DeepPhoPred: Accurate Deep Learning Model to Predict Microbial Phosphorylation.

Phosphorylation of proteins is one of the most significant post-translational modifications (PTMs) and plays a crucial role in plant functionality due to its impact on signaling, gene expression, enzyme kinetics, protein stability and interactions. Accurate prediction of plant phosphorylation sites (p-sites) is vital as abnormal regulation of phosphorylation usually leads to plant diseases. However, current experimental methods for PTM prediction suffers from high-computational cost and are error-prone. The present study develops machine learning-based prediction techniques, including a high-performance interpretable deep tabular learning network (TabNet) to improve the prediction of protein p-sites in soybean. Moreover, we use a hybrid feature set of sequential-based features, physicochemical properties and position-specific scoring matrices to predict serine (Ser/S), threonine (Thr/T) and tyrosine (Tyr/Y) p-sites in soybean for the first time. The experimentally verified p-sites data of soybean proteins are collected from the eukaryotic phosphorylation sites database and database post-translational modification. We then remove the redundant set of positive and negative samples by dropping protein sequences with >40% similarity. It is found that the developed techniques perform >70% in terms of accuracy. The results demonstrate that the TabNet model is the best performing classifier using hybrid features and with window size of 13, resulted in 78.96 and 77.24% sensitivity and specificity, respectively. The results indicate that the TabNet method has advantages in terms of high-performance and interpretability. The proposed technique can automatically analyze the data without any measurement errors and any human intervention. Furthermore, it can be used to predict putative protein p-sites in plants effectively. The collected dataset and source code are publicly deposited at https://github.com/Elham-khalili/Soybean-P-sites-Prediction.

Phosphorylation ranks among the most crucial post-translational modifications (PTMs), significantly influencing the conformation, activity, and functionality of proteins, and is intricately associated with numerous pathophysiological processes. Therefore, proposing a scientifically valid computational method for precise prediction of phosphorylation sites carries substantial importance. In this study, we propose and evaluate a novel method, PEGN-PSP, based on graph techniques and language models for the prediction of general phosphorylation sites. This method employs an adaptive feature fusion strategy, combining sequence embeddings and pretrained model embeddings to enhance the model's capability in representing features. The use of graph neural network attention mechanisms not only captures local patterns of protein sequences, but also effectively captures long-range dependency information between residues in protein sequences. Independent test results indicate that, in comparison to the current state-of-the-art methods for general phosphorylation site prediction, PEGN-PSP improves the Matthews correlation coefficient for S/T sites by 4.6% and for Y sites by 6.5%. Additionally, PEGN-PSP has good robustness in predicting lysine crotonylation sites, indicating that our method PEGN-PSP has strong potential in predicting other protein post-translational modification sites.

Protein phosphorylation is one of the most critical post-translational modifications of proteins in eukaryotes, which is essential for a variety of biological processes. Plenty of attempts have been made to improve the performance of computational predictors for phosphorylation site prediction. However, most of them are based on extra domain knowledge or feature selection. In this article, we present a novel deep learning-based predictor, named TransPhos, which is constructed using a transformer encoder and densely connected convolutional neural network blocks, for predicting phosphorylation sites. Data experiments are conducted on the datasets of PPA (version 3.0) and Phospho. ELM. The experimental results show that our TransPhos performs better than several deep learning models, including Convolutional Neural Networks (CNN), Long-term and short-term memory networks (LSTM), Recurrent neural networks (RNN) and Fully connected neural networks (FCNN), and some state-of-the-art deep learning-based prediction tools, including GPS2.1, NetPhos, PPRED, Musite, PhosphoSVM, SKIPHOS, and DeepPhos. Our model achieves a good performance on the training datasets of Serine (S), Threonine (T), and Tyrosine (Y), with AUC values of 0.8579, 0.8335, and 0.6953 using 10-fold cross-validation tests, respectively, and demonstrates that the presented TransPhos tool considerably outperforms competing predictors in general protein phosphorylation site prediction.

DeepPPSite: A deep learning-based model for analysis and prediction of phosphorylation sites using efficient sequence information

GPS: a novel group-based phosphorylation predicting and scoring method

In this paper, we propose several active learning strategies to train classifiers for phosphorylation site prediction. When combined with support vector machine, we show that active learning with SVM is able to produce classifiers that give comparable or better phosphorylation site prediction performance than conventional SVM techniques and, at the same time, require a significantly less number of annotated protein training samples. The result has both conceptual and practical implications in protein prediction: it exploits information inherent in the large scale database of non-annotated protein samples and reduces the amount of manual labor required for protein annotation. To the best of our knowledge, active learning has not been explored in phosphorylation sites prediction. Several active learning strategies: single-running mode, batch-running mode with sample and support vector diversity, were investigated for phosphorylation sites prediction in this work. Our experiments have shown that active learning with SVM is able to reduce the effort of protein annotation by 6.6% to 25.7% to yield similar prediction performance as compared with conventional SVM technique.

Accurate identificationof phosphorylation sites is fundamentalfor advancing the study of protein phosphorylation, which plays acritical role in elucidating protein function and facilitating rationaldrug design. While mass spectrometry-based experimental techniquesare considered the effective method for phosphorylation site identification,their widespread application is often hindered by high costs, limitedthroughput, and the requirement for specialized equipment. Computationalprediction methods have become a popular alternative strategy forproviding candidate phosphorylation sites by inferring their presencefrom the entire protein sequence. However, many current methods struggleto extract sufficient contextual features, limiting their predictivereliability in complex biological settings. To address these issues,we present a novel protein large language model (pLLM)-based model,called ProtPSP, for phosphorylation site prediction. By leveraginga pretrained pLLM, ProtPSP effectively captures complex sequence contextsthat conventional models trained on limited phosphorylation data maymiss. Comprehensive benchmarking on curated data sets demonstratesthat ProtPSP achieves reliable phosphorylation site prediction forboth serine/threonine and tyrosine sites, outperforming other commonlyused methods across multiple evaluation metrics. Ablation studiessubstantiate the critical contribution of both pLLM-driven featuresand the fusion model architecture to overall performance improvements.Moreover, case studies demonstrate that ProtPSP consistently identifiesall true phosphorylation sites, underscoring its significant potentialas a complementary approach to mass spectrometry-based techniquesin biomedical research and drug discovery.

Protein phosphorylation is one of the most pervasive post-translational modifications and regulates diverse cellular processes in organisms. Under the catalysis of protein kinases, protein phosphorylation usually occurred in the residues serine (S), threonine (T), or tyrosine (Y). In this contribution, we proposed a novel scheme (named KMPhos) for the theoretical prediction of protein phosphorylation sites. First, the numerical matrix was obtained from a protein sequence fragment by replacing the characters of the residues with the chemical descriptors of amino acid molecules to approximately describe the chemical environment of the protein fragment, which was turned to the grayscale image. Then the Krawtchouk image moments were calculated and used to establish the support vector machine models. The accuracies of 10-fold cross validation for the obtained models on the training set are up to 89.7%, 88.6%, and 90.1% for the residues S, Y, and T, respectively. For the independent test set, the prediction accuracies are up to 90.7% (S), 87.8% (T), and 89.3% (Y). The results of ROC and other evaluations are also satisfactory. Compared with several specialized prediction tools, KMPhos provided the higher accuracy and reliability. An available KMPhos package is provided and can be used directly for phosphorylation sites prediction.

Reversible protein phosphorylation is one of the most important post-translational modifications, which regulates various biological cellular processes. Identification of the kinase-specific phosphorylation sites is helpful for understanding the phosphorylation mechanism and regulation processes. Although a number of computational approaches have been developed, currently few studies are concerned about hierarchical structures of kinases, and most of the existing tools use only local sequence information to construct predictive models. In this work, we conduct a systematic and hierarchy-specific investigation of protein phosphorylation site prediction in which protein kinases are clustered into hierarchical structures with four levels including kinase, subfamily, family and group. To enhance phosphorylation site prediction at all hierarchical levels, functional information of proteins, including gene ontology (GO) and protein-protein interaction (PPI), is adopted in addition to primary sequence to construct prediction models based on random forest. Analysis of selected GO and PPI features shows that functional information is critical in determining protein phosphorylation sites for every hierarchical level. Furthermore, the prediction results of Phospho.ELM and additional testing dataset demonstrate that the proposed method remarkably outperforms existing phosphorylation prediction methods at all hierarchical levels. The proposed method is freely available at http://bioinformatics.ustc.edu.cn/phos_pred/.

Defining Human Tyrosine Kinase Phosphorylation Networks Using Yeast as an In Vivo Model Substrate.

Signaling by growth factor receptor tyrosine kinases is manifest through networks of proteins that are substrates and/or bind to the activated receptors. FGF receptor-3 (FGFR3) is a drug target in a subset of human multiple myelomas (MM) and is mutationally activated in some cervical and colon and many bladder cancers and in certain skeletal dysplasias. To define the FGFR3 network in multiple myeloma, mass spectrometry was used to identify and quantify phosphotyrosine (pY) sites modulated by FGFR3 activation and inhibition in myeloma-derived KMS11 cells. Label-free quantification of peptide ion currents indicated the activation of FGFR3 by phosphorylation of tandem tyrosines in the kinase domain activation loop when cellular pY phosphatases were inhibited by pervanadate. Among the 175 proteins that accumulated pY in response to pervanadate was a subset of 52 including FGFR3 that contained a total of 61 pY sites that were sensitive to inhibition by the FGFR3 inhibitor PD173074. The FGFR3 isoform containing the tandem pY motif in its activation loop was targeted by PD173074. Forty of the drug-sensitive pY sites, including two located within the 35-residue cytoplasmic domain of the transmembrane growth factor binding proteoglycan (and multiple myeloma biomarker) Syndecan-1/CD138, were also stimulated in cells treated with the ligand FGF1, providing additional validation of their link to FGFR3. The identification of these overlapping sets of co-modulated tyrosine phosphorylations presents an outline of an FGFR3 network in the MM model and demonstrates the potential for pharmacodynamic monitoring by label-free quantitative phospho-proteomics.

Protein phosphorylation, catalyzed by protein kinases (PKs), is one of the most important post-translational modifications (PTMs), and involved in regulating almost all of biological processes. Here, we report an updated server, Group-based Prediction System (GPS) 6.0, for prediction of PK-specific phosphorylation sites (p-sites) in eukaryotes. First, we pre-trained a general model using penalized logistic regression (PLR), deep neural network (DNN), and Light Gradient Boosting Machine (LightGMB) on 490 762 non-redundant p-sites in 71 407 proteins. Then, transfer learning was conducted to obtain 577 PK-specific predictors at the group, family and single PK levels, using a well-curated data set of 30 043 known site-specific kinase-substrate relations in 7041 proteins. Together with the evolutionary information, GPS 6.0 could hierarchically predict PK-specific p-sites for 44046 PKs in 185 species. Besides the basic statistics, we also offered the knowledge from 22 public resources to annotate the prediction results, including the experimental evidence, physical interactions, sequence logos, and p-sites in sequences and 3D structures. The GPS 6.0 server is freely available at https://gps.biocuckoo.cn. We believe that GPS 6.0 could be a highly useful service for further analysis of phosphorylation.

BackgroundProtein phosphorylation catalyzed by kinases plays crucial regulatory roles in cellular processes. Given the high-throughput mass spectrometry-based experiments, the desire to annotate the catalytic kinases for in vivo phosphorylation sites has motivated. Thus, a variety of computational methods have been developed for performing a large-scale prediction of kinase-specific phosphorylation sites. However, most of the proposed methods solely rely on the local amino acid sequences surrounding the phosphorylation sites. An increasing number of three-dimensional structures make it possible to physically investigate the structural environment of phosphorylation sites.ResultsIn this work, all of the experimental phosphorylation sites are mapped to the protein entries of Protein Data Bank by sequence identity. It resulted in a total of 4508 phosphorylation sites containing the protein three-dimensional (3D) structures. To identify phosphorylation sites on protein 3D structures, this work incorporates support vector machines (SVMs) with the information of linear motifs and spatial amino acid composition, which is determined for each kinase group by calculating the relative frequencies of 20 amino acid types within a specific radial distance from central phosphorylated amino acid residue. After the cross-validation evaluation, most of the kinase-specific models trained with the consideration of structural information outperform the models considering only the sequence information. Furthermore, the independent testing set which is not included in training set has demonstrated that the proposed method could provide a comparable performance to other popular tools.ConclusionThe proposed method is shown to be capable of predicting kinase-specific phosphorylation sites on 3D structures and has been implemented as a web server which is freely accessible at http://csb.cse.yzu.edu.tw/PhosK3D/. Due to the difficulty of identifying the kinase-specific phosphorylation sites with similar sequenced motifs, this work also integrates the 3D structural information to improve the cross classifying specificity.

Protein phosphorylation represents a ubiquitous control mechanism in living cells. The structural prerequisites and consequences of this important post-translational modification, however, are poorly understood. Oncoprotein 18/stathmin (Op18) is a globally disordered phosphoprotein that is involved in the regulation of the microtubule (MT) filament system. Here we document that phosphorylation of Ser63, which is located within a helix initiation site in Op18, disrupts the transiently formed amphipathic helix. The phosphoryl group reduces tubulin binding 10-fold and suppresses the MT polymerization inhibition activity of Op18's C-terminal domain. Op18 represents an example where phosphorylation occurs within a regular secondary structural element. Together, our findings have implications for the prediction of phosphorylation sites and give insights into the molecular behavior of a globally disordered protein.

The ERK1/2 MAP kinase pathway is an evolutionarily conserved signaling module that controls many fundamental physiological processes. Deregulated activity of ERK1/2 MAP kinases is associated with developmental syndromes and several human diseases. Despite the importance of this pathway, a comprehensive picture of the natural substrate repertoire and biochemical mechanisms regulated by ERK1/2 is still lacking. In this study, we used large-scale quantitative phosphoproteomics and bioinformatics analyses to identify novel candidate ERK1/2 substrates based on their phosphorylation signature and kinetic profiles in epithelial cells. We identified a total of 7936 phosphorylation sites within 1861 proteins, of which 155 classify as candidate ERK1/2 substrates, including 128 new targets. Candidate ERK1/2 substrates are involved in diverse cellular processes including transcriptional regulation, chromatin remodeling, RNA splicing, cytoskeleton dynamics, cellular junctions and cell signaling. Detailed characterization of one newly identified substrate, the transcriptional regulator JunB, revealed that ERK1/2 phosphorylate JunB on a serine adjacent to the DNA-binding domain, resulting in increased DNA-binding affinity and transcriptional activity. Our study expands the spectrum of cellular functions controlled by ERK1/2 kinases.