence-based intrinsic disorder prediction. We define intrinsic disorder, explain the format of computational prediction of disorder, and identify and describe several accurate predictors. We also introduce recently released databases of intrinsic disorder predictions and use an illustrative example to provide insights into how predictions should be interpreted and combined. Lastly, we summarize key experimental methods that can be used to validate computational predictions. © 2023 Wiley Periodicals LLC.

DEPICTER2: a comprehensive webserver for intrinsic disorder and disorder function prediction

Intrinsic disorder in proteins is relatively abundant in nature and essential for a broad spectrum of cellular functions. While disorder can be accurately predicted from protein sequences, as it was empirically demonstrated in recent community-organized assessments, it is rather challenging to collect and compile a comprehensive prediction that covers multiple disorder functions. To this end, we introduce the DEPICTER2 (DisorderEd PredictIon CenTER) webserver that offers convenient access to a curated collection of fast and accurate disorder and disorder function predictors. This server includes a state-of-the-art disorder predictor, flDPnn, and five modern methods that cover all currently predictable disorder functions: disordered linkers and protein, peptide, DNA, RNA and lipid binding. DEPICTER2 allows selection of any combination of the six methods, batch predictions of up to 25 proteins per request and provides interactive visualization of the resulting predictions. The webserver is freely available at http://biomine.cs.vcu.edu/servers/DEPICTER2/.

HybridRNAbind: prediction of RNA interacting residues across structure-annotated and disorder-annotated proteins

The sequence-based predictors of RNA-binding residues (RBRs) are trained on either structure-annotated or disorder-annotated binding regions. A recent study of predictors of protein-binding residues shows that they are plagued by high levels of cross-predictions (protein binding residues are predicted as nucleic acid binding) and that structure-trained predictors perform poorly for the disorder-annotated regions and vice versa. Consequently, we analyze a representative set of the structure and disorder trained predictors of RBRs to comprehensively assess quality of their predictions. Our empirical analysis that relies on a new and low-similarity benchmark dataset reveals that the structure-trained predictors of RBRs perform well for the structure-annotated proteins while the disorder-trained predictors provide accurate results for the disorder-annotated proteins. However, these methods work only modestly well on the opposite types of annotations, motivating the need for new solutions. Using an empirical approach, we design HybridRNAbind meta-model that generates accurate predictions and low amounts of cross-predictions when tested on data that combines structure and disorder-annotated RBRs. We release this meta-model as a convenient webserver which is available at https://www.csuligroup.com/hybridRNAbind/.

Computational prediction of disordered binding regions

One of the key features of intrinsically disordered regions (IDRs) is their ability to interact with a broad range of partner molecules. Multiple types of interacting IDRs were identified including molecular recognition fragments (MoRFs), short linear sequence motifs (SLiMs), and protein-, nucleic acids- and lipid-binding regions. Prediction of binding IDRs in protein sequences is gaining momentum in recent years. We survey 38 predictors of binding IDRs that target interactions with a diverse set of partners, such as peptides, proteins, RNA, DNA and lipids. We offer a historical perspective and highlight key events that fueled efforts to develop these methods. These tools rely on a diverse range of predictive architectures that include scoring functions, regular expressions, traditional and deep machine learning and meta-models. Recent efforts focus on the development of deep neural network-based architectures and extending coverage to RNA, DNA and lipid-binding IDRs. We analyze availability of these methods and show that providing implementations and webservers results in much higher rates of citations/use. We also make several recommendations to take advantage of modern deep network architectures, develop tools that bundle predictions of multiple and different types of binding IDRs, and work on algorithms that model structures of the resulting complexes.

Rapid prediction and analysis of protein intrinsic disorder

Protein intrinsic disorder is found in all kingdoms of life and is known to underpin numerous physiological and pathological processes. Computational methods play an important role in characterizing and identifying intrinsically disordered proteins and protein regions. Herein, we present a new high-efficiency web-based disorder predictor named Rapid Intrinsic Disorder Analysis Online (RIDAO) that is designed to facilitate the application of protein intrinsic disorder analysis in genome-scale structural bioinformatics and comparative genomics/proteomics. RIDAO integrates six established disorder predictors into a single, unified platform that reproduces the results of individual predictors with near-perfect fidelity. To demonstrate the potential applications, we construct a test set containing more than one million sequences from one hundred organisms comprising over 420 million residues. Using this test set, we compare the efficiency and accessibility (i.e., ease of use) of RIDAO to five well-known and popular disorder predictors, namely: AUCpreD, IUPred3, metapredict V2, flDPnn, and SPOT-Disorder2. We show that RIDAO yields per-residue predictions at a rate two to six orders of magnitude greater than the other predictors and completely processes the test set in under an hour. RIDAO can be accessed free of charge at https://ridao.app.

AlphaFold, Artificial Intelligence (AI), and Allostery

AlphaFold has burst into our lives. A powerful algorithm that underscores the strength of biological sequence data and artificial intelligence (AI). AlphaFold has appended projects and research directions. The database it has been creating promises an untold number of applications with vast potential impacts that are still difficult to surmise. AI approaches can revolutionize personalized treatments and usher in better-informed clinical trials. They promise to make giant leaps toward reshaping and revamping drug discovery strategies, selecting and prioritizing combinations of drug targets. Here, we briefly overview AI in structural biology, including in molecular dynamics simulations and prediction of microbiota–human protein–protein interactions. We highlight the advancements accomplished by the deep-learning-powered AlphaFold in protein structure prediction and their powerful impact on the life sciences. At the same time, AlphaFold does not resolve the decades-long protein folding challenge, nor does it identify the folding pathways. The models that AlphaFold provides do not capture conformational mechanisms like frustration and allostery, which are rooted in ensembles, and controlled by their dynamic distributions. Allostery and signaling are properties of populations. AlphaFold also does not generate ensembles of intrinsically disordered proteins and regions, instead describing them by their low structural probabilities. Since AlphaFold generates single ranked structures, rather than conformational ensembles, it cannot elucidate the mechanisms of allosteric activating driver hotspot mutations nor of allosteric drug resistance. However, by capturing key features, deep learning techniques can use the single predicted conformation as the basis for generating a diverse ensemble.

Compositional Bias of Intrinsically Disordered Proteins and Regions and Their Predictions

Intrinsically disordered regions (IDRs) carry out many cellular functions and vary in length and placement in protein sequences. This diversity leads to variations in the underlying compositional biases, which were demonstrated for the short vs. long IDRs. We analyze compositional biases across four classes of disorder: fully disordered proteins; short IDRs; long IDRs; and binding IDRs. We identify three distinct biases: for the fully disordered proteins, the short IDRs and the long and binding IDRs combined. We also investigate compositional bias for putative disorder produced by leading disorder predictors and find that it is similar to the bias of the native disorder. Interestingly, the accuracy of disorder predictions across different methods is correlated with the correctness of the compositional bias of their predictions highlighting the importance of the compositional bias. The predictive quality is relatively low for the disorder classes with compositional bias that is the most different from the “generic” disorder bias, while being much higher for the classes with the most similar bias. We discover that different predictors perform best across different classes of disorder. This suggests that no single predictor is universally best and motivates the development of new architectures that combine models that target specific disorder classes.

Complementarity of the residue-level protein function and structure predictions in human proteins

Sequence-based predictors of the residue-level protein function and structure cover a broad spectrum of characteristics including intrinsic disorder, secondary structure, solvent accessibility and binding to nucleic acids. They were catalogued and evaluated in numerous surveys and assessments. However, methods focusing on a given characteristic are studied separately from predictors of other characteristics, while they are typically used on the same proteins. We fill this void by studying complementarity of a representative collection of methods that target different predictions using a large, taxonomically consistent, and low similarity dataset of human proteins. First, we bridge the gap between the communities that develop structure-trained vs. disorder-trained predictors of binding residues. Motivated by a recent study of the protein-binding residue predictions, we empirically find that combining the structure-trained and disorder-trained predictors of the DNA-binding and RNA-binding residues leads to substantial improvements in predictive quality. Second, we investigate whether diverse predictors generate results that accurately reproduce relations between secondary structure, solvent accessibility, interaction sites, and intrinsic disorder that are present in the experimental data. Our empirical analysis concludes that predictions accurately reflect all combinations of these relations. Altogether, this study provides unique insights that support combining results produced by diverse residue-level predictors of protein function and structure.