ProKinO: an ontology for integrative analysis of protein kinases in cancer

Informatic challenges and advances in illuminating the druggable proteome

The understudied members of the druggable proteomes offer promising prospects for drug discovery efforts. While large-scale initiatives have generated valuable functional information on understudied members of the druggable gene families, translating this information into actionable knowledge for drug discovery requires specialized informatics tools and resources. Here, we review the unique informatics challenges and advances in annotating understudied members of the druggable proteome. We demonstrate the application of statistical evolutionary inference tools, knowledge graph mining approaches, and protein language models in illuminating understudied protein kinases, pseudokinases, and ion channels.

Dark kinase annotation, mining, and visualization using the Protein Kinase Ontology

The Protein Kinase Ontology (ProKinO) is an integrated knowledge graph that conceptualizes the complex relationships among protein kinase sequence, structure, function, and disease in a human and machine-readable format. In this study, we have significantly expanded ProKinO by incorporating additional data on expression patterns and drug interactions. Furthermore, we have developed a completely new browser from the ground up to render the knowledge graph visible and interactive on the web. We have enriched ProKinO with new classes and relationships that capture information on kinase ligand binding sites, expression patterns, and functional features. These additions extend ProKinO's capabilities as a discovery tool, enabling it to uncover novel insights about understudied members of the protein kinase family. We next demonstrate the application of ProKinO. Specifically, through graph mining and aggregate SPARQL queries, we identify the p21-activated protein kinase 5 (PAK5) as one of the most frequently mutated dark kinases in human cancers with abnormal expression in multiple cancers, including a previously unappreciated role in acute myeloid leukemia. We have identified recurrent oncogenic mutations in the PAK5 activation loop predicted to alter substrate binding and phosphorylation. Additionally, we have identified common ligand/drug binding residues in PAK family kinases, underscoring ProKinO's potential application in drug discovery. The updated ontology browser and the addition of a web component, ProtVista, which enables interactive mining of kinase sequence annotations in 3D structures and Alphafold models, provide a valuable resource for the signaling community. The updated ProKinO database is accessible at https://prokino.uga.edu.

KinFams: De-Novo Classification of Protein Kinases Using CATH Functional Units

Protein kinases are important targets for treating human disorders, and they are the second most targeted families after G-protein coupled receptors. Several resources provide classification of kinases into evolutionary families (based on sequence homology); however, very few systematically classify functional families (FunFams) comprising evolutionary relatives that share similar functional properties. We have developed the FunFam-MARC (Multidomain ARchitecture-based Clustering) protocol, which uses multi-domain architectures of protein kinases and specificity-determining residues for functional family classification. FunFam-MARC predicts 2210 kinase functional families (KinFams), which have increased functional coherence, in terms of EC annotations, compared to the widely used KinBase classification. Our protocol provides a comprehensive classification for kinase sequences from >10,000 organisms. We associate human KinFams with diseases and drugs and identify 28 druggable human KinFams, i.e., enriched in clinically approved drugs. Since relatives in the same druggable KinFam tend to be structurally conserved, including the drug-binding site, these KinFams may be valuable for shortlisting therapeutic targets. Information on the human KinFams and associated 3D structures from AlphaFold2 are provided via our CATH FTP website and Zenodo. This gives the domain structure representative of each KinFam together with information on any drug compounds available. For 32% of the KinFams, we provide information on highly conserved residue sites that may be associated with specificity.

Recent advances in targeting protein kinases and pseudokinases in cancer biology

Kinases still remain the most favorable members of the druggable genome, and there are an increasing number of kinase inhibitors approved by the FDA to treat a variety of cancers. Here, we summarize recent developments in targeting kinases and pseudokinases with some examples. Targeting the cell cycle machinery garnered significant clinical success, however, a large section of the kinome remains understudied. We also review recent developments in the understanding of pseudokinases and discuss approaches on how to effectively target in cancer.

Computational tools and resources for pseudokinase research

Pseudokinases regulate diverse cellular processes associated with normal cellular functions and disease. They are defined bioinformatically based on the absence of one or more catalytic residues that are required for canonical protein kinase functions. The ability to define pseudokinases based on primary sequence comparison has enabled the systematic mapping and cataloging of pseudokinase orthologs across the tree of life. While these sequences contain critical information regarding pseudokinase evolution and functional specialization, extracting this information and generating testable hypotheses based on integrative mining of sequence and structural data requires specialized computational tools and resources. In this chapter, we review recent advances in the development and application of open-source tools and resources for pseudokinase research. Specifically, we describe the application of an interactive data analytics framework, KinView, for visualizing the patterns of conservation and variation in the catalytic domain motifs of pseudokinases and evolutionarily related canonical kinases using a consistent set of curated alignments organized based on the widely used kinome evolutionary hierarchy. We also demonstrate the application of an integrated Protein Kinase Ontology (ProKinO) and an interactive viewer, ProtVista, for mapping and analyzing primary sequence motifs and annotations in the context of 3D structures and AlphaFold2 models. We provide examples and protocols for generating testable hypotheses on pseudokinase functions both for bench biologists and advanced users.

Emerging concepts in pseudoenzyme classification, evolution, and signaling

The 21st century is witnessing an explosive surge in our understanding of pseudoenzyme-driven regulatory mechanisms in biology. Pseudoenzymes are proteins that have sequence homology with enzyme families but that are proven or predicted to lack enzyme activity due to mutations in otherwise conserved catalytic amino acids. The best-studied pseudoenzymes are pseudokinases, although examples from other families are emerging at a rapid rate as experimental approaches catch up with an avalanche of freely available informatics data. Kingdom-wide analysis in prokaryotes, archaea and eukaryotes reveals that between 5 and 10% of proteins that make up enzyme families are pseudoenzymes, with notable expansions and contractions seemingly associated with specific signaling niches. Pseudoenzymes can allosterically activate canonical enzymes, act as scaffolds to control assembly of signaling complexes and their localization, serve as molecular switches, or regulate signaling networks through substrate or enzyme sequestration. Molecular analysis of pseudoenzymes is rapidly advancing knowledge of how they perform noncatalytic functions and is enabling the discovery of unexpected, and previously unappreciated, functions of their intensively studied enzyme counterparts. Notably, upon further examination, some pseudoenzymes have previously unknown enzymatic activities that could not have been predicted a priori. Pseudoenzymes can be targeted and manipulated by small molecules and therefore represent new therapeutic targets (or anti-targets, where intervention should be avoided) in various diseases. In this review, which brings together broad bioinformatics and cell signaling approaches in the field, we highlight a selection of findings relevant to a contemporary understanding of pseudoenzyme-based biology.

The dynamic switch mechanism that leads to activation of LRRK2 is embedded in the DFGψ motif in the kinase domain

Little is known about the regulation of Leucine-rich repeat kinase 2 (LRRK2) associated with familial Parkinson’s disease (PD). To test whether the kinase domain drives LRRK2 activation, we applied the spine concept that describes the core architecture of every protein kinase. We discovered that mutation of Y2018, a regulatory spine residue, to Phe in the DFGψ motif created a hyperactive kinase similar to the PD-associated mutation G2019S. The hydroxyl moiety of Y2018 thus serves as a “brake,” stabilizing the inactive conformation; simply removing it destroys a key inhibitory hydrogen-bonding node. These data reveal an LRRK2-specific regulatory mechanism, confirming that the kinase domain functions as a classical kinase that controls overall conformational dynamics in full-length LRRK2 and drives therapeutic strategies.

Leucine-rich repeat kinase 2 (LRRK2) is a large multidomain protein, and LRRK2 mutants are recognized risk factors for Parkinson’s disease (PD). Although the precise mechanisms that control LRRK2 regulation and function are unclear, the importance of the kinase domain is strongly implicated, since 2 of the 5 most common familial LRRK2 mutations (G2019S and I2020T) are localized to the conserved DFGψ motif in the kinase core, and kinase inhibitors are under development. Combining the concept of regulatory (R) and catalytic (C) spines with kinetic and cell-based assays, we discovered a major regulatory mechanism embedded within the kinase domain and show that the DFG motif serves as a conformational switch that drives LRRK2 activation. LRRK2 is quite unusual in that the highly conserved Phe in the DFGψ motif, which is 1 of the 4 R-spine residues, is replaced with tyrosine (DY²⁰¹⁸GI). A Y2018F mutation creates a hyperactive phenotype similar to the familial mutation G2019S. The hydroxyl moiety of Y2018 thus serves as a “brake” that stabilizes an inactive conformation; simply removing it destroys a key hydrogen-bonding node. Y2018F, like the pathogenic mutant I2020T, spontaneously forms LRRK2-decorated microtubules in cells, while the wild type and G2019S require kinase inhibitors to form filaments. We also explored 3 different mechanisms that create kinase-dead pseudokinases, including D2017A, which further emphasizes the highly synergistic role of key hydrophobic and hydrophilic/charged residues in the assembly of active LRRK2. We thus hypothesize that LRRK2 harbors a classical protein kinase switch mechanism that drives the dynamic activation of full-length LRRK2.

Advances in kinome research of parasitic worms - implications for fundamental research and applied biotechnological outcomes

Protein kinases are enzymes that play essential roles in the regulation of many cellular processes. Despite expansions in the fields of genomics, transcriptomics and bioinformatics, there is limited information on the kinase complements (kinomes) of most eukaryotic organisms, including parasitic worms that cause serious diseases of humans and animals. The biological uniqueness of these worms and the draft status of their genomes pose challenges for the identification and classification of protein kinases using established tools. In this article, we provide an account of kinase biology, the roles of kinases in diseases and their importance as drug targets, and drug discovery efforts in key socioeconomically important parasitic worms. In this context, we summarise methods and resources commonly used for the curation, identification, classification and functional annotation of protein kinase sequences from draft genomes; review recent advances made in the characterisation of the worm kinomes; and discuss the implications of these advances for investigating kinase signalling and developing small-molecule inhibitors as new anti-parasitic drugs.

Classifying kinase conformations using a machine learning approach

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].

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-017-1506-2) contains supplementary material, which is available to authorized users.

Computational and Experimental Characterization of Patient Derived Mutations Reveal an Unusual Mode of Regulatory Spine Assembly and Drug Sensitivity in EGFR Kinase

The catalytic activation of protein kinases requires precise positioning of key conserved catalytic and regulatory motifs in the kinase core. The Regulatory Spine (RS) is one such structural motif that is dynamically assembled upon kinase activation. The RS is also a mutational hotspot in cancers; however, the mechanisms by which cancer mutations impact RS assembly and kinase activity are not fully understood. In this study, through mutational analysis of patient derived mutations in the RS of EGFR kinase, we identify an activating mutation, M766T, at the RS3 position. RS3 is located in the regulatory αC-helix, and a series of mutations at the RS3 position suggest a strong correlation between the amino acid type present at the RS3 position and ligand (EGF) independent EGFR activation. Small polar amino acids increase ligand independent activity, while large aromatic amino acids decrease kinase activity. M766T relies on the canonical asymmetric dimer for full activation. Molecular modeling and molecular dynamics simulations of WT and mutant EGFR suggest a model in which M766T activates the kinase domain by disrupting conserved autoinhibitory interactions between M766 and hydrophobic residues in the activation segment. In addition, a water mediated hydrogen bond network between T766, the conserved K745-E762 salt bridge, and the backbone amide of the DFG motif is identified as a key determinant of M766T-mediated activation. M766T is resistant to FDA approved EGFR inhibitors such as gefitinib and erlotinib, and computational estimation of ligand binding free energy identifies key residues associated with drug sensitivity. In sum, our studies suggest an unusual mode of RS assembly and oncogenic EGFR activation, and provide new clues for the design of allosteric protein kinase inhibitors.

KinView: a visual comparative sequence analysis tool for integrated kinome research

Multiple sequence alignments (MSAs) are a fundamental analysis tool used throughout biology to investigate relationships between protein sequence, structure, function, evolutionary history, and patterns of disease-associated variants. However, their widespread application in systems biology research is currently hindered by the lack of user-friendly tools to simultaneously visualize, manipulate and query the information conceptualized in large sequence alignments, and the challenges in integrating MSAs with multiple orthogonal data such as cancer variants and post-translational modifications, which are often stored in heterogeneous data sources and formats. Here, we present the Multiple Sequence Alignment Ontology (MSAOnt), which represents a profile or consensus alignment in an ontological format. Subsets of the alignment are easily selected through the SPARQL Protocol and RDF Query Language for downstream statistical analysis or visualization. We have also created the Kinome Viewer (KinView), an interactive integrative visualization that places eukaryotic protein kinase cancer variants in the context of natural sequence variation and experimentally determined post-translational modifications, which play central roles in the regulation of cellular signaling pathways. Using KinView, we identified differential phosphorylation patterns between tyrosine and serine/threonine kinases in the activation segment, a major kinase regulatory region that is often mutated in proliferative diseases. We discuss cancer variants that disrupt phosphorylation sites in the activation segment, and show how KinView can be used as a comparative tool to identify differences and similarities in natural variation, cancer variants and post-translational modifications between kinase groups, families and subfamilies. Based on KinView comparisons, we identify and experimentally characterize a regulatory tyrosine (Y177PLK4) in the PLK4 C-terminal activation segment region termed the P+1 loop. To further demonstrate the application of KinView in hypothesis generation and testing, we formulate and validate a hypothesis explaining a novel predicted loss-of-function variant (D523NPKCβ) in the regulatory spine of PKCβ, a recently identified tumor suppressor kinase. KinView provides a novel, extensible interface for performing comparative analyses between subsets of kinases and for integrating multiple types of residue specific annotations in user friendly formats.

Multi-kinase inhibitors, AURKs and cancer

Inhibitors that impact function of kinases are valuable both for the biological research as well as therapy of kinase-associated diseases, such as different cancers. There are quite a number of inhibitors, which are quite specific for certain kinases and several of them are either already approved for the cancer therapy or are in clinical studies of various phases. However, that does not mean that each single kinase inhibitor is suitable for targeted therapy. Some of them are not effective others might be toxic or fail some other criteria for the use in vivo. On the other hand, even in case of successful therapy, many responders eventually develop resistance to the inhibitors. The limitations of various single kinase inhibitors can be fought using compounds which target multiple kinases. This tactics can increase effectiveness of the inhibitors by the synergistic effect or help to diminish the likelihood of drug resistance. To date, several families of kinases are quite popular targets of the inhibition in cancers, such as tyrosine kinases, cycle-dependent kinases, mitogen-activated protein kinases, phosphoinositide 3-kinases as well as their pathway "players" and aurora kinases. Aurora kinases play an important role in the control of the mitosis and are often altered in diverse human cancers. Here, we will describe the most interesting multi-kinase inhibitors which inhibit aurora kinases among other targets and their use in preclinical and clinical cancer studies.

ProKinO: a unified resource for mining the cancer kinome

Protein kinases represent a large and diverse family of evolutionarily related proteins that are abnormally regulated in human cancers. Although genome sequencing studies have revealed thousands of variants in protein kinases, translating "big" genomic data into biological knowledge remains a challenge. Here, we describe an ontological framework for integrating and conceptualizing diverse forms of information related to kinase activation and regulatory mechanisms in a machine readable, human understandable form. We demonstrate the utility of this framework in analyzing the cancer kinome, and in generating testable hypotheses for experimental studies. Through the iterative process of aggregate ontology querying, hypothesis generation and experimental validation, we identify a novel mutational hotspot in the αC-β4 loop of the kinase domain and demonstrate the functional impact of the identified variants in epidermal growth factor receptor (EGFR) constitutive activity and inhibitor sensitivity. We provide a unified resource for the kinase and cancer community, ProKinO, housed at http://vulcan.cs.uga.edu/prokino.

The Tribbles 2 (TRB2) pseudokinase binds to ATP and autophosphorylates in a metal-independent manner

The human Tribbles (TRB)-related pseudokinases are CAMK (calcium/calmodulin-dependent protein kinase)-related family members that have evolved a series of highly unusual motifs in the 'pseudocatalytic' domain. In canonical kinases, conserved amino acids bind to divalent metal ions and align ATP prior to efficient phosphoryl-transfer to substrates. However, in pseudokinases, atypical residues give rise to diverse and often unstudied biochemical and structural features that are thought to be central to cellular functions. TRB proteins play a crucial role in multiple signalling networks and overexpression confers cancer phenotypes on human cells, marking TRB pseudokinases out as a novel class of drug target. In the present paper, we report that the human pseudokinase TRB2 retains the ability to both bind and hydrolyse ATP weakly in vitro. Kinase activity is metal-independent and involves a catalytic lysine residue, which is conserved in TRB proteins throughout evolution alongside several unique amino acids in the active site. A similar low level of autophosphorylation is also preserved in the closely related human TRB3. By employing chemical genetics, we establish that the nucleotide-binding site of an 'analogue-sensitive' (AS) TRB2 mutant can be targeted with specific bulky ligands of the pyrazolo-pyrimidine (PP) chemotype. Our analysis confirms that TRB2 retains low levels of ATP binding and/or catalysis that is targetable with small molecules. Given the significant clinical successes associated with targeting of cancer-associated kinases with small molecule inhibitors, it is likely that similar approaches will be useful for further evaluating the TRB pseudokinases, with the translation of this information likely to furnish new leads for drug discovery.

Integration of signaling in the kinome: Architecture and regulation of the αC Helix

Eukaryotic protein kinases have evolved to be highly regulated and dynamic molecular switches that are typically kept in an inactive state and then activated in response to extracellular signals. The hallmark signature of an active kinase is a hydrophobic spine called the regulatory (R) spine, which consists of four residues, two in the N-lobe and two in the C-lobe. RS1 is in the catalytic loop, RS2 is the Phe in the DFG motif, RS3 is at the C-terminus of the αC-Helix, and RS4 is at the beginning of β4. Assembly of the R-spine is typically facilitated by phosphorylation of the Activation Loop. The assembled R-spine brings together all of the functional motifs that are essential for transferring the phosphate from ATP to a tethered protein substrate. This includes the G-Loop, which anchors the ATP, the catalytic loop, the DFG motif fused to the Activation Loop, and the αC-Helix. We focus here on the properties of the αC-Helix showing 1) how residues communicate with different parts of the molecule, 2) how it is recruited to the active site as a consequence of assembling of the R-spine, and 3) how it is regulated by linkers/motifs/proteins that lie outside the conserved kinase core. This article is part of a Special Issue entitled: Inhibitors of Protein Kinases.

Going for broke: targeting the human cancer pseudokinome

Protein phosphorylation lies at the heart of cell signalling, and somatic mutation(s) in kinases drives and sustains a multitude of human diseases, including cancer. The human protein kinase superfamily (the kinome) encodes approximately 50 'pseudokinases', which were initially predicted to be incapable of dynamic cell signalling when compared with canonical enzymatically active kinases. This assumption was supported by bioinformatics, which showed that amino acid changes at one or more key loci, making up the nucleotide-binding site or phosphotransferase machinery, were conserved in multiple vertebrate and non-vertebrate pseudokinase homologues. Protein kinases are highly attractive targets for drug discovery, as evidenced by the approval of almost 30 kinase inhibitors in oncology, and the successful development of the dual JAK1/2 (Janus kinase 1/2) inhibitor ruxolitinib for inflammatory indications. However, for such a large (>550) protein family, a remarkable number have still not been analysed at the molecular level, and only a surprisingly small percentage of kinases have been successfully targeted clinically. This is despite evidence that many are potential candidates for the development of new therapeutics. Indeed, several recent reports confirm that disease-associated pseudokinases can bind to nucleotide co-factors at concentrations achievable in the cell. Together, these findings suggest that drug targeting using either ATP-site or unbiased ligand-discovery approaches should now be attempted using the validation technology currently employed to evaluate their classic protein kinase counterparts. In the present review, we discuss members of the human pseudokinome repertoire, and catalogue somatic amino acid pseudokinase mutations that are emerging as the depth and clinical coverage of the human cancer pseudokinome expand.

Kin-Driver: a database of driver mutations in protein kinases

Somatic mutations in protein kinases (PKs) are frequent driver events in many human tumors, while germ-line mutations are associated with hereditary diseases. Here we present Kin-driver, the first database that compiles driver mutations in PKs with experimental evidence demonstrating their functional role. Kin-driver is a manual expert-curated database that pays special attention to activating mutations (AMs) and can serve as a validation set to develop new generation tools focused on the prediction of gain-of-function driver mutations. It also offers an easy and intuitive environment to facilitate the visualization and analysis of mutations in PKs. Because all mutations are mapped onto a multiple sequence alignment, analogue positions between kinases can be identified and tentative new mutations can be proposed for studying by transferring annotation. Finally, our database can also be of use to clinical and translational laboratories, helping them to identify uncommon AMs that can correlate with response to new antitumor drugs. The website was developed using PHP and JavaScript, which are supported by all major browsers; the database was built using MySQL server. Kin-driver is available at: http://kin-driver.leloir.org.ar/

Kinase regulation by hydrophobic spine assembly in cancer

A new model of kinase regulation based on the assembly of hydrophobic spines has been proposed. Changes in their positions can explain the mechanism of kinase activation. Here, we examined mutations in human cancer for clues about the regulation of the hydrophobic spines by focusing initially on mutations to Phe. We identified a selected number of Phe mutations in a small group of kinases that included BRAF, ABL1, and the epidermal growth factor receptor. Testing some of these mutations in BRAF, we found that one of the mutations impaired ATP binding and catalytic activity but promoted noncatalytic allosteric functions. Other Phe mutations functioned to promote constitutive catalytic activity. One of these mutations revealed a previously underappreciated hydrophobic surface that functions to position the dynamic regulatory αC-helix. This supports the key role of the C-helix as a signal integration motif for coordinating multiple elements of the kinase to create an active conformation. The importance of the hydrophobic space around the αC-helix was further tested by studying a V600F mutant, which was constitutively active in the absence of the negative charge that is associated with the common V600E mutation. Many hydrophobic mutations strategically localized along the C-helix can thus drive kinase activation.

Dynamic architecture of a protein kinase

Protein kinases are dynamically regulated signaling proteins that act as switches in the cell by phosphorylating target proteins. To establish a framework for analyzing linkages between structure, function, dynamics, and allostery in protein kinases, we carried out multiple microsecond-scale molecular-dynamics simulations of protein kinase A (PKA), an exemplar active kinase. We identified residue-residue correlated motions based on the concept of mutual information and used the Girvan-Newman method to partition PKA into structurally contiguous "communities." Most of these communities included 40-60 residues and were associated with a particular protein kinase function or a regulatory mechanism, and well-known motifs based on sequence and secondary structure were often split into different communities. The observed community maps were sensitive to the presence of different ligands and provide a new framework for interpreting long-distance allosteric coupling. Communication between different communities was also in agreement with the previously defined architecture of the protein kinase core based on the "hydrophobic spine" network. This finding gives us confidence in suggesting that community analyses can be used for other protein kinases and will provide an efficient tool for structural biologists. The communities also allow us to think about allosteric consequences of mutations that are linked to disease.

Prediction and prioritization of rare oncogenic mutations in the cancer Kinome using novel features and multiple classifiers

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.

Cancer progresses by accumulation of mutations in a subset of genes that confer growth advantage. The 518 protein kinase genes encoded in the human genome, collectively called the kinome, represent one of the largest families of oncogenes. Targeted sequencing studies of many different cancers have shown that the mutational landscape comprises both cancer-causing “driver” mutations and harmless “passenger” mutations. While the frequent recurrence of some driver mutations in human cancers helps distinguish them from the large number of passenger mutations, a significant challenge is to identify the rare “driver” mutations that are less frequently observed in patient samples and yet are causative. Here we combine computational and experimental approaches to identify rare cancer-associated mutations in Epidermal Growth Factor receptor kinase (EGFR), a signaling protein frequently mutated in cancers. Specifically, we evaluate a novel multiple-classifier approach and features specific to the protein kinase super-family in distinguishing known cancer-associated mutations from benign mutations. We then apply the multiple classifier to identify and test the functional impact of rare cancer-associated mutations in EGFR. We report, for the first time, that the EGFR mutations T725M and L861R, which are infrequently observed in cancers, constitutively activate EGFR in a manner analogous to the frequently observed driver mutations.

MONTAGE: CREATING SELF-POPULATING DOMAIN ONTOLOGIES FROM LINKED OPEN DATA

Domain-specific ontologies have become integral components of numerous semantic- and knowledge-based applications. However, creating such ontologies and populating them with correct individuals is a difficult and time-consuming process. Recently, a vast amount of knowledge has become available as part of the Linked Open Data (LOD) project, which includes data sets in multiple areas. In this paper, we present mOntage, a novel ontology design and population framework, which allows a domain expert to easily define a domain ontology schema and specify the ontology's classes and properties in terms of the subsets of the existing LOD data sources. The classes and properties of the ontology being created can be defined either directly, in terms of existing LOD-available classes and properties, or can be newly constructed by the domain expert. The definitions, called maps, are encoded as part of the ontology itself, effectively converting it into a self-populating ontology. Finally, a dedicated software system automatically populates the ontology with instances obtained from the selected LOD sources by executing suitable SPARQL queries. We illustrate our framework by creating Cancer Treatment ontology in the area of biomedicine.

Kinannote, a computer program to identify and classify members of the eukaryotic protein kinase superfamily

Motivation: Kinases of the eukaryotic protein kinase superfamily are key regulators of most aspects eukaryotic cellular behavior and have provided several drug targets including kinases dysregulated in cancers. The rapid increase in the number of genomic sequences has created an acute need to identify and classify members of this important class of enzymes efficiently and accurately.

Results: Kinannote produces a draft kinome and comparative analyses for a predicted proteome using a single line command, and it is currently the only tool that automatically classifies protein kinases using the controlled vocabulary of Hanks and Hunter [Hanks and Hunter (1995)]. A hidden Markov model in combination with a position-specific scoring matrix is used by Kinannote to identify kinases, which are subsequently classified using a BLAST comparison with a local version of KinBase, the curated protein kinase dataset from www.kinase.com. Kinannote was tested on the predicted proteomes from four divergent species. The average sensitivity and precision for kinome retrieval from the test species are 94.4 and 96.8%. The ability of Kinannote to classify identified kinases was also evaluated, and the average sensitivity and precision for full classification of conserved kinases are 71.5 and 82.5%, respectively. Kinannote has had a significant impact on eukaryotic genome annotation, providing protein kinase annotations for 36 genomes made public by the Broad Institute in the period spanning 2009 to the present.

Availability: Kinannote is freely available at http://sourceforge.net/projects/kinannote.

Contact:jmgold@broadinstitute.org

Supplementary information:Supplementary data are available at Bioinformatics online.