Classifying kinase conformations using a machine learning approach

Predicting the functional state of protein kinases using interpretable graph neural networks from sequence and structural data

Protein kinases are central to cellular activities and are actively pursued as drug targets for several conditions including cancer and autoimmune diseases. Despite the availability of a large structural database for kinases, methodologies to elucidate the structure-function relationship of these proteins (without manual intervention) are lacking. Such techniques are essential in structural biology and to accelerate drug discovery efforts. Here, we implement an interpretable graph neural network (GNN) framework for classifying the functionally active and inactive states of a large set of protein kinases by only using their tertiary structure and amino acid sequence. We show that the GNN models can classify kinase structures with high accuracy (>97%). We implement the Gradient-weighted Class Activation Mapping for graphs (Graph Grad-CAM) to automatically identify structurally important residues and residue-residue contacts of the kinases without any a priori input. We show that the motifs identified through the Graph Grad-CAM methodology are functionally critical, consistent with the existing kinase literature. Notably, the highly conserved DFG and HRD motifs of the well-known hydrophobic spine are identified by the interpretable framework in addition to some of the lesser known motifs. Further, using Grad-CAM maps as the vector embedding of the protein structures, we identify the subtle differences in the crystal structures among different sub-classes of kinases in the Protein Data Bank (PDB). Frameworks such as the one implemented here, for high-throughput identification of protein structure-function relationships are essential in designing targeted small molecules therapies as well as in engineering new proteins for novel applications.

Classifying protein kinase conformations with machine learning

Protein kinases are key actors of signaling networks and important drug targets. They cycle between active and inactive conformations, distinguished by a few elements within the catalytic domain. One is the activation loop, whose conserved DFG motif can occupy DFG-in, DFG-out, and some rarer conformations. Annotation and classification of the structural kinome are important, as different conformations can be targeted by different inhibitors and activators. Valuable resources exist; however, large-scale applications will benefit from increased automation and interpretability of structural annotation. Interpretable machine learning models are described for this purpose, based on ensembles of decision trees. To train them, a set of catalytic domain sequences and structures was collected, somewhat larger and more diverse than existing resources. The structures were clustered based on the DFG conformation and manually annotated. They were then used as training input. Two main models were constructed, which distinguished active/inactive and in/out/other DFG conformations. They considered initially 1692 structural variables, spanning the whole catalytic domain, then identified ("learned") a small subset that sufficed for accurate classification. The first model correctly labeled all but 3 of 3289 structures as active or inactive, while the second assigned the correct DFG label to all but 17 of 8826 structures. The most potent classifying variables were all related to well-known structural elements in or near the activation loop and their ranking gives insights into the conformational preferences. The models were used to automatically annotate 3850 kinase structures predicted recently with the Alphafold2 tool, showing that Alphafold2 reproduced the active/inactive but not the DFG-in proportions seen in the Protein Data Bank. We expect the models will be useful for understanding and engineering kinases.

Development of Kinase-Centric Drugs: A Computational Perspective

Kinases are prominent drug targets in the pharmaceutical and research community due to their involvement in signal transduction, physiological responses, and upon dysregulation, in diseases such as cancer, neurological and autoimmune disorders. Several FDA-approved small-molecule drugs have been developed to combat human diseases since Gleevec was approved for the treatment of chronic myelogenous leukemia. Kinases were considered "undruggable" in the beginning. Several FDA-approved small-molecule drugs have become available in recent years. Most of these drugs target ATP-binding sites, but a few target allosteric sites. Among kinases that belong to the same family, the catalytic domain shows high structural and sequence conservation. Inhibitors of ATP-binding sites can cause off-target binding. Because members of the same family have similar sequences and structural patterns, often complex relationships between kinases and inhibitors are observed. To design and develop drugs with desired selectivity, it is essential to understand the target selectivity for kinase inhibitors. To create new inhibitors with the desired selectivity, several experimental methods have been designed to profile the kinase selectivity of small molecules. Experimental approaches are often expensive, laborious, time-consuming, and limited by the available kinases. Researchers have used computational methodologies to address these limitations in the design and development of effective therapeutics. Many computational methods have been developed over the last few decades, either to complement experimental findings or to forecast kinase inhibitor activity and selectivity. The purpose of this review is to provide insight into recent advances in theoretical/computational approaches for the design of new kinase inhibitors with the desired selectivity and optimization of existing inhibitors.

Molecular Dynamics and Machine Learning Give Insights on the Flexibility-Activity Relationships in Tyrosine Kinome

Tyrosine kinases are a subfamily of kinases with critical roles in cellular machinery. Dysregulation of their active or inactive forms is associated with diseases like cancer. This study aimed to holistically understand their flexibility-activity relationships, focusing on pockets and fluctuations. We studied 43 different tyrosine kinases by collecting 120 μs of molecular dynamics simulations, pocket and residue fluctuation analysis, and a complementary machine learning approach. We found that the inactive forms often have increased flexibility, particularly at the DFG motif level. Noteworthy, thanks to these long simulations combined with a decision tree, we identified a semiquantitative fluctuation threshold of the DGF+3 residue over which the kinase has a higher probability to be in the inactive form.

Beyond sequence: Structure-based machine learning

Recent breakthroughs in protein structure prediction demarcate the start of a new era in structural bioinformatics. Combined with various advances in experimental structure determination and the uninterrupted pace at which new structures are published, this promises an age in which protein structure information is as prevalent and ubiquitous as sequence. Machine learning in protein bioinformatics has been dominated by sequence-based methods, but this is now changing to make use of the deluge of rich structural information as input. Machine learning methods making use of structures are scattered across literature and cover a number of different applications and scopes; while some try to address questions and tasks within a single protein family, others aim to capture characteristics across all available proteins. In this review, we look at the variety of structure-based machine learning approaches, how structures can be used as input, and typical applications of these approaches in protein biology. We also discuss current challenges and opportunities in this all-important and increasingly popular field.

Specificity and Reactivity of Mycobacterium tuberculosis Serine/Threonine Kinases PknG and PknB

Mycobacterium tuberculosis (Mtb), the causative agent of Tuberculosis, has 11 eukaryotic-like serine/threonine protein kinases, which play essential roles in cell growth, signal transduction, and pathogenesis. Protein kinase G (PknG) regulates the carbon and nitrogen metabolism by phosphorylation of the glycogen accumulation regulator (GarA) protein at Thr21. Protein kinase B (PknB) is involved in cell wall synthesis and cell shape, as well as phosphorylates GarA but at Thr22. While PknG seems to be constitutively activated and recognition of GarA requires phosphorylation in its unstructured tail, PknB activation is triggered by phosphorylation of its activation loop, which allows binding of the forkhead-associated domain of GarA. In the present work, we used molecular dynamics and quantum-mechanics/molecular mechanics simulations of the catalytically competent complex and kinase activity assays to understand PknG/PknB specificity and reactivity toward GarA. Two hydrophobic residues in GarA, Val24 and Phe25, seem essential for PknG binding and allow specificity for Thr21 phosphorylation. On the other hand, phosphorylated residues in PknB bind Arg26 in GarA and regulate its specificity for Thr22. We also provide a detailed analysis of the free energy profile for the phospho-transfer reaction and show why PknG has a constitutively active conformation not requiring priming phosphorylation in contrast to PknB. Our results provide new insights into these two key enzymes relevant for Mtb and the mechanisms of serine/threonine phosphorylation in bacteria.

Kincore: a web resource for structural classification of protein kinases and their inhibitors

The active form of kinases is shared across different family members, as are several commonly observed inactive forms. We previously performed a clustering of the conformation of the activation loop of all protein kinase structures in the Protein Data Bank (PDB) into eight classes based on the dihedral angles that place the Phe side chain of the DFG motif at the N-terminus of the activation loop. Our clusters are strongly associated with the placement of the activation loop, the C-helix, and other structural elements of kinases. We present Kincore, a web resource providing access to our conformational assignments for kinase structures in the PDB. While other available databases provide conformational states or drug type but not both, KinCore includes the conformational state and the inhibitor type (Type 1, 1.5, 2, 3, allosteric) for each kinase chain. The user can query and browse the database using these attributes or determine the conformational labels of a kinase structure using the web server or a standalone program. The database and labeled structure files can be downloaded from the server. Kincore will help in understanding the conformational dynamics of these proteins and guide development of inhibitors targeting specific states. Kincore is available at http://dunbrack.fccc.edu/kincore.

Pandemic drugs at pandemic speed: infrastructure for accelerating COVID-19 drug discovery with hybrid machine learning- and physics-based simulations on high-performance computers

The race to meet the challenges of the global pandemic has served as a reminder that the existing drug discovery process is expensive, inefficient and slow. There is a major bottleneck screening the vast number of potential small molecules to shortlist lead compounds for antiviral drug development. New opportunities to accelerate drug discovery lie at the interface between machine learning methods, in this case, developed for linear accelerators, and physics-based methods. The two in silico methods, each have their own advantages and limitations which, interestingly, complement each other. Here, we present an innovative infrastructural development that combines both approaches to accelerate drug discovery. The scale of the potential resulting workflow is such that it is dependent on supercomputing to achieve extremely high throughput. We have demonstrated the viability of this workflow for the study of inhibitors for four COVID-19 target proteins and our ability to perform the required large-scale calculations to identify lead antiviral compounds through repurposing on a variety of supercomputers.

Pseudokinases repurpose flexibility signatures associated with the protein kinase fold for noncatalytic roles

The bilobal protein kinase-like fold in pseudokinases lack one or more catalytic residues, conserved in canonical protein kinases, and are considered enzymatically deficient. Tertiary structures of pseudokinases reveal that their loops topologically equivalent to activation segments of kinases adopt contracted configurations, which is typically extended in active conformation of kinases. Herein, anisotropic network model based normal mode analysis (NMA) was conducted on 51 active conformation structures of protein kinases and 26 crystal structures of pseudokinases. Our observations indicate that although backbone fluctuation profiles are similar for individual kinase-pseudokinase families, low intensity mean square fluctuations in pseudo-activation segment and other sub-structures impart rigidity to pseudokinases. Analyses of collective motions from functional modes reveal that pseudokinases, compared to active kinases, undergo distinct conformational transitions using the same structural fold. All-atom NMA of protein kinase-pseudokinase pairs from each family, sharing high amino acid sequence identities, yielded distinct community clusters, partitioned by residues exhibiting highly correlated fluctuations. It appears that atomic fluctuations from equivalent activation segments guide community membership and network topologies for respective kinase and pseudokinase. Our findings indicate that such adaptations in backbone and side-chain fluctuations render pseudokinases competent for catalysis-independent roles.

The mechanism of activation of monomeric B-Raf V600E

Oncogenic mutations in the serine/threonine kinase B-Raf, particularly the V600E mutation, are frequent in cancer, making it a major drug target. Although much is known about B-Raf's active and inactive states, questions remain about the mechanism by which the protein changes between these two states. Here, we utilize molecular dynamics to investigate both wild-type and V600E B-Raf to gain mechanistic insights into the impact of the Val to Glu mutation. The results show that the wild-type and mutant follow similar activation pathways involving an extension of the activation loop and an inward motion of the αC-helix. The V600E mutation, however, destabilizes the inactive state by disrupting hydrophobic interactions present in the wild-type structure while the active state is stabilized through the formation of a salt bridge between Glu600 and Lys507. Additionally, when the activation loop is extended, the αC-helix is able to move between an inward and outward orientation as long as the DFG motif adopts a specific orientation. In that orientation Phe595 rotates away from the αC-helix, allowing the formation of a salt bridge between Lys483 and Glu501. These mechanistic insights have implications for the development of new Raf inhibitors.

Computational Approaches for the Design of (Mutant-)Selective Tyrosine Kinase Inhibitors: State-of-the-Art and Future Prospects

Kinases remain one of the major attractive therapeutic targets for a large number of indications such as cancer, rheumatoid arthritis, cardiac failure and many others. Design and development of kinase inhibitors (ATP-competitive, allosteric or covalent) is a clinically validated and successful strategy in the pharmaceutical industry. The perks come with limitations, particularly the development of resistance to highly potent and selective inhibitors. When this happens, the cycle needs to be repeated, i.e., the design and development of kinase inhibitors active against the mutated forms. The complexity of tumor milieu makes it awfully difficult for these molecularly-targeted therapies to work. Every year newer and better versions of these agents are introduced in the clinic. Several computational approaches such as structure-, ligand-based or hybrid ones continue to live up to their potential in discovering novel kinase inhibitors. New schools of thought in this area continue to emerge, e.g., development of dual-target kinase inhibitors. But there are fundamental issues with this approach. It is indeed difficult to selectively optimize binding at two entirely different or related kinases. In addition to the conventional strategies, modern technologies (machine learning, deep learning, artificial intelligence, etc.) started yielding the results and building success stories. Computational tools invariably played a critical role in catalysing the phenomenal progress in kinase drug discovery field. The present review summarized the progress in utilizing computational methods and tools for discovering (mutant-)selective tyrosine kinase inhibitor drugs in the last three years (2017-2019). Representative investigations have been discussed, while others are merely listed. The author believes that the enthusiastic reader will be inspired to dig out the cited literature extensively to appreciate the progress made so far and the future prospects of the field.

Tuning Local Hydration Enables a Deeper Understanding of Protein-Ligand Binding: The PP1-Src Kinase Case

Water plays a key role in biomolecular recognition and binding. Despite the development of several computational and experimental approaches, it is still challenging to comprehensively characterize water-mediated effects on the binding process. Here, we investigate how water affects the binding of Src kinase to one of its inhibitors, PP1. Src kinase is a target for treating several diseases, including cancer. We use biased molecular dynamics simulations, where the hydration of predetermined regions is tuned at will. This computational technique efficiently accelerates the SRC-PP1 binding simulation and allows us to identify several key and yet unexplored aspects of the solvent's role. This study provides a further perspective on the binding phenomenon, which may advance the current drug design approaches for the development of new kinase inhibitors.

Prediction of kinase inhibitors binding modes with machine learning and reduced descriptor sets

Protein kinases are receiving wide research interest, from drug perspective, due to their important roles in human body. Available kinase-inhibitor data, including crystallized structures, revealed many details about the mechanism of inhibition and binding modes. The understanding and analysis of these binding modes are expected to support the discovery of kinase-targeting drugs. The huge amounts of data made it possible to utilize computational techniques, including machine learning, to help in the discovery of kinase-targeting drugs. Machine learning gave reasonable predictions when applied to differentiate between the binding modes of kinase inhibitors, promoting a wider application in that domain. In this study, we applied machine learning supported by feature selection techniques to classify kinase inhibitors according to their binding modes. We represented inhibitors as a large number of molecular descriptors, as features, and systematically reduced these features in a multi-step manner while trying to attain high classification accuracy. Our predictive models could satisfy both goals by achieving high accuracy while utilizing at most 5% of the modeling features. The models could differentiate between binding mode types with MCC values between 0.67 and 0.92, and balanced accuracy values between 0.78 and 0.97 for independent test sets.

Characterizing an allosteric inhibitor-induced inactive state in with-no-lysine kinase 1 using Gaussian accelerated molecular dynamics simulations

The binding of an allosteric inhibitor in WNK1 leads to the inactive state.

Diversity of Long-Lived Intermediates along the Binding Pathway of Imatinib to Abl Kinase Revealed by MD Simulations

Imatinib, a drug used for the treatment of chronic myeloid leukemia and other cancers, works by blocking the catalytic site of pathological constitutively active Abl kinase. While the binding pose is known from X-ray crystallography, the different steps leading to the formation of the complex are not well understood. The results from extensive molecular dynamics simulations show that imatinib can primarily exit the known crystallographic binding pose through the cleft of the binding site or by sliding under the αC helix. Once displaced from the crystallographic binding pose, imatinib becomes trapped in intermediate states. These intermediates are characterized by a high diversity of ligand orientations and conformations, and relaxation timescales within this region may exceed 3-4 ms. Analysis indicates that the metastable intermediate states should be spectroscopically indistinguishable from the crystallographic binding pose, in agreement with tryptophan stopped-flow fluorescence experiments.

Cataloguing the dead: breathing new life into pseudokinase research

Pseudoenzymes are present within many, but not all, known enzyme families and lack one or more conserved canonical amino acids that help define their catalytically active counterparts. Recent findings in the pseudokinase field confirm that evolutionary repurposing of the structurally defined bilobal protein kinase fold permits distinct biological functions to emerge, many of which rely on conformational switching, as opposed to canonical catalysis. In this analysis, we evaluate progress in evaluating several members of the 'dark' pseudokinome that are pertinent to help drive this expanding field. Initially, we discuss how adaptions in erythropoietin-producing hepatocellular carcinoma (Eph) receptor tyrosine kinase domains resulted in two vertebrate pseudokinases, EphA10 and EphB6, in which co-evolving sequences generate new motifs that are likely to be important for both nucleotide binding and catalysis-independent signalling. Secondly, we discuss how conformationally flexible Tribbles pseudokinases, which have radiated in the complex vertebrates, control fundamental aspects of cell signalling that may be targetable with covalent small molecules. Finally, we show how species-level adaptions in the duplicated canonical kinase protein serine kinase histone (PSKH)1 sequence have led to the appearance of the pseudokinase PSKH2, whose physiological role remains mysterious. In conclusion, we show how the patterns we discover are selectively conserved within specific pseudokinases, and that when they are modelled alongside closely related canonical kinases, many are found to be located in functionally important regions of the conserved kinase fold. Interrogation of these patterns will be useful for future evaluation of these, and other, members of the unstudied human kinome.

Using phosphoproteomics data to understand cellular signaling: a comprehensive guide to bioinformatics resources

Mass spectrometry-based phosphoproteomics is becoming an essential methodology for the study of global cellular signaling. Numerous bioinformatics resources are available to facilitate the translation of phosphopeptide identification and quantification results into novel biological and clinical insights, a critical step in phosphoproteomics data analysis. These resources include knowledge bases of kinases and phosphatases, phosphorylation sites, kinase inhibitors, and sequence variants affecting kinase function, and bioinformatics tools that can predict phosphorylation sites in addition to the kinase that phosphorylates them, infer kinase activity, and predict the effect of mutations on kinase signaling. However, these resources exist in silos and it is challenging to select among multiple resources with similar functions. Therefore, we put together a comprehensive collection of resources related to phosphoproteomics data interpretation, compared the use of tools with similar functions, and assessed the usability from the standpoint of typical biologists or clinicians. Overall, tools could be improved by standardization of enzyme names, flexibility of data input and output format, consistent maintenance, and detailed manuals.

Caretta - A multiple protein structure alignment and feature extraction suite

The vast number of protein structures currently available opens exciting opportunities for machine learning on proteins, aimed at predicting and understanding functional properties. In particular, in combination with homology modelling, it is now possible to not only use sequence features as input for machine learning, but also structure features. However, in order to do so, robust multiple structure alignments are imperative.

Here we present Caretta, a multiple structure alignment suite meant for homologous but sequentially divergent protein families which consistently returns accurate alignments with a higher coverage than current state-of-the-art tools. Caretta is available as a GUI and command-line application and additionally outputs an aligned structure feature matrix for a given set of input structures, which can readily be used in downstream steps for supervised or unsupervised machine learning. We show Caretta’s performance on two benchmark datasets, and present an example application of Caretta in predicting the conformational state of cyclin-dependent kinases.

Identification of Druggable Kinase Target Conformations Using Markov Model Metastable States Analysis of apo-Abl

Kinases are important targets for drug development. However, accounting for the impact of possible structural rearrangements on the binding of kinase inhibitors is complicated by the extensive flexibility of their catalytic domain. The dynamic N-lobe contains four particular mobile structural elements: the Asp-Phe-Gly (DFG) motif, the phosphate (P) positioning loop, the activation (A) loop, and the αC helix. In our previous study [Meng et al. J. Chem. Theory Comput. 2018 14, 2721-2732], we combined various simulation techniques with Markov state modeling (MSM) to explore the free energy landscape of Abl kinase beyond conformations that are known from X-ray crystallography. Here we examine the resulting Markov model in greater detail by analyzing its metastable states. A characterization of the states in terms of their DFG state, P-loop, and αC conformations is presented and compared to existing classification schemes. Several metastable states are found to be structurally close to known crystal structures of different kinases in complex with a variety of inhibitors. These results suggest that the set of conformations accessible to tyrosine kinases may be shared within the entire family and that the conformational dynamics of one kinase in the absence of any ligand can provide meaningful information about possible target conformations for inhibitors of any member of the kinase family.

Structural Consequences of Multisite Phosphorylation in the BAK1 Kinase Domain

Multisite phosphorylation is an important mechanism of post-translational control of protein kinases. The effects of combinations of possible phosphorylation states on protein kinase activity are difficult to study experimentally because of challenges in isolating a particular phosphorylation state; surprising little effort on this topic has been expended in computational studies. To understand the effects of multisite phosphorylation on the plant protein kinase brassinosteroid insensitive 1-associated kinase 1 (BAK1) conformational ensemble, we performed Gaussian accelerated molecular dynamics simulations on eight BAK1 mod-forms involving phosphorylation of the four activation-loop threonine residues and binding of ATP-Mg²⁺. We find that unphosphorylated BAK1 transitions into an inactive conformation with a "cracked" activation loop and with the αC helix swung away from the active site. T450 phosphorylation can prevent the activation loop from cracking and keep the αC helix in an active-like conformation, whereas phosphorylation of T455 only slightly stabilizes the activation loop. There is a general trend of reduced flexibility in interlobe motion with increased phosphorylation. Interestingly, the αC helix is destabilized when the activation loop is fully phosphorylated but is again stabilized with ATP-Mg²⁺ bound. Our results provide insight into the mechanism of phosphorylation-controlled BAK1 activation while at the same time represent the first, to our knowledge, comprehensive, comparative study of the effects of combinatorial phosphorylation states on protein kinase conformational dynamics.

A Structurally-Validated Multiple Sequence Alignment of 497 Human Protein Kinase Domains

Studies on the structures and functions of individual kinases have been used to understand the biological properties of other kinases that do not yet have experimental structures. The key factor in accurate inference by homology is an accurate sequence alignment. We present a parsimonious, structure-based multiple sequence alignment (MSA) of 497 human protein kinase domains excluding atypical kinases. The alignment is arranged in 17 blocks of conserved regions and unaligned blocks in between that contain insertions of varying lengths present in only a subset of kinases. The aligned blocks contain well-conserved elements of secondary structure and well-known functional motifs, such as the DFG and HRD motifs. From pairwise, all-against-all alignment of 272 human kinase structures, we estimate the accuracy of our MSA to be 97%. The remaining inaccuracy comes from a few structures with shifted elements of secondary structure, and from the boundaries of aligned and unaligned regions, where compromises need to be made to encompass the majority of kinases. A new phylogeny of the protein kinase domains in the human genome based on our alignment indicates that ten kinases previously labeled as "OTHER" can be confidently placed into the CAMK group. These kinases comprise the Aurora kinases, Polo kinases, and calcium/calmodulin-dependent kinase kinases.

Kinase Activation by Small Conformational Changes

Protein kinases (PKs) are allosteric enzymes that play an essential role in signal transduction by regulating a variety of key cellular processes. Most PKs suffer conformational rearrangements upon phosphorylation that strongly enhance the catalytic activity. Generally, it involves the movement of the phosphorylated loop toward the active site and the rotation of the whole C-terminal lobe. However, not all kinases undergo such a large configurational change: The MAPK extracellular signal-regulated protein kinases ERK1 and ERK2 achieve a 50 000 fold increase in kinase activity with only a small motion of the C-terminal region. In the present work, we used a combination of molecular simulation tools to characterize the conformational landscape of ERK2 in the active (phosphorylated) and inactive (unphosphorylated) states in solution in agreement with NMR experiments. We show that the chemical reaction barrier is strongly dependent on ATP conformation and that the "active" low-barrier configuration is subtly regulated by phosphorylation, which stabilizes a key salt bridge between the conserved Lys52 and Glu69 belonging to helix-C and promotes binding of a second Mg ion. Our study highlights that the on-off switch embedded in the kinase fold can be regulated by small, medium, and large conformational changes.

LF4/MOK and a CDK-related kinase regulate the number and length of cilia in Tetrahymena

The length of cilia is controlled by a poorly understood mechanism that involves members of the conserved RCK kinase group, and among them, the LF4/MOK kinases. The multiciliated protist model, Tetrahymena, carries two types of cilia (oral and locomotory) and the length of the locomotory cilia is dependent on their position with the cell. In Tetrahymena, loss of an LF4/MOK ortholog, LF4A, lengthened the locomotory cilia, but also reduced their number. Without LF4A, cilia assembled faster and showed signs of increased intraflagellar transport (IFT). Consistently, overproduced LF4A shortened cilia and downregulated IFT. GFP-tagged LF4A, expressed in the native locus and imaged by total internal reflection microscopy, was enriched at the basal bodies and distributed along the shafts of cilia. Within cilia, most LF4A-GFP particles were immobile and a few either diffused or moved by IFT. We suggest that the distribution of LF4/MOK along the cilium delivers a uniform dose of inhibition to IFT trains that travel from the base to the tip. In a longer cilium, the IFT machinery may experience a higher cumulative dose of inhibition by LF4/MOK. Thus, LF4/MOK activity could be a readout of cilium length that helps to balance the rate of IFT-driven assembly with the rate of disassembly at steady state. We used a forward genetic screen to identify a CDK-related kinase, CDKR1, whose loss-of-function suppressed the shortening of cilia caused by overexpression of LF4A, by reducing its kinase activity. Loss of CDKR1 alone lengthened both the locomotory and oral cilia. CDKR1 resembles other known ciliary CDK-related kinases: LF2 of Chlamydomonas, mammalian CCRK and DYF-18 of C. elegans, in lacking the cyclin-binding motif and acting upstream of RCKs. The new genetic tools we developed here for Tetrahymena have potential for further dissection of the principles of cilia length regulation in multiciliated cells.

Two Antagonistic Hippo Signaling Circuits Set the Division Plane at the Medial Position in the Ciliate Tetrahymena

In a single cell, ciliates maintain a complex pattern of cortical organelles that are arranged along the anteroposterior and circumferential axes. The underlying molecular mechanisms of intracellular pattern formation in ciliates are largely unknown. Ciliates divide by tandem duplication, a process that remodels the parental cell into two daughters aligned head-to-tail. In the elo1-1 mutant of Tetrahymena thermophila, the segmentation boundary/division plane forms too close to the posterior end of the parental cell, producing a large anterior and a small posterior daughter cell, respectively. We show that ELO1 encodes a Lats/NDR kinase that marks the posterior segment of the cell cortex, where the division plane does not form in the wild-type. Elo1 acts independently of CdaI, a Hippo/Mst kinase that marks the anterior half of the parental cell, and whose loss shifts the division plane anteriorly. We propose that, in Tetrahymena, two antagonistic Hippo circuits focus the segmentation boundary/division plane at the equatorial position, by excluding divisional morphogenesis from the cortical areas that are too close to cell ends.

Perspectives and applications of machine learning for evolutionary developmental biology

Evolutionary Developmental Biology (Evo-Devo) is an ever-expanding field that aims to understand how development was modulated by the evolutionary process. In this sense, "omic" studies emerged as a powerful ally to unravel the molecular mechanisms underlying development. In this scenario, bioinformatics tools become necessary to analyze the growing amount of information. Among computational approaches, machine learning stands out as a promising field to generate knowledge and trace new research perspectives for bioinformatics. In this review, we aim to expose the current advances of machine learning applied to evolution and development. We draw clear perspectives and argue how evolution impacted machine learning techniques.

Altered conformational landscape and dimerization dependency underpins the activation of EGFR by αC-β4 loop insertion mutations

Mutational activation of epidermal growth factor receptor (EGFR) in human cancers involves both point mutations and complex mutations (insertions and deletions). In particular, short in-frame insertion mutations within a conserved αC-β4 loop in the EGFR kinase domain are frequently observed in tumor samples and patients harboring these mutations are insensitive to first-generation EGFR inhibitors. Despite the prevalence and clinical relevance of insertion mutations, the mechanisms by which these mutations regulate EGFR activity and contribute to drug sensitivity are poorly understood. Using cell-based mutation screening, we find that the precise location, length, and sequence of the inserted segment are critical for ligand-independent EGFR activation and downstream signaling. We identify three insertion mutations (N771_P772insN, D770_N771insG, and D770>GY) that activate EGFR in a unique way by relying more on the "acceptor" interface for kinase activation. Our drug inhibition studies indicate that these activating insertion mutations respond more favorably to osimertinib, a recently Food and Drug Administration-approved EGFR inhibitor for T790M-positive patients with lung cancer. Molecular dynamics simulations and umbrella sampling of WT and mutant EGFR suggest a model in which activating insertion mutations increase catalytic activity by relieving key autoinhibitory interactions associated with αC-helix movement and by lowering the transition free energy ([Formula: see text]) between active and inactive states. Our studies also identify a transition state sampled by activating insertion mutations that can be exploited in the design of mutant-selective EGFR inhibitors.

Redefining the Protein Kinase Conformational Space with Machine Learning

Protein kinases are dynamic, adopting different conformational states that are critical for their catalytic activity. We assess a range of structural features derived from the conserved αC helix and DFG motif to define the conformational space of the catalytic domain of protein kinases. We then construct Kinformation, a random forest classifier, to annotate the conformation of 3,708 kinase structures in the PDB. Our classification scheme captures known active and inactive kinase conformations and defines an additional conformational state, thereby refining the current understanding of the kinase conformational space. Furthermore, network analysis of the small molecules recognized by each conformation captures chemical substructures that are associated with each conformation type. Our description of the kinase conformational space is expected to improve modeling of protein kinase structures, as well as guide the development of conformation-specific kinase inhibitors with optimal pharmacological profiles.