How Machine Learning Will Transform Biomedicine.

This Perspective explores the application of machine learning toward improved diagnosis and treatment. We outline a vision for how machine learning can transform three broad areas of biomedicine: clinical diagnostics, precision treatments, and health monitoring, where the goal is to maintain health through a range of diseases and the normal aging process. For each area, early instances of successful machine learning applications are discussed, as well as opportunities and challenges for machine learning. When these challenges are met, machine learning promises a future of rigorous, outcomes-based medicine with detection, diagnosis, and treatment strategies that are continuously adapted to individual and environmental differences.

RhoA GEF Mcf2lb regulates rosette integrity during collective cell migration.

Multicellular rosettes are transient epithelial structures that serve as important cellular intermediates in the formation of diverse organs. Using the zebrafish posterior lateral line primordium (pLLP) as a model system, we investigated the role of the RhoA GEF Mcf2lb in rosette morphogenesis. The pLLP is a group of ∼150 cells that migrates along the zebrafish trunk and is organized into epithelial rosettes; these are deposited along the trunk and will differentiate into sensory organs called neuromasts (NMs). Using single-cell RNA-sequencing and whole-mount in situ hybridization, we showed that mcf2lb is expressed in the pLLP during migration. Live imaging and subsequent 3D analysis of mcf2lb mutant pLLP cells showed disrupted apical constriction and subsequent rosette organization. This resulted in an excess number of deposited NMs along the trunk of the zebrafish. Cell polarity markers ZO-1 and Par-3 were apically localized, indicating that pLLP cells are properly polarized. In contrast, RhoA activity, as well as signaling components downstream of RhoA, Rock2a and non-muscle Myosin II, were diminished apically. Thus, Mcf2lb-dependent RhoA activation maintains the integrity of epithelial rosettes.

Single-Cell Analysis of Rohon-Beard Neurons Implicates Fgf Signaling in Axon Maintenance and Cell Survival.

Peripheral sensory neurons are a critical part of the nervous system that transmit a multitude of sensory stimuli to the central nervous system. During larval and juvenile stages in zebrafish, this function is mediated by Rohon-Beard somatosensory neurons (RBs). RBs are optically accessible and amenable to experimental manipulation, making them a powerful system for mechanistic investigation of sensory neurons. Previous studies provided evidence that RBs fall into multiple subclasses; however, the number and molecular makeup of these potential RB subtypes have not been well defined. Using a single-cell RNA sequencing (scRNA-seq) approach, we demonstrate that larval RBs in zebrafish fall into three, largely nonoverlapping classes of neurons. We also show that RBs are molecularly distinct from trigeminal neurons in zebrafish. Cross-species transcriptional analysis indicates that one RB subclass is similar to a mammalian group of A-fiber sensory neurons. Another RB subclass is predicted to sense multiple modalities, including mechanical stimulation and chemical irritants. We leveraged our scRNA-seq data to determine that the fibroblast growth factor (Fgf) pathway is active in RBs. Pharmacological and genetic inhibition of this pathway led to defects in axon maintenance and RB cell death. Moreover, this can be phenocopied by treatment with dovitinib, an FDA-approved Fgf inhibitor with a common side effect of peripheral neuropathy. Importantly, dovitinib-mediated axon loss can be suppressed by loss of Sarm1, a positive regulator of neuronal cell death and axonal injury. This offers a molecular target for future clinical intervention to fight neurotoxic effects of this drug.

An altered extracellular matrix-integrin interface contributes to Huntington's disease-associated CNS dysfunction in glial and vascular cells.

Astrocytes and brain endothelial cells are components of the neurovascular unit that comprises the blood-brain barrier (BBB) and their dysfunction contributes to pathogenesis in Huntington's disease (HD). Defining the contribution of these cells to disease can inform cell-type-specific effects and uncover new disease-modifying therapeutic targets. These cells express integrin (ITG) adhesion receptors that anchor the cells to the extracellular matrix (ECM) to maintain the integrity of the BBB. We used HD patient-derived induced pluripotent stem cell (iPSC) modeling to study the ECM-ITG interface in astrocytes and brain microvascular endothelial cells and found ECM-ITG dysregulation in human iPSC-derived cells that may contribute to the dysfunction of the BBB in HD. This disruption has functional consequences since reducing ITG expression in glia in an HD Drosophila model suppressed disease-associated CNS dysfunction. Since ITGs can be targeted therapeutically and manipulating ITG signaling prevents neurodegeneration in other diseases, defining the role of ITGs in HD may provide a novel strategy of intervention to slow CNS pathophysiology to treat HD.

Systematic replication enables normalization of high-throughput imaging assays.

MOTIVATION: High-throughput fluorescent microscopy is a popular class of techniques for studying tissues and cells through automated imaging and feature extraction of hundreds to thousands of samples. Like other high-throughput assays, these approaches can suffer from unwanted noise and technical artifacts that obscure the biological signal. In this work, we consider how an experimental design incorporating multiple levels of replication enables the removal of technical artifacts from such image-based platforms. RESULTS: We develop a general approach to remove technical artifacts from high-throughput image data that leverages an experimental design with multiple levels of replication. To illustrate the methods, we consider microenvironment microarrays (MEMAs), a high-throughput platform designed to study cellular responses to microenvironmental perturbations. In application to MEMAs, our approach removes unwanted spatial artifacts and thereby enhances the biological signal. This approach has broad applicability to diverse biological assays. AVAILABILITY AND IMPLEMENTATION: Raw data are on synapse (syn2862345), analysis code is on github: gjhunt/mema_norm, a reproducible Docker image is available on dockerhub: gjhunt/mema_norm. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

LSD1 activates a lethal prostate cancer gene network independently of its demethylase function.

Medical castration that interferes with androgen receptor (AR) function is the principal treatment for advanced prostate cancer. However, clinical progression is universal, and tumors with AR-independent resistance mechanisms appear to be increasing in frequency. Consequently, there is an urgent need to develop new treatments targeting molecular pathways enriched in lethal prostate cancer. Lysine-specific demethylase 1 (LSD1) is a histone demethylase and an important regulator of gene expression. Here, we show that LSD1 promotes the survival of prostate cancer cells, including those that are castration-resistant, independently of its demethylase function and of the AR. Importantly, this effect is explained in part by activation of a lethal prostate cancer gene network in collaboration with LSD1's binding protein, ZNF217. Finally, that a small-molecule LSD1 inhibitor-SP-2509-blocks important demethylase-independent functions and suppresses castration-resistant prostate cancer cell viability demonstrates the potential of LSD1 inhibition in this disease.

Annot: a Django-based sample, reagent, and experiment metadata tracking system.

BACKGROUND: In biological experiments, comprehensive experimental metadata tracking - which comprises experiment, reagent, and protocol annotation with controlled vocabulary from established ontologies - remains a challenge, especially when the experiment involves multiple laboratory scientists who execute different steps of the protocol. Here we describe Annot, a novel web application designed to provide a flexible solution for this task. RESULTS: Annot enforces the use of controlled vocabulary for sample and reagent annotation while enabling robust investigation, study, and protocol tracking. The cornerstone of Annot's implementation is a json syntax-compatible file format, which can capture detailed metadata for all aspects of complex biological experiments. Data stored in this json file format can easily be ported into spreadsheet or data frame files that can be loaded into R ( https://www.r-project.org/ ) or Pandas, Python's data analysis library ( https://pandas.pydata.org/ ). Annot is implemented in Python3 and utilizes the Django web framework, Postgresql, Nginx, and Debian. It is deployed via Docker and supports all major browsers. CONCLUSIONS: Annot offers a robust solution to annotate samples, reagents, and experimental protocols for established assays where multiple laboratory scientists are involved. Further, it provides a framework to store and retrieve metadata for data analysis and integration, and therefore ensures that data generated in different experiments can be integrated and jointly analyzed. This type of solution to metadata tracking can enhance the utility of large-scale datasets, which we demonstrate here with a large-scale microenvironment microarray study.

Inferring causal molecular networks: empirical assessment through a community-based effort.

It remains unclear whether causal, rather than merely correlational, relationships in molecular networks can be inferred in complex biological settings. Here we describe the HPN-DREAM network inference challenge, which focused on learning causal influences in signaling networks. We used phosphoprotein data from cancer cell lines as well as in silico data from a nonlinear dynamical model. Using the phosphoprotein data, we scored more than 2,000 networks submitted by challenge participants. The networks spanned 32 biological contexts and were scored in terms of causal validity with respect to unseen interventional data. A number of approaches were effective, and incorporating known biology was generally advantageous. Additional sub-challenges considered time-course prediction and visualization. Our results suggest that learning causal relationships may be feasible in complex settings such as disease states. Furthermore, our scoring approach provides a practical way to empirically assess inferred molecular networks in a causal sense.

FOXA1 overexpression mediates endocrine resistance by altering the ER transcriptome and IL-8 expression in ER-positive breast cancer.

Forkhead box protein A1 (FOXA1) is a pioneer factor of estrogen receptor α (ER)-chromatin binding and function, yet its aberration in endocrine-resistant (Endo-R) breast cancer is unknown. Here, we report preclinical evidence for a role of FOXA1 in Endo-R breast cancer as well as evidence for its clinical significance. FOXA1 is gene-amplified and/or overexpressed in Endo-R derivatives of several breast cancer cell line models. Induced FOXA1 triggers oncogenic gene signatures and proteomic profiles highly associated with endocrine resistance. Integrated omics data reveal IL8 as one of the most perturbed genes regulated by FOXA1 and ER transcriptional reprogramming in Endo-R cells. IL-8 knockdown inhibits tamoxifen-resistant cell growth and invasion and partially attenuates the effect of overexpressed FOXA1. Our study highlights a role of FOXA1 via IL-8 signaling as a potential therapeutic target in FOXA1-overexpressing ER-positive tumors.

Metrics other than potency reveal systematic variation in responses to cancer drugs.

Large-scale analysis of cellular response to anticancer drugs typically focuses on variation in potency (half-maximum inhibitory concentration, (IC50)), assuming that it is the most important difference between effective and ineffective drugs or sensitive and resistant cells. We took a multiparametric approach involving analysis of the slope of the dose-response curve, the area under the curve and the maximum effect (Emax). We found that some of these parameters vary systematically with cell line and others with drug class. For cell-cycle inhibitors, Emax often but not always correlated with cell proliferation rate. For drugs targeting the Akt/PI3K/mTOR pathway, dose-response curves were unusually shallow. Classical pharmacology has no ready explanation for this phenomenon, but single-cell analysis showed that it correlated with significant and heritable cell-to-cell variability in the extent of target inhibition. We conclude that parameters other than potency should be considered in the comparative analysis of drug response, particularly at clinically relevant concentrations near and above the IC50.

Subtype and pathway specific responses to anticancer compounds in breast cancer.

Breast cancers are comprised of molecularly distinct subtypes that may respond differently to pathway-targeted therapies now under development. Collections of breast cancer cell lines mirror many of the molecular subtypes and pathways found in tumors, suggesting that treatment of cell lines with candidate therapeutic compounds can guide identification of associations between molecular subtypes, pathways, and drug response. In a test of 77 therapeutic compounds, nearly all drugs showed differential responses across these cell lines, and approximately one third showed subtype-, pathway-, and/or genomic aberration-specific responses. These observations suggest mechanisms of response and resistance and may inform efforts to develop molecular assays that predict clinical response.

Overcoming endocrine resistance due to reduced PTEN levels in estrogen receptor-positive breast cancer by co-targeting mammalian target of rapamycin, protein kinase B, or mitogen-activated protein kinase kinase.

INTRODUCTION: Activation of the phosphatidylinositol 3-kinase (PI3K) pathway in estrogen receptor α (ER)-positive breast cancer is associated with reduced ER expression and activity, luminal B subtype, and poor outcome. Phosphatase and tensin homolog (PTEN), a negative regulator of this pathway, is typically lost in ER-negative breast cancer. We set out to clarify the role of reduced PTEN levels in endocrine resistance, and to explore the combination of newly developed PI3K downstream kinase inhibitors to overcome this resistance. METHODS: Altered cellular signaling, gene expression, and endocrine sensitivity were determined in inducible PTEN-knockdown ER-positive/human epidermal growth factor receptor 2 (HER2)-negative breast cancer cell and/or xenograft models. Single or two-agent combinations of kinase inhibitors were examined to improve endocrine therapy. RESULTS: Moderate PTEN reduction was sufficient to enhance PI3K signaling, generate a gene signature associated with the luminal B subtype of breast cancer, and cause endocrine resistance in vitro and in vivo. The mammalian target of rapamycin (mTOR), protein kinase B (AKT), or mitogen-activated protein kinase kinase (MEK) inhibitors, alone or in combination, improved endocrine therapy, but the efficacy varied by PTEN levels, type of endocrine therapy, and the specific inhibitor(s). A single-agent AKT inhibitor combined with fulvestrant conferred superior efficacy in overcoming resistance, inducing apoptosis and tumor regression. CONCLUSIONS: Moderate reduction in PTEN, without complete loss, can activate the PI3K pathway to cause endocrine resistance in ER-positive breast cancer, which can be overcome by combining endocrine therapy with inhibitors of the PI3K pathway. Our data suggests that the ER degrader fulvestrant, to block both ligand-dependent and -independent ER signaling, combined with an AKT inhibitor is an effective strategy to test in patients.

TNBC response to paclitaxel phenocopies interferon response which reveals cell cycle-associated resistance mechanisms.

Paclitaxel is a standard of care neoadjuvant therapy for patients with triple negative breast cancer (TNBC); however, it shows limited benefit for locally advanced or metastatic disease. Here we used a coordinated experimental-computational approach to explore the influence of paclitaxel on the cellular and molecular responses of TNBC cells. We found that escalating doses of paclitaxel resulted in multinucleation, promotion of senescence, and initiation of DNA damage induced apoptosis. Single-cell RNA sequencing (scRNA-seq) of TNBC cells after paclitaxel treatment revealed upregulation of innate immune programs canonically associated with interferon response and downregulation of cell cycle progression programs. Systematic exploration of transcriptional responses to paclitaxel and cancer-associated microenvironmental factors revealed common gene programs induced by paclitaxel, IFNB, and IFNG. Transcription factor (TF) enrichment analysis identified 13 TFs that were both enriched based on activity of downstream targets and also significantly upregulated after paclitaxel treatment. Functional assessment with siRNA knockdown confirmed that the TFs FOSL1, NFE2L2 and ELF3 mediate cellular proliferation and also regulate nuclear structure. We further explored the influence of these TFs on paclitaxel-induced cell cycle behavior via live cell imaging, which revealed altered progression rates through G1, S/G2 and M phases. We found that ELF3 knockdown synergized with paclitaxel treatment to lock cells in a G1 state and prevent cell cycle progression. Analysis of publicly available breast cancer patient data showed that high ELF3 expression was associated with poor prognosis and enrichment programs associated with cell cycle progression. Together these analyses disentangle the diverse aspects of paclitaxel response and identify ELF3 upregulation as a putative biomarker of paclitaxel resistance in TNBC.

Single-cell morphodynamical trajectories enable prediction of gene expression accompanying cell state change.

Extracellular signals induce changes to molecular programs that modulate multiple cellular phenotypes, including proliferation, motility, and differentiation status. The connection between dynamically adapting phenotypic states and the molecular programs that define them is not well understood. Here we develop data-driven models of single-cell phenotypic responses to extracellular stimuli by linking gene transcription levels to "morphodynamics" - changes in cell morphology and motility observable in time-lapse image data. We adopt a dynamics-first view of cell state by grouping single-cell trajectories into states with shared morphodynamic responses. The single-cell trajectories enable development of a first-of-its-kind computational approach to map live-cell dynamics to snapshot gene transcript levels, which we term MMIST, Molecular and Morphodynamics-Integrated Single-cell Trajectories. The key conceptual advance of MMIST is that cell behavior can be quantified based on dynamically defined states and that extracellular signals alter the overall distribution of cell states by altering rates of switching between states. We find a cell state landscape that is bound by epithelial and mesenchymal endpoints, with distinct sequences of epithelial to mesenchymal transition (EMT) and mesenchymal to epithelial transition (MET) intermediates. The analysis yields predictions for gene expression changes consistent with curated EMT gene sets and provides a prediction of thousands of RNA transcripts through extracellular signal-induced EMT and MET with near-continuous time resolution. The MMIST framework leverages true single-cell dynamical behavior to generate molecular-level omics inferences and is broadly applicable to other biological domains, time-lapse imaging approaches and molecular snapshot data. Summary: Epithelial cells change behavior and state in response to signals, which is necessary for the function of healthy tissue, while aberrant responses can drive diseases like cancer. To decode and potentially steer these responses, there is a need to link live-cell behavior to molecular programs, but high-throughput molecular measurement is generally destructive or requires fixation. Here we present a novel method which connects single-cell morphology and motility over time to bulk molecular readouts. Our model predicts gene expression from the observation of label-free live-cell imaging, as a step toward understanding and ultimately controlling cell state change.

Functional delineation of the luminal epithelial microenvironment in breast using cell-based screening in combinatorial microenvironments.

Microenvironment signals are potent determinants of cell fate and arbiters of tissue homeostasis, however understanding how different microenvironment factors coordinately regulate cellular phenotype has been experimentally challenging. Here we used a high-throughput microenvironment microarray comprised of 2640 unique pairwise signals to identify factors that support proliferation and maintenance of primary human mammary luminal epithelial cells. Multiple microenvironment factors that modulated luminal cell number were identified, including: HGF, NRG1, BMP2, CXCL1, TGFB1, FGF2, PDGFB, RANKL, WNT3A, SPP1, HA, VTN, and OMD. All of these factors were previously shown to modulate luminal cell numbers in painstaking mouse genetics experiments, or were shown to have a role in breast cancer, demonstrating the relevance and power of our high-dimensional approach to dissect key microenvironmental signals. RNA-sequencing of primary epithelial and stromal cell lineages identified the cell types that express these signals and the cognate receptors in vivo. Cell-based functional studies confirmed which effects from microenvironment factors were reproducible and robust to individual variation. Hepatocyte growth factor (HGF) was the factor most robust to individual variation and drove expansion of luminal cells via cKit+ progenitor cells, which expressed abundant MET receptor. Luminal cells from women who are genetically high risk for breast cancer had significantly more MET receptor and may explain the characteristic expansion of the luminal lineage in those women. In ensemble, our approach provides proof of principle that microenvironment signals that control specific cellular states can be dissected with high-dimensional cell-based approaches.

Bayesian inference of signaling network topology in a cancer cell line.

MOTIVATION: Protein signaling networks play a key role in cellular function, and their dysregulation is central to many diseases, including cancer. To shed light on signaling network topology in specific contexts, such as cancer, requires interrogation of multiple proteins through time and statistical approaches to make inferences regarding network structure. RESULTS: In this study, we use dynamic Bayesian networks to make inferences regarding network structure and thereby generate testable hypotheses. We incorporate existing biology using informative network priors, weighted objectively by an empirical Bayes approach, and exploit a connection between variable selection and network inference to enable exact calculation of posterior probabilities of interest. The approach is computationally efficient and essentially free of user-set tuning parameters. Results on data where the true, underlying network is known place the approach favorably relative to existing approaches. We apply these methods to reverse-phase protein array time-course data from a breast cancer cell line (MDA-MB-468) to predict signaling links that we independently validate using targeted inhibition. The methods proposed offer a general approach by which to elucidate molecular networks specific to biological context, including, but not limited to, human cancers. AVAILABILITY: http://mukherjeelab.nki.nl/DBN (code and data).

MOBILE pipeline enables identification of context-specific networks and regulatory mechanisms.

Robust identification of context-specific network features that control cellular phenotypes remains a challenge. We here introduce MOBILE (Multi-Omics Binary Integration via Lasso Ensembles) to nominate molecular features associated with cellular phenotypes and pathways. First, we use MOBILE to nominate mechanisms of interferon-γ (IFNγ) regulated PD-L1 expression. Our analyses suggest that IFNγ-controlled PD-L1 expression involves BST2, CLIC2, FAM83D, ACSL5, and HIST2H2AA3 genes, which were supported by prior literature. We also compare networks activated by related family members transforming growth factor-beta 1 (TGFß1) and bone morphogenetic protein 2 (BMP2) and find that differences in ligand-induced changes in cell size and clustering properties are related to differences in laminin/collagen pathway activity. Finally, we demonstrate the broad applicability and adaptability of MOBILE by analyzing publicly available molecular datasets to investigate breast cancer subtype specific networks. Given the ever-growing availability of multi-omics datasets, we envision that MOBILE will be broadly useful for identification of context-specific molecular features and pathways.

Single Cell Spatial Biology for Precision Cancer Medicine.

In cancer, complex ecosystems of interacting cell types play fundamental roles in tumor development, progression, and response to therapy. However, the cellular organization, community structure, and spatially defined microenvironments of human tumors remain poorly understood. With the emergence of new technologies for high-throughput spatial profiling of complex tissue specimens, it is now possible to identify clinically significant spatial features with high granularity. In this PSB workshop, we will highlight recent advances in this area and explore how single cell spatial profiling can advance precision cancer medicine.

Morphodynamical cell state description via live-cell imaging trajectory embedding.

Time-lapse imaging is a powerful approach to gain insight into the dynamic responses of cells, but the quantitative analysis of morphological changes over time remains challenging. Here, we exploit the concept of "trajectory embedding" to analyze cellular behavior using morphological feature trajectory histories-that is, multiple time points simultaneously, rather than the more common practice of examining morphological feature time courses in single timepoint (snapshot) morphological features. We apply this approach to analyze live-cell images of MCF10A mammary epithelial cells after treatment with a panel of microenvironmental perturbagens that strongly modulate cell motility, morphology, and cell cycle behavior. Our morphodynamical trajectory embedding analysis constructs a shared cell state landscape revealing ligand-specific regulation of cell state transitions and enables quantitative and descriptive models of single-cell trajectories. Additionally, we show that incorporation of trajectories into single-cell morphological analysis enables (i) systematic characterization of cell state trajectories, (ii) better separation of phenotypes, and (iii) more descriptive models of ligand-induced differences as compared to snapshot-based analysis. This morphodynamical trajectory embedding is broadly applicable to the quantitative analysis of cell responses via live-cell imaging across many biological and biomedical applications.

RhoA GEF Mcf2lb regulates rosette integrity during collective cell migration.

During development, multicellular rosettes serve as important cellular intermediates in the formation of diverse organ systems. Multicellular rosettes are transient epithelial structures that are defined by the apical constriction of cells towards the rosette center. Due to the important role these structures play during development, understanding the molecular mechanisms by which rosettes are formed and maintained is of high interest. Utilizing the zebrafish posterior lateral line primordium (pLLP) as a model system, we identify the RhoA GEF Mcf2lb as a regulator of rosette integrity. The pLLP is a group of ~150 cells that migrates along the zebrafish trunk and is organized into epithelial rosettes; these are deposited along the trunk and will differentiate into sensory organs called neuromasts (NMs). Using single-cell RNA sequencing and whole-mount in situ hybridization, we showed that mcf2lb is expressed in the pLLP during migration. Given the known role of RhoA in rosette formation, we asked whether Mcf2lb plays a role in regulating apical constriction of cells within rosettes. Live imaging and subsequent 3D analysis of mcf2lb mutant pLLP cells showed disrupted apical constriction and subsequent rosette organization. This in turn resulted in a unique posterior Lateral Line phenotype: an excess number of deposited NMs along the trunk of the zebrafish. Cell polarity markers ZO-1 and Par-3 were apically localized, indicating that pLLP cells are normally polarized. In contrast, signaling components that mediate apical constriction downstream of RhoA, Rock-2a and non-muscle Myosin II were diminished apically. Altogether our results suggest a model whereby Mcf2lb activates RhoA, which in turn activates downstream signaling machinery to induce and maintain apical constriction in cells incorporated into rosettes.