SPOT-Peptide: Template-Based Prediction of Peptide-Binding Proteins and Peptide-Binding Sites

VISH-Pred: an ensemble of fine-tuned ESM models for protein toxicity prediction

Peptide- and protein-based therapeutics are becoming a promising treatment regimen for myriad diseases. Toxicity of proteins is the primary hurdle for protein-based therapies. Thus, there is an urgent need for accurate in silico methods for determining toxic proteins to filter the pool of potential candidates. At the same time, it is imperative to precisely identify non-toxic proteins to expand the possibilities for protein-based biologics. To address this challenge, we proposed an ensemble framework, called VISH-Pred, comprising models built by fine-tuning ESM2 transformer models on a large, experimentally validated, curated dataset of protein and peptide toxicities. The primary steps in the VISH-Pred framework are to efficiently estimate protein toxicities taking just the protein sequence as input, employing an under sampling technique to handle the humongous class-imbalance in the data and learning representations from fine-tuned ESM2 protein language models which are then fed to machine learning techniques such as Lightgbm and XGBoost. The VISH-Pred framework is able to correctly identify both peptides/proteins with potential toxicity and non-toxic proteins, achieving a Matthews correlation coefficient of 0.737, 0.716 and 0.322 and F1-score of 0.759, 0.696 and 0.713 on three non-redundant blind tests, respectively, outperforming other methods by over $10\%$ on these quality metrics. Moreover, VISH-Pred achieved the best accuracy and area under receiver operating curve scores on these independent test sets, highlighting the robustness and generalization capability of the framework. By making VISH-Pred available as an easy-to-use web server, we expect it to serve as a valuable asset for future endeavors aimed at discerning the toxicity of peptides and enabling efficient protein-based therapeutics.

Bioinformatics and bioactive peptides from foods: Do they work together?

We live in the Big Data Era which affects many aspects of science, including research on bioactive peptides derived from foods, which during the last few decades have been a focus of interest for scientists. These two issues, i.e., the development of computer technologies and progress in the discovery of novel peptides with health-beneficial properties, are closely interrelated. This Chapter presents the example applications of bioinformatics for studying biopeptides, focusing on main aspects of peptide analysis as the starting point, including: (i) the role of peptide databases; (ii) aspects of bioactivity prediction; (iii) simulation of peptide release from proteins. Bioinformatics can also be used for predicting other features of peptides, including ADMET, QSAR, structure, and taste. To answer the question asked "bioinformatics and bioactive peptides from foods: do they work together?", currently it is almost impossible to find examples of peptide research with no bioinformatics involved. However, theoretical predictions are not equivalent to experimental work and always require critical scrutiny. The aspects of compatibility of in silico and in vitro results are also summarized herein.

To Improve Prediction of Binding Residues With DNA, RNA, Carbohydrate, and Peptide Via Multi-Task Deep Neural Networks

MOTIVATION

The interactions of proteins with DNA, RNA, peptide, and carbohydrate play key roles in various biological processes. The studies of uncharacterized protein-molecules interactions could be aided by accurate predictions of residues that bind with partner molecules. However, the existing methods for predicting binding residues on proteins remain of relatively low accuracies due to the limited number of complex structures in databases. As different types of molecules partially share chemical mechanisms, the predictions for each molecular type should benefit from the binding information with other molecule types.

RESULTS

In this study, we employed a multiple task deep learning strategy to develop a new sequence-based method for simultaneously predicting binding residues/sites with multiple important molecule types named MTDsite. By combining four training sets for DNA, RNA, peptide, and carbohydrate-binding proteins, our method yielded accurate and robust predictions with AUC values of 0.852, 0836, 0.758, and 0.776 on their respective independent test sets, which are 0.52 to 6.6% better than other state-of-the-art methods. To my best knowledge, this is the first method using multi-task framework to predict multiple molecular binding sites simultaneously.

InterPepScore: A Deep Learning Score for Improving the FlexPepDock Refinement Protocol

Motivation

Interactions between peptide fragments and protein receptors are vital to cell function yet difficult to experimentally determine in structural details of. As such, many computational methods have been developed to aid in peptide–protein docking or structure prediction. One such method is Rosetta FlexPepDock which consistently refines coarse peptide–protein models into sub-Ångström precision using Monte-Carlo simulations and statistical potentials. Deep learning has recently seen increased use in protein structure prediction, with graph neural networks used for protein model quality assessment.

Results

Here, we introduce a graph neural network, InterPepScore, as an additional scoring term to complement and improve the Rosetta FlexPepDock refinement protocol. InterPepScore is trained on simulation trajectories from FlexPepDock refinement starting from thousands of peptide–protein complexes generated by a wide variety of docking schemes. The addition of InterPepScore into the refinement protocol consistently improves the quality of models created, and on an independent benchmark on 109 peptide–protein complexes its inclusion results in an increase in the number of complexes for which the top-scoring model had a DockQ-score of 0.49 (Medium quality) or better from 14.8% to 26.1%.

Availability and implementation

InterPepScore is available online at http://wallnerlab.org/InterPepScore.

Supplementary information

Supplementary data are available at Bioinformatics online.

Efficient 3D conformer generation of cyclic peptides formed by a disulfide bond

Cyclic peptides formed by disulfide bonds have been one large group of common drug candidates in drug development. Structural information of a peptide is essential to understand its interaction with its target. However, due to the high flexibility of peptides, it is difficult to sample the near-native conformations of a peptide. Here, we have developed an extended version of our MODPEP approach, named MODPEP2.0, to fast generate the conformations of cyclic peptides formed by a disulfide bond. MODPEP2.0 builds the three-dimensional (3D) structures of a cyclic peptide from scratch by assembling amino acids one by one onto the cyclic fragment based on the constructed rotamer and cyclic backbone libraries. Being tested on a data set of 193 diverse cyclic peptides, MODPEP2.0 obtained a considerable advantage in both accuracy and computational efficiency, compared with other sampling algorithms including PEP-FOLD, ETKDG, and modified ETKDG (mETKDG). MODPEP2.0 achieved a high sampling accuracy with an average C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}α RMSD of 2.20 Å and 1.66 Å when 10 and 100 conformations were considered, respectively, compared with 3.41 Å and 2.62 Å for PEP-FOLD, 3.44 Å and 3.16 Å for ETKDG, 3.09 Å and 2.72 Å for mETKDG. MODPEP2.0 also reproduced experimental peptide structures for 81.35% of the test cases when an ensemble of 100 conformations were considered, compared with 54.95%, 37.50% and 50.00% for PEP-FOLD, ETKDG, and mETKDG. MODPEP2.0 is computationally efficient and can generate 100 peptide conformations in one second. MODPEP2.0 will be useful in sampling cyclic peptide structures and modeling related protein-peptide interactions, facilitating the development of cyclic peptide drugs.

InterPepRank: Assessment of Docked Peptide Conformations by a Deep Graph Network

Peptide-protein interactions between a smaller or disordered peptide stretch and a folded receptor make up a large part of all protein-protein interactions. A common approach for modeling such interactions is to exhaustively sample the conformational space by fast-Fourier-transform docking, and then refine a top percentage of decoys. Commonly, methods capable of ranking the decoys for selection fast enough for larger scale studies rely on first-principle energy terms such as electrostatics, Van der Waals forces, or on pre-calculated statistical potentials. We present InterPepRank for peptide-protein complex scoring and ranking. InterPepRank is a machine learning-based method which encodes the structure of the complex as a graph; with physical pairwise interactions as edges and evolutionary and sequence features as nodes. The graph network is trained to predict the LRMSD of decoys by using edge-conditioned graph convolutions on a large set of peptide-protein complex decoys. InterPepRank is tested on a massive independent test set with no targets sharing CATH annotation nor 30% sequence identity with any target in training or validation data. On this set, InterPepRank has a median AUC of 0.86 for finding coarse peptide-protein complexes with LRMSD < 4Å. This is an improvement compared to other state-of-the-art ranking methods that have a median AUC between 0.65 and 0.79. When included as a selection-method for selecting decoys for refinement in a previously established peptide docking pipeline, InterPepRank improves the number of medium and high quality models produced by 80% and 40%, respectively. The InterPepRank program as well as all scripts for reproducing and retraining it are available from: http://wallnerlab.org/InterPepRank.

NLDock: a Fast Nucleic Acid-Ligand Docking Algorithm for Modeling RNA/DNA-Ligand Complexes

Nucleic acid-ligand interactions play an important role in numerous cellular processes such as gene function expression and regulation. Therefore, nucleic acids such as RNAs have become more and more important drug targets, where the structural determination of nucleic acid-ligand complexes is pivotal for understanding their functions and thus developing therapeutic interventions. Molecular docking has been a useful computational tool in predicting the complex structure between molecules. However, although a number of docking algorithms have been developed for protein-ligand interactions, only a few docking programs were presented for nucleic acid-ligand interactions. Here, we have developed a fast nucleic acid-ligand docking algorithm, named NLDock, by implementing our intrinsic scoring function ITScoreNL for nucleic acid-ligand interactions into a modified version of the MDock program. NLDock was extensively evaluated on four test sets and compared with five other state-of-the-art docking algorithms including AutoDock, DOCK 6, rDock, GOLD, and Glide. It was shown that our NLDock algorithm obtained a significantly better performance than the other docking programs in binding mode predictions and achieved the success rates of 73%, 36%, and 32% on the largest test set of 77 complexes for local rigid-, local flexible-, and global flexible-ligand docking, respectively. In addition, our NLDock approach is also computationally efficient and consumed an average of as short as 0.97 and 2.08 min for a local flexible-ligand docking job and a global flexible-ligand docking job, respectively. These results suggest the good performance of our NLDock in both docking accuracy and computational efficiency.

A deep-learning framework for multi-level peptide-protein interaction prediction

Peptide-protein interactions are involved in various fundamental cellular functions and their identification is crucial for designing efficacious peptide therapeutics. Recently, a number of computational methods have been developed to predict peptide-protein interactions. However, most of the existing prediction approaches heavily depend on high-resolution structure data. Here, we present a deep learning framework for multi-level peptide-protein interaction prediction, called CAMP, including binary peptide-protein interaction prediction and corresponding peptide binding residue identification. Comprehensive evaluation demonstrated that CAMP can successfully capture the binary interactions between peptides and proteins and identify the binding residues along the peptides involved in the interactions. In addition, CAMP outperformed other state-of-the-art methods on binary peptide-protein interaction prediction. CAMP can serve as a useful tool in peptide-protein interaction prediction and identification of important binding residues in the peptides, which can thus facilitate the peptide drug discovery process.

Protein-Peptide Binding Site Detection Using 3D Convolutional Neural Networks

Peptides and peptide-based molecules represent a promising therapeutic modality targeting intracellular protein-protein interactions, potentially combining the beneficial properties of biologics and small-molecule drugs. Protein-peptide complexes occupy a unique niche of interaction interfaces with respect to protein-protein and protein-small molecule complexes. Protein-peptide binding site identification resembles image object detection, a field that had been revolutionalized with computer vision techniques. We present a new protein-peptide binding site detection method called BiteNet_Pp by harnessing the power of 3D convolutional neural network. Our method employs a tensor-based representation of spatial protein structures, which is fed to 3D convolutional neural network, resulting in probability scores and coordinates of the binding "hot spots" in the input structures. We used the domain adaptation technique to fine-tune model trained on protein-small molecule complexes using a manually curated set of protein-peptide structures. BiteNet_Pp consistently outperforms existing state-of-the-art methods in the independent test benchmark. It takes less than a second to analyze a single-protein structure, making BiteNet_Pp suitable for the large-scale analysis of protein-peptide binding sites.

Computational approaches to predict protein functional families and functional sites

Understanding the mechanisms of protein function is indispensable for many biological applications, such as protein engineering and drug design. However, experimental annotations are sparse, and therefore, theoretical strategies are needed to fill the gap. Here, we present the latest developments in building functional subclassifications of protein superfamilies and using evolutionary conservation to detect functional determinants, for example, catalytic-, binding- and specificity-determining residues important for delineating the functional families. We also briefly review other features exploited for functional site detection and new machine learning strategies for combining multiple features.

Methods for Molecular Modelling of Protein Complexes

Biological processes are often mediated by complexes formed between proteins and various biomolecules. The 3D structures of such protein-biomolecule complexes provide insights into the molecular mechanism of their action. The structure of these complexes can be predicted by various computational methods. Choosing an appropriate method for modelling depends on the category of biomolecule that a protein interacts with and the availability of structural information about the protein and its interacting partner. We intend for the contents of this chapter to serve as a guide as to what software would be the most appropriate for the type of data at hand and the kind of 3D complex structure required. Particularly, we have dealt with protein-small molecule ligand, protein-peptide, protein-protein, and protein-nucleic acid interactions.Most, if not all, model building protocols perform some sampling and scoring. Typically, several alternate conformations and configurations of the interactors are sampled. Each such sample is then scored for optimization. To boost the confidence in these predicted models, their assessment using other independent scoring schemes besides the inbuilt/default ones would prove to be helpful. This chapter also lists such software and serves as a guide to gauge the fidelity of modelled structures of biomolecular complexes.

ITScore-NL: An Iterative Knowledge-Based Scoring Function for Nucleic Acid-Ligand Interactions

Nucleic acid-ligand complexes underlie numerous cellular processes, such as gene function expression and regulation, in which their three-dimensional structures are important to understand their functions and thus to develop therapeutic interventions. Given the high cost and technical difficulties in experimental methods, computational methods such as molecular docking have been actively used to investigate nucleic acid-ligand interactions in which an accurate scoring function is crucial. However, because of the limited number of experimental nucleic acid-ligand binding data and structures, the scoring function development for nucleic acid-ligand interactions falls far behind that for protein-protein and protein-ligand interactions. Here, based on our statistical mechanics-based iterative approach, we have developed an iterative knowledge-based scoring function for nucleic acid-ligand interactions, named as ITScore-NL, by explicitly including stacking and electrostatic potentials. Our ITScore-NL scoring function was extensively evaluated for its ability in the binding mode and binding affinity predictions on three diverse test sets and compared with state-of-the-art scoring functions. Overall, ITScore-NL obtained significantly better performance than the other 12 scoring functions and predicted near-native poses with rmsd ≤ 1.5 Å for 71.43% of the cases when the top three binding modes were considered and a good correlation of R = 0.64 in binding affinity prediction on the large test set of 77 nucleic acid-ligand complexes. These results suggested the accuracy of ITScore-NL and the necessity of explicitly including stacking and electrostatic potentials.

InterPep2: global peptide-protein docking using interaction surface templates

Motivation

Interactions between proteins and peptides or peptide-like intrinsically disordered regions are involved in many important biological processes, such as gene expression and cell life-cycle regulation. Experimentally determining the structure of such interactions is time-consuming and difficult because of the inherent flexibility of the peptide ligand. Although several prediction-methods exist, most are limited in performance or availability.

Results

InterPep2 is a freely available method for predicting the structure of peptide–protein interactions. Improved performance is obtained by using templates from both peptide–protein and regular protein–protein interactions, and by a random forest trained to predict the DockQ-score for a given template using sequence and structural features. When tested on 252 bound peptide–protein complexes from structures deposited after the complexes used in the construction of the training and templates sets of InterPep2, InterPep2-Refined correctly positioned 67 peptides within 4.0 Å LRMSD among top10, similar to another state-of-the-art template-based method which positioned 54 peptides correctly. However, InterPep2 displays a superior ability to evaluate the quality of its own predictions. On a previously established set of 27 non-redundant unbound-to-bound peptide–protein complexes, InterPep2 performs on-par with leading methods. The extended InterPep2-Refined protocol managed to correctly model 15 of these complexes within 4.0 Å LRMSD among top10, without using templates from homologs. In addition, combining the template-based predictions from InterPep2 with ab initio predictions from PIPER-FlexPepDock resulted in 22% more near-native predictions compared to the best single method (22 versus 18).

Availability and implementation

The program is available from: http://wallnerlab.org/InterPep2.

Supplementary information

Supplementary data are available at Bioinformatics online.

Improving Protein-Peptide Docking Results via Pose-Clustering and Rescoring with a Combined Knowledge-Based and MM-GBSA Scoring Function

Protein-peptide docking, which predicts the complex structure between a protein and a peptide, is a valuable computational tool in peptide therapeutics development and the mechanistic investigation of peptides involved in cellular processes. Although current peptide docking approaches are often able to sample near-native peptide binding modes, correctly identifying those near-native modes from decoys is still challenging because of the extremely high complexity of the peptide binding energy landscape. In this study, we have developed an efficient postdocking rescoring protocol using a combined scoring function of knowledge-based ITScorePP potentials and physics-based MM-GBSA energies. Tested on five benchmark/docking test sets, our postdocking strategy showed an overall significantly better performance in binding mode prediction and score-rmsd correlation than original docking approaches. Specifically, our postdocking protocol outperformed original docking approaches with success rates of 15.8 versus 10.5% for pepATTRACT on the Global_57 benchmark, 5.3 versus 5.3% for CABS-dock on the Global_57 benchmark, 17.0 versus 11.3% for FlexPepDock on the LEADS-PEP data set, 40.3 versus 33.9% for HPEPDOCK on the Local_62 benchmark, and 64.2 versus 52.8% for HPEPDOCK on the LEADS-PEP data set when the top prediction was considered. These results demonstrated the efficacy and robustness of our postdocking protocol.