Visible Machine Learning for Biomedicine.

A major ambition of artificial intelligence lies in translating patient data to successful therapies. Machine learning models face particular challenges in biomedicine, however, including handling of extreme data heterogeneity and lack of mechanistic insight into predictions. Here, we argue for "visible" approaches that guide model structure with experimental biology.

Heterogeneity of Myc expression in breast cancer exposes pharmacological vulnerabilities revealed through executable mechanistic modeling.

Cells with higher levels of Myc proliferate more rapidly and supercompetitively eliminate neighboring cells. Nonetheless, tumor cells in aggressive breast cancers typically exhibit significant and stable heterogeneity in their Myc levels, which correlates with refractoriness to therapy and poor prognosis. This suggests that Myc heterogeneity confers some selective advantage on breast tumor growth and progression. To investigate this, we created a traceable MMTV-Wnt1-driven in vivo chimeric mammary tumor model comprising an admixture of low-Myc- and reversibly switchable high-Myc-expressing clones. We show that such tumors exhibit interclonal mutualism wherein cells with high-Myc expression facilitate tumor growth by promoting protumorigenic stroma yet concomitantly suppress Wnt expression, which renders them dependent for survival on paracrine Wnt provided by low-Myc-expressing clones. To identify any therapeutic vulnerabilities arising from such interdependency, we modeled Myc/Ras/p53/Wnt signaling cross talk as an executable network for low-Myc, for high-Myc clones, and for the 2 together. This executable mechanistic model replicated the observed interdependence of high-Myc and low-Myc clones and predicted a pharmacological vulnerability to coinhibition of COX2 and MEK. This was confirmed experimentally. Our study illustrates the power of executable models in elucidating mechanisms driving tumor heterogeneity and offers an innovative strategy for identifying combination therapies tailored to the oligoclonal landscape of heterogenous tumors.

Single-cell analyses of regulatory network perturbations using enhancer-targeting TALEs suggest novel roles for PU.1 during haematopoietic specification.

Transcription factors (TFs) act within wider regulatory networks to control cell identity and fate. Numerous TFs, including Scl (Tal1) and PU.1 (Spi1), are known regulators of developmental and adult haematopoiesis, but how they act within wider TF networks is still poorly understood. Transcription activator-like effectors (TALEs) are a novel class of genetic tool based on the modular DNA-binding domains of Xanthomonas TAL proteins, which enable DNA sequence-specific targeting and the manipulation of endogenous gene expression. Here, we report TALEs engineered to target the PU.1-14kb and Scl+40kb transcriptional enhancers as efficient new tools to perturb the expression of these key haematopoietic TFs. We confirmed the efficiency of these TALEs at the single-cell level using high-throughput RT-qPCR, which also allowed us to assess the consequences of both PU.1 activation and repression on wider TF networks during developmental haematopoiesis. Combined with comprehensive cellular assays, these experiments uncovered novel roles for PU.1 during early haematopoietic specification. Finally, transgenic mouse studies confirmed that the PU.1-14kb element is active at sites of definitive haematopoiesis in vivo and PU.1 is detectable in haemogenic endothelium and early committing blood cells. We therefore establish TALEs as powerful new tools to study the functionality of transcriptional networks that control developmental processes such as early haematopoiesis.

BTR: training asynchronous Boolean models using single-cell expression data.

BACKGROUND: Rapid technological innovation for the generation of single-cell genomics data presents new challenges and opportunities for bioinformatics analysis. One such area lies in the development of new ways to train gene regulatory networks. The use of single-cell expression profiling technique allows the profiling of the expression states of hundreds of cells, but these expression states are typically noisier due to the presence of technical artefacts such as drop-outs. While many algorithms exist to infer a gene regulatory network, very few of them are able to harness the extra expression states present in single-cell expression data without getting adversely affected by the substantial technical noise present. RESULTS: Here we introduce BTR, an algorithm for training asynchronous Boolean models with single-cell expression data using a novel Boolean state space scoring function. BTR is capable of refining existing Boolean models and reconstructing new Boolean models by improving the match between model prediction and expression data. We demonstrate that the Boolean scoring function performed favourably against the BIC scoring function for Bayesian networks. In addition, we show that BTR outperforms many other network inference algorithms in both bulk and single-cell synthetic expression data. Lastly, we introduce two case studies, in which we use BTR to improve published Boolean models in order to generate potentially new biological insights. CONCLUSIONS: BTR provides a novel way to refine or reconstruct Boolean models using single-cell expression data. Boolean model is particularly useful for network reconstruction using single-cell data because it is more robust to the effect of drop-outs. In addition, BTR does not assume any relationship in the expression states among cells, it is useful for reconstructing a gene regulatory network with as few assumptions as possible. Given the simplicity of Boolean models and the rapid adoption of single-cell genomics by biologists, BTR has the potential to make an impact across many fields of biomedical research.

Processing, visualising and reconstructing network models from single-cell data.

New single-cell technologies readily permit gene expression profiling of thousands of cells at single-cell resolution. In this review, we will discuss methods for visualisation and interpretation of single-cell gene expression data, and the computational analysis needed to go from raw data to predictive executable models of gene regulatory network function. We will focus primarily on single-cell real-time quantitative PCR and RNA-sequencing data, but much of what we cover will also be relevant to other platforms, such as the mass cytometry technology for high-dimensional single-cell proteomics.

Emergent stem cell homeostasis in the C. elegans germline is revealed by hybrid modeling.

The establishment of homeostasis among cell growth, differentiation, and apoptosis is of key importance for organogenesis. Stem cells respond to temporally and spatially regulated signals by switching from mitotic proliferation to asymmetric cell division and differentiation. Executable computer models of signaling pathways can accurately reproduce a wide range of biological phenomena by reducing detailed chemical kinetics to a discrete, finite form. Moreover, coordinated cell movements and physical cell-cell interactions are required for the formation of three-dimensional structures that are the building blocks of organs. To capture all these aspects, we have developed a hybrid executable/physical model describing stem cell proliferation, differentiation, and homeostasis in the Caenorhabditis elegans germline. Using this hybrid model, we are able to track cell lineages and dynamic cell movements during germ cell differentiation. We further show how apoptosis regulates germ cell homeostasis in the gonad, and propose a role for intercellular pressure in developmental control. Finally, we use the model to demonstrate how an executable model can be developed from the hybrid system, identifying a mechanism that ensures invariance in fate patterns in the presence of instability.

WormQTL--public archive and analysis web portal for natural variation data in Caenorhabditis spp.

Here, we present WormQTL (http://www.wormqtl.org), an easily accessible database enabling search, comparative analysis and meta-analysis of all data on variation in Caenorhabditis spp. Over the past decade, Caenorhabditis elegans has become instrumental for molecular quantitative genetics and the systems biology of natural variation. These efforts have resulted in a valuable amount of phenotypic, high-throughput molecular and genotypic data across different developmental worm stages and environments in hundreds of C. elegans strains. WormQTL provides a workbench of analysis tools for genotype-phenotype linkage and association mapping based on but not limited to R/qtl (http://www.rqtl.org). All data can be uploaded and downloaded using simple delimited text or Excel formats and are accessible via a public web user interface for biologists and R statistic and web service interfaces for bioinformaticians, based on open source MOLGENIS and xQTL workbench software. WormQTL welcomes data submissions from other worm researchers.

Transcriptional hierarchies regulating early blood cell development.

Hematopoiesis represents one of the paradigmatic systems for studying stem cell biology, but our understanding of how the hematopoietic system develops during embryogenesis is still incomplete. While many lessons have been learned from studying the mouse embryo, embryonic stem cells have come to the fore as an alternative and more tractable model to recapitulate hematopoietic development. Here we review what is known about the embryonic origin of blood from these complementary systems and how transcription factor networks regulate the emergence of hematopoietic tissue from the mesoderm. Furthermore, we have performed an integrated analysis of genome-wide microarray and ChIP-seq data sets from mouse embryos and embryonic stem (ES) cell lines deficient in key regulators and demonstrate how this type of analysis can be used to reconstruct regulatory hierarchies that both confirm existing regulatory linkages and suggest additional interactions.

Static network structure can be used to model the phenotypic effects of perturbations in regulatory networks.

MOTIVATION: Biological processes are dynamic, whereas the networks that depict them are typically static. Quantitative modeling using differential equations or logic-based functions can offer quantitative predictions of the behavior of biological systems, but they require detailed experimental characterization of interaction kinetics, which is typically unavailable. To determine to what extent complex biological processes can be modeled and analyzed using only the static structure of the network (i.e. the direction and sign of the edges), we attempt to predict the phenotypic effect of perturbations in biological networks from the static network structure. RESULTS: We analyzed three networks from different sources: The EGFR/MAPK and PI3K/AKT network from a detailed experimental study, the TNF regulatory network from the STRING database and a large network of all NCI-curated pathways from the Protein Interaction Database. Altogether, we predicted the effect of 39 perturbations (e.g. by one or two drugs) on 433 target proteins/genes. In up to 82% of the cases, an algorithm that used only the static structure of the network correctly predicted whether any given protein/gene is upregulated or downregulated as a result of perturbations of other proteins/genes. CONCLUSION: While quantitative modeling requires detailed experimental data and heavy computations, which limit its scalability for large networks, a wiring-based approach can use available data from pathway and interaction databases and may be scalable. These results lay the foundations for a large-scale approach of predicting phenotypes based on the schematic structure of networks.

Cell-cycle regulation of NOTCH signaling during C. elegans vulval development.

C. elegans vulval development is one of the best-characterized systems to study cell fate specification during organogenesis. The detailed knowledge of the signaling pathways determining vulval precursor cell (VPC) fates permitted us to create a computational model based on the antagonistic interactions between the epidermal growth factor receptor (EGFR)/RAS/MAPK and the NOTCH pathways that specify the primary and secondary fates, respectively. A key notion of our model is called bounded asynchrony, which predicts that a limited degree of asynchrony in the progression of the VPCs is necessary to break their equivalence. While searching for a molecular mechanism underlying bounded asynchrony, we discovered that the termination of NOTCH signaling is tightly linked to cell-cycle progression. When single VPCs were arrested in the G1 phase, intracellular NOTCH failed to be degraded, resulting in a mixed primary/secondary cell fate. Moreover, the G1 cyclins CYD-1 and CYE-1 stabilize NOTCH, while the G2 cyclin CYB-3 promotes NOTCH degradation. Our findings reveal a synchronization mechanism that coordinates NOTCH signaling with cell-cycle progression and thus permits the formation of a stable cell fate pattern.

Multi-scale characterisation of homologous recombination deficiency in breast cancer.

BACKGROUND: Homologous recombination is a robust, broadly error-free mechanism of double-strand break repair, and deficiencies lead to PARP inhibitor sensitivity. Patients displaying homologous recombination deficiency can be identified using 'mutational signatures'. However, these patterns are difficult to reliably infer from exome sequencing. Additionally, as mutational signatures are a historical record of mutagenic processes, this limits their utility in describing the current status of a tumour. METHODS: We apply two methods for characterising homologous recombination deficiency in breast cancer to explore the features and heterogeneity associated with this phenotype. We develop a likelihood-based method which leverages small insertions and deletions for high-confidence classification of homologous recombination deficiency for exome-sequenced breast cancers. We then use multinomial elastic net regression modelling to develop a transcriptional signature of heterogeneous homologous recombination deficiency. This signature is then applied to single-cell RNA-sequenced breast cancer cohorts enabling analysis of homologous recombination deficiency heterogeneity and differential patterns of tumour microenvironment interactivity. RESULTS: We demonstrate that the inclusion of indel events, even at low levels, improves homologous recombination deficiency classification. Whilst BRCA-positive homologous recombination deficient samples display strong similarities to those harbouring BRCA1/2 defects, they appear to deviate in microenvironmental features such as hypoxic signalling. We then present a 228-gene transcriptional signature which simultaneously characterises homologous recombination deficiency and BRCA1/2-defect status, and is associated with PARP inhibitor response. Finally, we show that this signature is applicable to single-cell transcriptomics data and predict that these cells present a distinct milieu of interactions with their microenvironment compared to their homologous recombination proficient counterparts, typified by a decreased cancer cell response to TNFα signalling. CONCLUSIONS: We apply multi-scale approaches to characterise homologous recombination deficiency in breast cancer through the development of mutational and transcriptional signatures. We demonstrate how indels can improve homologous recombination deficiency classification in exome-sequenced breast cancers. Additionally, we demonstrate the heterogeneity of homologous recombination deficiency, especially in relation to BRCA1/2-defect status, and show that indications of this feature can be captured at a single-cell level, enabling further investigations into interactions between DNA repair deficient cells and their tumour microenvironment.

Localized immune surveillance of primary melanoma in the skin deciphered through executable modeling.

While skin is a site of active immune surveillance, primary melanomas often escape detection. Here, we have developed an in silico model to determine the local cross-talk between melanomas and Langerhans cells (LCs), the primary antigen-presenting cells at the site of melanoma development. The model predicts that melanomas fail to activate LC migration to lymph nodes until tumors reach a critical size, which is determined by a positive TNF-α feedback loop within melanomas, in line with our observations of murine tumors. In silico drug screening, supported by subsequent experimental testing, shows that treatment of primary tumors with MAPK pathway inhibitors may further prevent LC migration. In addition, our in silico model predicts treatment combinations that bypass LC dysfunction. In conclusion, our combined approach of in silico and in vivo studies suggests a molecular mechanism that explains how early melanomas develop under the radar of immune surveillance by LC.

The role of proteosome-mediated proteolysis in modulating potentially harmful transcription factor activity in Saccharomyces cerevisiae.

MOTIVATION: The appropriate modulation of the stress response to variable environmental conditions is necessary to maintain sustained viability in Saccharomyces cerevisiae. Particularly, controlling the abundance of proteins that may have detrimental effects on cell growth is crucial for rapid recovery from stress-induced quiescence. RESULTS: Prompted by qualitative modeling of the nutrient starvation response in yeast, we investigated in vivo the effect of proteolysis after nutrient starvation showing that, for the Gis1 transcription factor at least, proteasome-mediated control is crucial for a rapid return to growth. Additional bioinformatics analyses show that potentially toxic transcriptional regulators have a significantly lower protein half-life, a higher fraction of unstructured regions and more potential PEST motifs than the non-detrimental ones. Furthermore, inhibiting proteasome activity tends to increase the expression of genes induced during the Environmental Stress Response more than those in the rest of the genome. Our combined results suggest that proteasome-mediated proteolysis of potentially toxic transcription factors tightly modulates the stress response in yeast. CONTACT: jasmin.fisher@microsoft.com

A dynamic physical model of cell migration, differentiation and apoptosis in Caenorhabditis elegans.

The germ line of the nematode C. elegans provides a paradigm to study essential developmental concepts like stem cell differentiation and apoptosis. Here, we have created a computational model encompassing these developmental landmarks and the resulting movement of germ cells along the gonadal tube. We have used a technique based on molecular dynamics (MD) to model the physical movement of cells solely based on the force that arises from dividing cells. This novel way of using MD to drive the model enables calibration of simulation and experimental time. Based on this calibration, the analysis of our model shows that it is in accordance with experimental observations. In addition, the model provides insights into kinetics of molecular pathways within individual cells as well as into physical aspects like the cell density along the germ line and in local neighbourhoods of individual germ cells. In the future, the presented model can be used to test hypotheses about diverse aspects of development like stem cell division or programmed cell death. An iterative process of evolving this model and experimental testing in the model system C. elegans will provide new insights into key developmental aspects.

HOXA9 has the hallmarks of a biological switch with implications in blood cancers.

Blood malignancies arise from the dysregulation of haematopoiesis. The type of blood cell and the specific order of oncogenic events initiating abnormal growth ultimately determine the cancer subtype and subsequent clinical outcome. HOXA9 plays an important role in acute myeloid leukaemia (AML) prognosis by promoting blood cell expansion and altering differentiation; however, the function of HOXA9 in other blood malignancies is still unclear. Here, we highlight the biological switch and prognosis marker properties of HOXA9 in AML and chronic myeloproliferative neoplasms (MPN). First, we establish the ability of HOXA9 to stratify AML patients with distinct cellular and clinical outcomes. Then, through the use of a computational network model of MPN, we show that the self-activation of HOXA9 and its relationship to JAK2 and TET2 can explain the branching progression of JAK2/TET2 mutant MPN patients towards divergent clinical characteristics. Finally, we predict a connection between the RUNX1 and MYB genes and a suppressive role for the NOTCH pathway in MPN diseases.

Executable network of SARS-CoV-2-host interaction predicts drug combination treatments.

The COVID-19 pandemic has pushed healthcare systems globally to a breaking point. The urgent need for effective and affordable COVID-19 treatments calls for repurposing combinations of approved drugs. The challenge is to identify which combinations are likely to be most effective and at what stages of the disease. Here, we present the first disease-stage executable signalling network model of SARS-CoV-2-host interactions used to predict effective repurposed drug combinations for treating early- and late stage severe disease. Using our executable model, we performed in silico screening of 9870 pairs of 140 potential targets and have identified nine new drug combinations. Camostat and Apilimod were predicted to be the most promising combination in effectively supressing viral replication in the early stages of severe disease and were validated experimentally in human Caco-2 cells. Our study further demonstrates the power of executable mechanistic modelling to enable rapid pre-clinical evaluation of combination therapies tailored to disease progression. It also presents a novel resource and expandable model system that can respond to further needs in the pandemic.

Mapping Rora expression in resting and activated CD4+ T cells.

The transcription factor Rora has been shown to be important for the development of ILC2 and the regulation of ILC3, macrophages and Treg cells. Here we investigate the role of Rora across CD4+ T cells in general, but with an emphasis on Th2 cells, both in vitro as well as in the context of several in vivo type 2 infection models. We dissect the function of Rora using overexpression and a CD4-conditional Rora-knockout mouse, as well as a RORA-reporter mouse. We establish the importance of Rora in CD4+ T cells for controlling lung inflammation induced by Nippostrongylus brasiliensis infection, and have measured the effect on downstream genes using RNA-seq. Using a systematic stimulation screen of CD4+ T cells, coupled with RNA-seq, we identify upstream regulators of Rora, most importantly IL-33 and CCL7. Our data suggest that Rora is a negative regulator of the immune system, possibly through several downstream pathways, and is under control of the local microenvironment.

Executable cell biology.

Computational modeling of biological systems is becoming increasingly important in efforts to better understand complex biological behaviors. In this review, we distinguish between two types of biological models--mathematical and computational--which differ in their representations of biological phenomena. We call the approach of constructing computational models of biological systems 'executable biology', as it focuses on the design of executable computer algorithms that mimic biological phenomena. We survey the main modeling efforts in this direction, emphasize the applicability and benefits of executable models in biological research and highlight some of the challenges that executable biology poses for biology and computer science. We claim that for executable biology to reach its full potential as a mainstream biological technique, formal and algorithmic approaches must be integrated into biological research. This will drive biology toward a more precise engineering discipline.

Executable cancer models: successes and challenges.

Making decisions on how best to treat cancer patients requires the integration of different data sets, including genomic profiles, tumour histopathology, radiological images, proteomic analysis and more. This wealth of biological information calls for novel strategies to integrate such information in a meaningful, predictive and experimentally verifiable way. In this Perspective we explain how executable computational models meet this need. Such models provide a means for comprehensive data integration, can be experimentally validated, are readily interpreted both biologically and clinically, and have the potential to predict effective therapies for different cancer types and subtypes. We explain what executable models are and how they can be used to represent the dynamic biological behaviours inherent in cancer, and demonstrate how such models, when coupled with automated reasoning, facilitate our understanding of the mechanisms by which oncogenic signalling pathways regulate tumours. We explore how executable models have impacted the field of cancer research and argue that extending them to represent a tumour in a specific patient (that is, an avatar) will pave the way for improved personalized treatments and precision medicine. Finally, we highlight some of the ongoing challenges in developing executable models and stress that effective cross-disciplinary efforts are key to forward progress in the field.

Constructing and Analyzing Computational Models of Cell Signaling with BioModelAnalyzer.

BioModelAnalyzer (BMA) is an open-source graphical tool for the development of executable models of protein and gene networks within cells. Based upon the Qualitative Networks formalism, the user can rapidly construct large networks, either manually or by connecting motifs selected from a built-in library. After the appropriate functions for each variable are defined, the user has access to three analysis engines to test the model. In addition to standard simulation tools, BMA includes an interface to the stability-testing algorithm and to a graphical Linear Temporal Logic (LTL) editor and analysis tool. Alongside this, we have developed a novel ChatBot to aid users constructing LTL queries and to explain the interface and run through tutorials. Here we present worked examples of model construction and testing via the interface. As an initial example, we discuss fate decisions in Dictyostelium discoidum and cAMP signaling. We go on to describe the workflow leading to the construction of a published model of the germline of C. elegans. Finally, we demonstrate how to construct simple models from the built-in network motif library. © 2020 by John Wiley & Sons, Inc. Basic Protocol 1: Modeling the signaling network of Dictyostelium discoidum Basic Protocol 2: Modeling the germline progression of Caenorhabditis elegans Basic Protocol 3: Constructing a model of the cell cycle using motifs.