MOTIVATION: Phosphorylation, a post-translational modification regulated by protein kinase enzymes, plays an essential role in almost all cellular processes. Understanding how each of the nearly 500 human protein kinases selectively phosphorylates their substrates is a foundational challenge in bioinformatics and cell signaling. Although deep learning models have been a popular means to predict kinase-substrate relationships, existing models often lack interpretability and are trained on datasets skewed toward a subset of well-studied kinases. RESULTS: Here we leverage recent peptide library datasets generated to determine substrate specificity profiles of 300 serine/threonine kinases to develop an explainable Transformer model for kinase-peptide interaction prediction. The model, trained solely on primary sequences, achieved state-of-the-art performance. Its unique multitask learning paradigm built within the model enables predictions on virtually any kinase-peptide pair, including predictions on 139 kinases not used in peptide library screens. Furthermore, we employed explainable machine learning methods to elucidate the model's inner workings. Through analysis of learned embeddings at different training stages, we demonstrate that the model employs a unique strategy of substrate prediction considering both substrate motif patterns and kinase evolutionary features. SHapley Additive exPlanation (SHAP) analysis reveals key specificity determining residues in the peptide sequence. Finally, we provide a web interface for predicting kinase-substrate associations for user-defined sequences and a resource for visualizing the learned kinase-substrate associations. AVAILABILITY AND IMPLEMENTATION: All code and data are available at https://github.com/esbgkannan/Phosformer-ST. Web server is available at https://phosformer.netlify.app.

Protein language modeling is a fast-emerging deep learning method in bioinformatics with diverse applications such as structure prediction and protein design. However, application toward estimating sequence conservation for functional site prediction has not been systematically explored. Here, we present a method for the alignment-free estimation of sequence conservation using sequence embeddings generated from protein language models. Comprehensive benchmarks across publicly available protein language models reveal that ESM2 models provide the best performance to computational cost ratio for conservation estimation. Applying our method to full-length protein sequences, we demonstrate that embedding-based methods are not sensitive to the order of conserved elements-conservation scores can be calculated for multidomain proteins in a single run, without the need to separate individual domains. Our method can also identify conserved functional sites within fast-evolving sequence regions (such as domain inserts), which we demonstrate through the identification of conserved phosphorylation motifs in variable insert segments in protein kinases. Overall, embedding-based conservation analysis is a broadly applicable method for identifying potential functional sites in any full-length protein sequence and estimating conservation in an alignment-free manner. To run this on your protein sequence of interest, try our scripts at https://github.com/esbgkannan/kibby.

Protein language models, trained on millions of biologically observed sequences, generate feature-rich numerical representations of protein sequences. These representations, called sequence embeddings, can infer structure-functional properties, despite protein language models being trained on primary sequence alone. While sequence embeddings have been applied toward tasks such as structure and function prediction, applications toward alignment-free sequence classification have been hindered by the lack of studies to derive, quantify and evaluate relationships between protein sequence embeddings. Here, we develop workflows and visualization methods for the classification of protein families using sequence embedding derived from protein language models. A benchmark of manifold visualization methods reveals that Neighbor Joining (NJ) embedding trees are highly effective in capturing global structure while achieving similar performance in capturing local structure compared with popular dimensionality reduction techniques such as t-SNE and UMAP. The statistical significance of hierarchical clusters on a tree is evaluated by resampling embeddings using a variational autoencoder (VAE). We demonstrate the application of our methods in the classification of two well-studied enzyme superfamilies, phosphatases and protein kinases. Our embedding-based classifications remain consistent with and extend upon previously published sequence alignment-based classifications. We also propose a new hierarchical classification for the S-Adenosyl-L-Methionine (SAM) enzyme superfamily which has been difficult to classify using traditional alignment-based approaches. Beyond applications in sequence classification, our results further suggest NJ trees are a promising general method for visualizing high-dimensional data sets.

MOTIVATION: The human genome encodes over 500 distinct protein kinases which regulate nearly all cellular processes by the specific phosphorylation of protein substrates. While advances in mass spectrometry and proteomics studies have identified thousands of phosphorylation sites across species, information on the specific kinases that phosphorylate these sites is currently lacking for the vast majority of phosphosites. Recently, there has been a major focus on the development of computational models for predicting kinase-substrate associations. However, most current models only allow predictions on a subset of well-studied kinases. Furthermore, the utilization of hand-curated features and imbalances in training and testing datasets pose unique challenges in the development of accurate predictive models for kinase-specific phosphorylation prediction. Motivated by the recent development of universal protein language models which automatically generate context-aware features from primary sequence information, we sought to develop a unified framework for kinase-specific phosphosite prediction, allowing for greater investigative utility and enabling substrate predictions at the whole kinome level. RESULTS: We present a deep learning model for kinase-specific phosphosite prediction, termed Phosformer, which predicts the probability of phosphorylation given an arbitrary pair of unaligned kinase and substrate peptide sequences. We demonstrate that Phosformer implicitly learns evolutionary and functional features during training, removing the need for feature curation and engineering. Further analyses reveal that Phosformer also learns substrate specificity motifs and is able to distinguish between functionally distinct kinase families. Benchmarks indicate that Phosformer exhibits significant improvements compared to the state-of-the-art models, while also presenting a more generalized, unified, and interpretable predictive framework. AVAILABILITY AND IMPLEMENTATION: Code and data are available at https://github.com/esbgkannan/phosformer. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Protein kinases are among the largest druggable family of signaling proteins, involved in various human diseases, including cancers and neurodegenerative disorders. Despite their clinical relevance, nearly 30% of the 545 human protein kinases remain highly understudied. Comparative genomics is a powerful approach for predicting and investigating the functions of understudied kinases. However, an incomplete knowledge of kinase orthologs across fully sequenced kinomes severely limits the application of comparative genomics approaches for illuminating understudied kinases. Here, we introduce KinOrtho, a query- and graph-based orthology inference method that combines full-length and domain-based approaches to map one-to-one kinase orthologs across 17 thousand species. RESULTS: Using multiple metrics, we show that KinOrtho performed better than existing methods in identifying kinase orthologs across evolutionarily divergent species and eliminated potential false positives by flagging sequences without a proper kinase domain for further evaluation. We demonstrate the advantage of using domain-based approaches for identifying domain fusion events, highlighting a case between an understudied serine/threonine kinase TAOK1 and a metabolic kinase PIK3C2A with high co-expression in human cells. We also identify evolutionary fission events involving the understudied OBSCN kinase domains, further highlighting the value of domain-based orthology inference approaches. Using KinOrtho-defined orthologs, Gene Ontology annotations, and machine learning, we propose putative biological functions of several understudied kinases, including the role of TP53RK in cell cycle checkpoint(s), the involvement of TSSK3 and TSSK6 in acrosomal vesicle localization, and potential functions for the ULK4 pseudokinase in neuronal development. CONCLUSIONS: In sum, KinOrtho presents a novel query-based tool to identify one-to-one orthologous relationships across thousands of proteomes that can be applied to any protein family of interest. We exploit KinOrtho here to identify kinase orthologs and show that its well-curated kinome ortholog set can serve as a valuable resource for illuminating understudied kinases, and the KinOrtho framework can be extended to any protein-family of interest.

BACKGROUND: Protein kinases are a large family of druggable proteins that are genomically and proteomically altered in many human cancers. Kinase-targeted drugs are emerging as promising avenues for personalized medicine because of the differential response shown by altered kinases to drug treatment in patients and cell-based assays. However, an incomplete understanding of the relationships connecting genome, proteome and drug sensitivity profiles present a major bottleneck in targeting kinases for personalized medicine. RESULTS: In this study, we propose a multi-component Quantitative Structure-Mutation-Activity Relationship Tests (QSMART) model and neural networks framework for providing explainable models of protein kinase inhibition and drug response ([Formula: see text]) profiles in cell lines. Using non-small cell lung cancer as a case study, we show that interaction terms that capture associations between drugs, pathways, and mutant kinases quantitatively contribute to the response of two EGFR inhibitors (afatinib and lapatinib). In particular, protein-protein interactions associated with the JNK apoptotic pathway, associations between lung development and axon extension, and interaction terms connecting drug substructures and the volume/charge of mutant residues at specific structural locations contribute significantly to the observed [Formula: see text] values in cell-based assays. CONCLUSIONS: By integrating multi-omics data in the QSMART model, we not only predict drug responses in cancer cell lines with high accuracy but also identify features and explainable interaction terms contributing to the accuracy. Although we have tested our multi-component explainable framework on protein kinase inhibitors, it can be extended across the proteome to investigate the complex relationships connecting genotypes and drug sensitivity profiles.

BACKGROUND: Signaling proteins such as protein kinases adopt a diverse array of conformations to respond to regulatory signals in signaling pathways. Perhaps the most fundamental conformational change of a kinase is the transition between active and inactive states, and defining the conformational features associated with kinase activation is critical for selectively targeting abnormally regulated kinases in diseases. While manual examination of crystal structures have led to the identification of key structural features associated with kinase activation, the large number of kinase crystal structures (~3,500) and extensive conformational diversity displayed by the protein kinase superfamily poses unique challenges in fully defining the conformational features associated with kinase activation. Although some computational approaches have been proposed, they are typically based on a small subset of crystal structures using measurements biased towards the active site geometry. RESULTS: We utilize an unbiased informatics based machine learning approach to classify all eukaryotic protein kinase conformations deposited in the PDB. We show that the orientation of the activation segment, measured by φ, ψ, χ1, and pseudo-dihedral angles more accurately classify kinase crystal conformations than existing methods. We show that the formation of the K-E salt bridge is statistically dependent upon the activation segment orientation and identify evolutionary differences between the activation segment conformation of tyrosine and serine/threonine kinases. We provide evidence that our method can identify conformational changes associated with the binding of allosteric regulatory proteins, and show that the greatest variation in inactive structures comes from kinase group and family specific side chain orientations. CONCLUSION: We have provided the first comprehensive machine learning based classification of protein kinase active/inactive conformations, taking into account more structures and measurements than any previous classification effort. Further, our unbiased classification of inactive structures reveals residues associated with kinase functional specificity. To enable classification of new crystal structures, we have made our classifier publicly accessible through a stand-alone program housed at https://github.com/esbg/kinconform [DOI: 10.5281/zenodo.249090 ].

Cancer is a genetic disease that develops through a series of somatic mutations, a subset of which drive cancer progression. Although cancer genome sequencing studies are beginning to reveal the mutational patterns of genes in various cancers, identifying the small subset of "causative" mutations from the large subset of "non-causative" mutations, which accumulate as a consequence of the disease, is a challenge. In this article, we present an effective machine learning approach for identifying cancer-associated mutations in human protein kinases, a class of signaling proteins known to be frequently mutated in human cancers. We evaluate the performance of 11 well known supervised learners and show that a multiple-classifier approach, which combines the performances of individual learners, significantly improves the classification of known cancer-associated mutations. We introduce several novel features related specifically to structural and functional characteristics of protein kinases and find that the level of conservation of the mutated residue at specific evolutionary depths is an important predictor of oncogenic effect. We consolidate the novel features and the multiple-classifier approach to prioritize and experimentally test a set of rare unconfirmed mutations in the epidermal growth factor receptor tyrosine kinase (EGFR). Our studies identify T725M and L861R as rare cancer-associated mutations inasmuch as these mutations increase EGFR activity in the absence of the activating EGF ligand in cell-based assays.

Protein kinases play a prominent role in cell regulation and disease, which has given rise to an abundance of information about the structure, function, disease, pathway, interaction and evolution of these proteins. This information, however, is currently spread across several heterogeneous resources, an obstacle to the kind of integrative approaches needed in utilizing existing knowledge for research related to diseases. We have designed and developed an ontology for protein kinases, ProKinO, that serves as a useful and efficient representation of the integrated knowledge about these complex proteins which are intimately involved in the genesis and behavior of cancer cells. Concepts and relationships in ProKinO capture important knowledge about kinases while ProKinO instances represent a wealth of data acquired from disparate resources, including KinBase, COSMIC, UniProt, and Reactome. We have created a customized ontology browser for ProKinO. Also, we used ProKinO to do a variety of integrative analyses using SPARQL queries.

This paper provides a novel approach to use URI fragment identifiers to enable HTTP clients to address and process content, independent of its original representation.