Cellular forgetting, desensitisation, stress and ageing in signalling networks. When do cells refuse to learn more?

Recent findings show that single, non-neuronal cells are also able to learn signalling responses developing cellular memory. In cellular learning nodes of signalling networks strengthen their interactions e.g. by the conformational memory of intrinsically disordered proteins, protein translocation, miRNAs, lncRNAs, chromatin memory and signalling cascades. This can be described by a generalized, unicellular Hebbian learning process, where those signalling connections, which participate in learning, become stronger. Here we review those scenarios, where cellular signalling is not only repeated in a few times (when learning occurs), but becomes too frequent, too large, or too complex and overloads the cell. This leads to desensitisation of signalling networks by decoupling signalling components, receptor internalization, and consequent downregulation. These molecular processes are examples of anti-Hebbian learning and 'forgetting' of signalling networks. Stress can be perceived as signalling overload inducing the desensitisation of signalling pathways. Ageing occurs by the summative effects of cumulative stress downregulating signalling. We propose that cellular learning desensitisation, stress and ageing may be placed along the same axis of more and more intensive (prolonged or repeated) signalling. We discuss how cells might discriminate between repeated and unexpected signals, and highlight the Hebbian and anti-Hebbian mechanisms behind the fold-change detection in the NF-κB signalling pathway. We list drug design methods using Hebbian learning (such as chemically-induced proximity) and clinical treatment modalities inducing (cancer, drug allergies) desensitisation or avoiding drug-induced desensitisation. A better discrimination between cellular learning, desensitisation and stress may open novel directions in drug design, e.g. helping to overcome drug resistance.

Learning of Signaling Networks: Molecular Mechanisms.

Molecular processes of neuronal learning have been well described. However, learning mechanisms of non-neuronal cells are not yet fully understood at the molecular level. Here, we discuss molecular mechanisms of cellular learning, including conformational memory of intrinsically disordered proteins (IDPs) and prions, signaling cascades, protein translocation, RNAs [miRNA and long noncoding RNA (lncRNA)], and chromatin memory. We hypothesize that these processes constitute the learning of signaling networks and correspond to a generalized Hebbian learning process of single, non-neuronal cells, and we discuss how cellular learning may open novel directions in drug design and inspire new artificial intelligence methods.

SignaLink3: a multi-layered resource to uncover tissue-specific signaling networks.

Signaling networks represent the molecular mechanisms controlling a cell's response to various internal or external stimuli. Most currently available signaling databases contain only a part of the complex network of intertwining pathways, leaving out key interactions or processes. Hence, we have developed SignaLink3 (http://signalink.org/), a value-added knowledge-base that provides manually curated data on signaling pathways and integrated data from several types of databases (interaction, regulation, localisation, disease, etc.) for humans, and three major animal model organisms. SignaLink3 contains over 400 000 newly added human protein-protein interactions resulting in a total of 700 000 interactions for Homo sapiens, making it one of the largest integrated signaling network resources. Next to H. sapiens, SignaLink3 is the only current signaling network resource to provide regulatory information for the model species Caenorhabditis elegans and Danio rerio, and the largest resource for Drosophila melanogaster. Compared to previous versions, we have integrated gene expression data as well as subcellular localization of the interactors, therefore uniquely allowing tissue-, or compartment-specific pathway interaction analysis to create more accurate models. Data is freely available for download in widely used formats, including CSV, PSI-MI TAB or SQL.

Probabilistic edge weights fine-tune Boolean network dynamics.

Biological systems are noisy by nature. This aspect is reflected in our experimental measurements and should be reflected in the models we build to better understand these systems. Noise can be especially consequential when trying to interpret specific regulatory interactions, i.e. regulatory network edges. In this paper, we propose a method to explicitly encode edge-noise in Boolean dynamical systems by probabilistic edge-weight (PEW) operators. PEW operators have two important features: first, they introduce a form of edge-weight into Boolean models through the noise, second, the noise is dependent on the dynamical state of the system, which enables more biologically meaningful modeling choices. Moreover, we offer a simple-to-use implementation in the already well-established BooleanNet framework. In two application cases, we show how the introduction of just a few PEW operators in Boolean models can fine-tune the emergent dynamics and increase the accuracy of qualitative predictions. This includes fine-tuning interactions which cause non-biological behaviors when switching between asynchronous and synchronous update schemes in dynamical simulations. Moreover, PEW operators also open the way to encode more exotic cellular dynamics, such as cellular learning, and to implementing edge-weights for regulatory networks inferred from omics data.

Efficient link prediction in the protein-protein interaction network using topological information in a generative adversarial network machine learning model.

BACKGROUND: The investigation of possible interactions between two proteins in intracellular signaling is an expensive and laborious procedure in the wet-lab, therefore, several in silico approaches have been implemented to narrow down the candidates for future experimental validations. Reformulating the problem in the field of network theory, the set of proteins can be represented as the nodes of a network, while the interactions between them as the edges. The resulting protein-protein interaction (PPI) network enables the use of link prediction techniques in order to discover new probable connections. Therefore, here we aimed to offer a novel approach to the link prediction task in PPI networks, utilizing a generative machine learning model. RESULTS: We created a tool that consists of two modules, the data processing framework and the machine learning model. As data processing, we used a modified breadth-first search algorithm to traverse the network and extract induced subgraphs, which served as image-like input data for our model. As machine learning, an image-to-image translation inspired conditional generative adversarial network (cGAN) model utilizing Wasserstein distance-based loss improved with gradient penalty was used, taking the combined representation from the data processing as input, and training the generator to predict the probable unknown edges in the provided induced subgraphs. Our link prediction tool was evaluated on the protein-protein interaction networks of five different species from the STRING database by calculating the area under the receiver operating characteristic, the precision-recall curves and the normalized discounted cumulative gain (AUROC, AUPRC, NDCG, respectively). Test runs yielded the averaged results of AUROC = 0.915, AUPRC = 0.176 and NDCG = 0.763 on all investigated species. CONCLUSION: We developed a software for the purpose of link prediction in PPI networks utilizing machine learning. The evaluation of our software serves as the first demonstration that a cGAN model, conditioned on raw topological features of the PPI network, is an applicable solution for the PPI prediction problem without requiring often unavailable molecular node attributes. The corresponding scripts are available at https://github.com/semmelweis-pharmacology/ppi_pred .

Screening and monitoring of the BTKC481S mutation in a real-world cohort of patients with relapsed/refractory chronic lymphocytic leukaemia during ibrutinib therapy.

The Bruton's tyrosine kinase (BTK) inhibitor ibrutinib has revolutionised the therapeutic landscape of chronic lymphocytic leukaemia (CLL). Acquired mutations emerging at position C481 in the BTK tyrosine kinase domain are the predominant genetic alterations associated with secondary ibrutinib resistance. To assess the correlation between disease progression, and the emergence and temporal dynamics of the most common resistance mutation BTKC481S , sensitive (10-4 ) time-resolved screening was performed in 83 relapsed/refractory CLL patients during single-agent ibrutinib treatment. With a median follow-up time of 40 months, BTKC481S was detected in 48·2% (40/83) of the patients, with 80·0% (32/40) of them showing disease progression during the examined period. In these 32 cases, representing 72·7% (32/44) of all patients experiencing relapse, emergence of the BTKC481S mutation preceded the symptoms of clinical relapse with a median of nine months. Subsequent Bcl-2 inhibition therapy applied in 28/32 patients harbouring BTKC481S and progressing on ibrutinib conferred clinical and molecular remission across the patients. Our study demonstrates the clinical value of sensitive BTKC481S monitoring with the largest longitudinally analysed real-world patient cohort reported to date and validates the feasibility of an early prediction of relapse in the majority of ibrutinib-treated relapsed/refractory CLL patients experiencing disease progression.

Discovering cooperative biomarkers for heterogeneous complex disease diagnoses.

Biomarkers with high reproducibility and accurate prediction performance can contribute to comprehending the underlying pathogenesis of related complex diseases and further facilitate disease diagnosis and therapy. Techniques integrating gene expression profiles and biological networks for the identification of network-based disease biomarkers are receiving increasing interest. The biomarkers for heterogeneous diseases often exhibit strong cooperative effects, which implies that a set of genes may achieve more accurate outcome prediction than any single gene. In this study, we evaluated various biomarker identification methods that consider gene cooperative effects implicitly or explicitly, and proposed the gene cooperation network to explicitly model the cooperative effects of gene combinations. The gene cooperation network-enhanced method, named as MarkRank, achieves superior performance compared with traditional biomarker identification methods in both simulation studies and real data sets. The biomarkers identified by MarkRank not only have a better prediction accuracy but also have stronger topological relationships in the biological network and exhibit high specificity associated with the related diseases. Furthermore, the top genes identified by MarkRank involve crucial biological processes of related diseases and give a good prioritization for known disease genes. In conclusion, MarkRank suggests that explicit modeling of gene cooperative effects can greatly improve biomarker identification for complex diseases, especially for diseases with high heterogeneity.

Synaptic polarity and sign-balance prediction using gene expression data in the Caenorhabditis elegans chemical synapse neuronal connectome network.

Graph theoretical analyses of nervous systems usually omit the aspect of connection polarity, due to data insufficiency. The chemical synapse network of Caenorhabditis elegans is a well-reconstructed directed network, but the signs of its connections are yet to be elucidated. Here, we present the gene expression-based sign prediction of the ionotropic chemical synapse connectome of C. elegans (3,638 connections and 20,589 synapses total), incorporating available presynaptic neurotransmitter and postsynaptic receptor gene expression data for three major neurotransmitter systems. We made predictions for more than two-thirds of these chemical synapses and observed an excitatory-inhibitory (E:I) ratio close to 4:1 which was found similar to that observed in many real-world networks. Our open source tool (http://EleganSign.linkgroup.hu) is simple but efficient in predicting polarities by integrating neuronal connectome and gene expression data.

Translocatome: a novel resource for the analysis of protein translocation between cellular organelles.

Here we present Translocatome, the first dedicated database of human translocating proteins (URL: http://translocatome.linkgroup.hu). The core of the Translocatome database is the manually curated data set of 213 human translocating proteins listing the source of their experimental validation, several details of their translocation mechanism, their local compartmentalized interactome, as well as their involvement in signalling pathways and disease development. In addition, using the well-established and widely used gradient boosting machine learning tool, XGBoost, Translocatome provides translocation probability values for 13 066 human proteins identifying 1133 and 3268 high- and low-confidence translocating proteins, respectively. The database has user-friendly search options with a UniProt autocomplete quick search and advanced search for proteins filtered by their localization, UniProt identifiers, translocation likelihood or data complexity. Download options of search results, manually curated and predicted translocating protein sets are available on its website. The update of the database is helped by its manual curation framework and connection to the previously published ComPPI compartmentalized protein-protein interaction database (http://comppi.linkgroup.hu). As shown by the application examples of merlin (NF2) and tumor protein 63 (TP63) Translocatome allows a better comprehension of protein translocation as a systems biology phenomenon and can be used as a discovery-tool in the protein translocation field.

The EntOptLayout Cytoscape plug-in for the efficient visualization of major protein complexes in protein-protein interaction and signalling networks.

MOTIVATION: Network visualizations of complex biological datasets usually result in 'hairball' images, which do not discriminate network modules. RESULTS: We present the EntOptLayout Cytoscape plug-in based on a recently developed network representation theory. The plug-in provides an efficient visualization of network modules, which represent major protein complexes in protein-protein interaction and signalling networks. Importantly, the tool gives a quality score of the network visualization by calculating the information loss between the input data and the visual representation showing a 3- to 25-fold improvement over conventional methods. AVAILABILITY AND IMPLEMENTATION: The plug-in (running on Windows, Linux, or Mac OS) and its tutorial (both in written and video forms) can be downloaded freely under the terms of the MIT license from: http://apps.cytoscape.org/apps/entoptlayout. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The Wisdom of Networks: A General Adaptation and Learning Mechanism of Complex Systems: The Network Core Triggers Fast Responses to Known Stimuli; Innovations Require the Slow Network Periphery and Are Encoded by Core-Remodeling.

I hypothesize that re-occurring prior experience of complex systems mobilizes a fast response, whose attractor is encoded by their strongly connected network core. In contrast, responses to novel stimuli are often slow and require the weakly connected network periphery. Upon repeated stimulus, peripheral network nodes remodel the network core that encodes the attractor of the new response. This "core-periphery learning" theory reviews and generalizes the heretofore fragmented knowledge on attractor formation by neural networks, periphery-driven innovation, and a number of recent reports on the adaptation of protein, neuronal, and social networks. The core-periphery learning theory may increase our understanding of signaling, memory formation, information encoding and decision-making processes. Moreover, the power of network periphery-related "wisdom of crowds" inventing creative, novel responses indicates that deliberative democracy is a slow yet efficient learning strategy developed as the success of a billion-year evolution. Also see the video abstract here: https://youtu.be/IIjP7zWGjVE.

Roles of heat shock factor 1 beyond the heat shock response.

Various stress factors leading to protein damage induce the activation of an evolutionarily conserved cell protective mechanism, the heat shock response (HSR), to maintain protein homeostasis in virtually all eukaryotic cells. Heat shock factor 1 (HSF1) plays a central role in the HSR. HSF1 was initially known as a transcription factor that upregulates genes encoding heat shock proteins (HSPs), also called molecular chaperones, which assist in refolding or degrading injured intracellular proteins. However, recent accumulating evidence indicates multiple additional functions for HSF1 beyond the activation of HSPs. Here, we present a nearly comprehensive list of non-HSP-related target genes of HSF1 identified so far. Through controlling these targets, HSF1 acts in diverse stress-induced cellular processes and molecular mechanisms, including the endoplasmic reticulum unfolded protein response and ubiquitin-proteasome system, multidrug resistance, autophagy, apoptosis, immune response, cell growth arrest, differentiation underlying developmental diapause, chromatin remodelling, cancer development, and ageing. Hence, HSF1 emerges as a major orchestrator of cellular stress response pathways.

Intracellular and intercellular signaling networks in cancer initiation, development and precision anti-cancer therapy: RAS acts as contextual signaling hub.

Cancer initiation and development are increasingly perceived as systems-level phenomena, where intra- and inter-cellular signaling networks of the ecosystem of cancer and stromal cells offer efficient methodologies for outcome prediction and intervention design. Within this framework, RAS emerges as a 'contextual signaling hub', i.e. the final result of RAS activation or inhibition is determined by the signaling network context. Current therapies often 'train' cancer cells shifting them to a novel attractor, which has increased metastatic potential and drug resistance. The few therapy-surviving cancer cells are surrounded by massive cell death triggering a primordial adaptive and reparative general wound healing response. Overall, dynamic analysis of patient- and disease-stage specific intracellular and intercellular signaling networks may open new areas of anticancer therapy using multitarget drugs, drugs combinations, edgetic drugs, as well as help design 'gentler', differentiation and maintenance therapies.

Oncogenic KRAS signaling and YAP1/ß-catenin: Similar cell cycle control in tumor initiation.

Why are YAP1 and c-Myc often overexpressed (or activated) in KRAS-driven cancers and drug resistance? Here, we propose that there are two independent pathways in tumor proliferation: one includes MAPK/ERK and PI3K/A kt/mTOR; and the other consists of pathways leading to the expression (or activation) of YAP1 and c-Myc. KRAS contributes through the first. MYC is regulated by e.g. ß-catenin, Notch and Hedgehog. We propose that YAP1 and ERK accomplish similar roles in cell cycle control, as do ß-catenin and PI3K. This point is compelling, since the question of how YAP1 rescues K-Ras or B-Raf ablation has recently captured much attention, as well as the mechanism of resistance to PI3K inhibitors. The similarity in cell cycle actions of ß-catenin and PI3K can also clarify the increased aggressiveness of lung cancer when both K-Ras and ß-catenin operate. Thus, we propose that the two pathways can substitute one another - or together amplify each other - in promoting proliferation. This new understanding of the independence and correspondence of the two pathways in cancer - MAPK/ERK and PI3K/Akt/mTOR; and YAP1 and c-Myc - provide a coherent and significant picture of signaling-driven oncogenic proliferation and may help in judicious, pathway-based drug discovery.

ComPPI: a cellular compartment-specific database for protein-protein interaction network analysis.

Here we present ComPPI, a cellular compartment-specific database of proteins and their interactions enabling an extensive, compartmentalized protein-protein interaction network analysis (URL: http://ComPPI.LinkGroup.hu). ComPPI enables the user to filter biologically unlikely interactions, where the two interacting proteins have no common subcellular localizations and to predict novel properties, such as compartment-specific biological functions. ComPPI is an integrated database covering four species (S. cerevisiae, C. elegans, D. melanogaster and H. sapiens). The compilation of nine protein-protein interaction and eight subcellular localization data sets had four curation steps including a manually built, comprehensive hierarchical structure of >1600 subcellular localizations. ComPPI provides confidence scores for protein subcellular localizations and protein-protein interactions. ComPPI has user-friendly search options for individual proteins giving their subcellular localization, their interactions and the likelihood of their interactions considering the subcellular localization of their interacting partners. Download options of search results, whole-proteomes, organelle-specific interactomes and subcellular localization data are available on its website. Due to its novel features, ComPPI is useful for the analysis of experimental results in biochemistry and molecular biology, as well as for proteome-wide studies in bioinformatics and network science helping cellular biology, medicine and drug design.

Cancer stem cells display extremely large evolvability: alternating plastic and rigid networks as a potential Mechanism: network models, novel therapeutic target strategies, and the contributions of hypoxia, inflammation and cellular senescence.

Cancer is increasingly perceived as a systems-level, network phenomenon. The major trend of malignant transformation can be described as a two-phase process, where an initial increase of network plasticity is followed by a decrease of plasticity at late stages of tumor development. The fluctuating intensity of stress factors, like hypoxia, inflammation and the either cooperative or hostile interactions of tumor inter-cellular networks, all increase the adaptation potential of cancer cells. This may lead to the bypass of cellular senescence, and to the development of cancer stem cells. We propose that the central tenet of cancer stem cell definition lies exactly in the indefinability of cancer stem cells. Actual properties of cancer stem cells depend on the individual "stress-history" of the given tumor. Cancer stem cells are characterized by an extremely large evolvability (i.e. a capacity to generate heritable phenotypic variation), which corresponds well with the defining hallmarks of cancer stem cells: the possession of the capacity to self-renew and to repeatedly re-build the heterogeneous lineages of cancer cells that comprise a tumor in new environments. Cancer stem cells represent a cell population, which is adapted to adapt. We argue that the high evolvability of cancer stem cells is helped by their repeated transitions between plastic (proliferative, symmetrically dividing) and rigid (quiescent, asymmetrically dividing, often more invasive) phenotypes having plastic and rigid networks. Thus, cancer stem cells reverse and replay cancer development multiple times. We describe network models potentially explaining cancer stem cell-like behavior. Finally, we propose novel strategies including combination therapies and multi-target drugs to overcome the Nietzschean dilemma of cancer stem cell targeting: "what does not kill me makes me stronger".

Teaching the bioinformatics of signaling networks: an integrated approach to facilitate multi-disciplinary learning.

The number of bioinformatics tools and resources that support molecular and cell biology approaches is continuously expanding. Moreover, systems and network biology analyses are accompanied more and more by integrated bioinformatics methods. Traditional information-centered university teaching methods often fail, as (1) it is impossible to cover all existing approaches in the frame of a single course, and (2) a large segment of the current bioinformation can become obsolete in a few years. Signaling network offers an excellent example for teaching bioinformatics resources and tools, as it is both focused and complex at the same time. Here, we present an outline of a university bioinformatics course with four sample practices to demonstrate how signaling network studies can integrate biochemistry, genetics, cell biology and network sciences. We show that several bioinformatics resources and tools, as well as important concepts and current trends, can also be integrated to signaling network studies. The research-type hands-on experiences we show enable the students to improve key competences such as teamworking, creative and critical thinking and problem solving. Our classroom course curriculum can be re-formulated as an e-learning material or applied as a part of a specific training course. The multi-disciplinary approach and the mosaic setup of the course have the additional benefit to support the advanced teaching of talented students.

Induced fit, conformational selection and independent dynamic segments: an extended view of binding events.

Single molecule and NMR measurements of protein dynamics increasingly uncover the complexity of binding scenarios. Here, we describe an extended conformational selection model that embraces a repertoire of selection and adjustment processes. Induced fit can be viewed as a subset of this repertoire, whose contribution is affected by the bond types stabilizing the interaction and the differences between the interacting partners. We argue that protein segments whose dynamics are distinct from the rest of the protein ('discrete breathers') can govern conformational transitions and allosteric propagation that accompany binding processes and, as such, might be more sensitive to mutational events. Additionally, we highlight the dynamic complexity of binding scenarios as they relate to events such as aggregation and signalling, and the crowded cellular environment.

Cancer-related networks: a help to understand, predict and change malignant transformation.

Cancer is increasingly described as a systems-level, network phenomenon. Genetic methods, such as next generation sequencing and RNA interference uncovered the complexity tumor-specific mutation-induced effects and the identification of multiple target sets. Network analysis of cancer-specific metabolic and signaling pathways highlighted the structural features of cancer-related proteins and their complexes to develop next-generation protein kinase inhibitors, as well as the modulation of inflammatory and autophagic pathways in anti-cancer therapies. Importantly, malignant transformation can be described as a two-phase process, where an initial increase of system plasticity is followed by a decrease of plasticity at late stages of tumor development. Late-stage tumors should be attacked by an indirect network influence strategy. On the contrary, the attack of early-stage tumors may target central network nodes. Cancer stem cells need special diagnosis and targeting, since they potentially have an extremely high ability to change the rigidity/plasticity of their networks. The early warning signals of the activation of fast growing tumor cell clones are important in personalized diagnosis and therapy. Multi-target attacks are needed to perturb cancer-specific networks to exit from cancer attractors and re-enter a normal attractor. However, the dynamic non-genetic heterogeneity of cancer cell population induces the replenishment of the cancer attractor with surviving, non-responsive cells from neighboring abnormal attractors. The development of drug resistance is further complicated by interactions of tumor clones and their microenvironment. Network analysis of intercellular cooperation using game theory approaches may open new areas of understanding tumor complexity. In conclusion, the above applications of the network approach open up new, and highly promising avenues in anti-cancer drug design.

Adaptation and learning of molecular networks as a description of cancer development at the systems-level: potential use in anti-cancer therapies.

There is a widening recognition that cancer cells are products of complex developmental processes. Carcinogenesis and metastasis formation are increasingly described as systems-level, network phenomena. Here we propose that malignant transformation is a two-phase process, where an initial increase of system plasticity is followed by a decrease of plasticity at late stages of carcinogenesis as a model of cellular learning. We describe the hallmarks of increased system plasticity of early, tumor initiating cells, such as increased noise, entropy, conformational and phenotypic plasticity, physical deformability, cell heterogeneity and network rearrangements. Finally, we argue that the large structural changes of molecular networks during cancer development necessitate a rather different targeting strategy in early and late phase of carcinogenesis. Plastic networks of early phase cancer development need a central hit, while rigid networks of late stage primary tumors or established metastases should be attacked by the network influence strategy, such as by edgetic, multi-target, or allo-network drugs. Cancer stem cells need special diagnosis and targeting, since their dormant and rapidly proliferating forms may have more rigid, or more plastic networks, respectively. The extremely high ability of cancer stem cells to change the rigidity/plasticity of their networks may be their key hallmark. The application of early stage-optimized anti-cancer drugs to late-stage patients may be a reason of many failures in anti-cancer therapies. Our hypotheses presented here underlie the need for patient-specific multi-target therapies applying the correct ratio of central hits and network influences - in an optimized sequence.