VEGF inhibitor drugs are part of standard care in oncology and ophthalmology, but not all patients respond to them. Combinations of drugs are likely to be needed for more effective therapies of angiogenesis-related diseases. In this paper we describe naturally occurring combinations of receptors in endothelial cells that might help to understand how cells communicate and to identify targets for drug combinations. We also develop and share a new software tool called DECNEO to identify them. Single-cell gene expression data are used to identify a set of co-expressed endothelial cell receptors, conserved among species (mice and humans) and enriched, within a network, of connections to up-regulated genes. This set includes several receptors previously shown to play a role in angiogenesis. Multiple statistical tests from large datasets, including an independent validation set, support the reproducibility, evolutionary conservation and role in angiogenesis of these naturally occurring combinations of receptors. We also show tissue-specific combinations and, in the case of choroid endothelial cells, consistency with both well-established and recent experimental findings, presented in a separate paper. The results and methods presented here advance the understanding of signaling to endothelial cells. The methods are generally applicable to the decoding of intercellular combinations of signals.

A patient diagnosed with multiple myeloma, bicuspid aortic valve, and Von Hippel-Lindau syndrome underwent whole-exome sequencing seeking a unified genetic cause for these three pathologies. The patient possessed a single-point mutation of arginine to cysteine (R24C) in the N-terminal region(pro-domain) of matrix metalloproteinase 9 (MMP-9). The pro-domain interacts with the catalytic site of this enzyme rendering it inactive. MMP-9 has previously been associated with all three pathologies suffered by the patient. We hypothesized that the observed mutation in the pro-domain would influence the activity of this enzyme. We expressed recombinant versions of MMP-9 and an investigation of their biochemical properties revealed that MMP-9 R24C is a constitutively active zymogen. To our knowledge, this is the first example of a mutation that discloses catalytic activity in the pro-form in any of the 24 human MMPs.

In the course of our studies aiming to discover vascular bed-specific endothelial cell (EC) mitogens, we identified leukemia inhibitory factor (LIF) as a mitogen for bovine choroidal EC (BCE), although LIF has been mainly characterized as an EC growth inhibitor and an anti-angiogenic molecule. LIF stimulated growth of BCE while it inhibited, as previously reported, bovine aortic EC (BAE) growth. The JAK-STAT3 pathway mediated LIF actions in both BCE and BAE cells, but a caspase-independent proapoptotic signal mediated by cathepsins was triggered in BAE but not in BCE. LIF administration directly promoted activation of STAT3 and increased blood vessel density in mouse eyes. LIF also had protective effects on the choriocapillaris in a model of oxidative retinal injury. Analysis of available single-cell transcriptomic datasets shows strong expression of the specific LIF receptor in mouse and human choroidal EC. Our data suggest that LIF administration may be an innovative approach to prevent atrophy associated with AMD, through protection of the choriocapillaris.

VEGF inhibitor drugs have been successful, especially in ophthalmology, but not all patients respond to them. Combinations of drugs are likely to be needed for a really effective therapy of angiogenesis-related diseases. In this paper we introduce a new concept, the comberon, a term named by analogy with the operon that refers to evolutionarily conserved combinations of co-expressed genes. These genes identify potential drug targets. Our results show that single-cell gene expression data can help to identify a set of co-expressed endothelial cell receptors, conserved among species (mice and human) and enriched, within a network, of connections to up-regulated genes. This set does include VEGF receptors and includes several other receptors previously shown to play a role in angiogenesis. The conclusions are supported by multiple highly significant statistical tests from large datasets, including an independent validation set. This discovery has broad pharmacological and socio-economic implications, going beyond angiogenesis therapy.

MOTIVATION:Analysis of singe cell RNA sequencing (scRNA-seq) typically consists of different steps including quality control, batch correction, clustering, cell identification and characterization, and visualization. The amount of scRNA-seq data is growing extremely fast, and novel algorithmic approaches improving these steps are key to extract more biological information. Here, we introduce: (i) two methods for automatic cell type identification (i.e., without expert curator) based on a voting algorithm and a Hopfield classifier, (ii) a method for cell anomaly quantification based on isolation forest, and (iii) a tool for the visualization of cell phenotypic landscapes based on Hopfield energy-like functions. These new approaches are integrated in a software platform that includes many other state-of-the-art methodologies and provides a self-contained toolkit for scRNA-seq analysis.RESULTS:We present a suite of software elements for the analysis of scRNA-seq data. This Python-based open source software, Digital Cell Sorter (DCS), consists in an extensive toolkit of methods for scRNA-seq analysis. We illustrate the capability of the software using data from large datasets of peripheral blood mononuclear cells (PBMC), as well as plasma cells of bone marrow samples from healthy donors and multiple myeloma patients. We test the novel algorithms by evaluating their ability to deconvolve cell mixtures and detect small numbers of anomalous cells in PBMC data.AVAILABILITY:The DCS toolkit is available for download and installation through the Python Package Index (PyPI). The software can be deployed using the Python import function following installation. Source code is also available for download on Zenodo: DOI 10.5281/zenodo.2533377.SUPPLEMENTARY INFORMATION:Supplemental Materials are available at PeerJ online.

The Hopfield neural network model is one of the simplest models able to mathematically implement Waddington’s interpretation of normal and anomalous cell phenotypes as dynamical attractors of epigenetic landscapes. Here, we propose a computational approach based on Hopfield’s associative memories that integrate gene expression data and gene interactome networks in (1) a model representing the dynamics and control of disease progression in multiple myeloma (MM), and (2) a model describing the control of angiogenesis. The MM model is built using single-cell RNA-seq data from bone marrow aspirates of MM patients as well as patients diagnosed with monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM), two medical conditions that often progress to full MM. We identify different clusters of MGUS, SMM, and MM cells, map them to Hopfield associative memory patterns, and model the dynamics of transition between the different patterns. The model is then used to identify combinations of genes whose simultaneous inhibition is associated with delayed disease progression. In the angiogenesis control model, we use single-cell RNA-seq data to predict specific combinations of targets for inhibition that could induce a cellular transition from tip-like to stalk-like endothelial cells, inhibiting the formation of capillary sprouts in blood vessels. The model generates novel hypotheses for combinations that could complement standard VEGF inhibitors and lead to a more efficient control of angiogenesis in cancer. Citation Format: Carlo Piermarocchi, Sergii Domanskyi, Alex Hakansson, Giovanni Paternostro. Modeling drug combination sensitivity with Hopfield networks and transcriptomics data [abstract]. In: Proceedings of the AACR Special Conference on the Evolving Landscape of Cancer Modeling; 2020 Mar 2-5; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2020;80(11 Suppl):Abstract nr A14.

The Hopfield neural network model is one of the simplest models able to mathematically implement Waddington’s interpretation of normal and anomalous cell phenotypes as dynamical attractors of epigenetic landscapes. Here, we propose a computational approach based on Hopfield’s associative memories that integrate gene expression data and gene interactome networks in (1) a model representing the dynamics and control of disease progression in multiple myeloma (MM), and (2) a model describing the control of angiogenesis. The MM model is built using single-cell RNA-seq data from bone marrow aspirates of MM patients as well as patients diagnosed with monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM), two medical conditions that often progress to full MM. We identify different clusters of MGUS, SMM, and MM cells, map them to Hopfield associative memory patterns, and model the dynamics of transition between the different patterns. The model is then used to identify combinations of genes whose simultaneous inhibition is associated with delayed disease progression. In the angiogenesis control model, we use single-cell RNA-seq data to predict specific combinations of targets for inhibition that could induce a cellular transition from tip-like to stalk-like endothelial cells, inhibiting the formation of capillary sprouts in blood vessels. The model generates novel hypotheses for combinations that could complement standard VEGF inhibitors and lead to a more efficient control of angiogenesis in cancer. Citation Format: Carlo Piermarocchi, Sergii Domanskyi, Alex Hakansson, Giovanni Paternostro. Modeling drug combination sensitivity with Hopfield networks and transcriptomics data [abstract]. In: Proceedings of the AACR Special Conference on the Evolving Landscape of Cancer Modeling; 2020 Mar 2-5; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2020;80(11 Suppl):Abstract nr A14.

Background Single cell RNA sequencing (scRNA-seq) brings unprecedented opportunities for mapping the heterogeneity of complex cellular environments such as bone marrow, and provides insight into many cellular processes. Single cell RNA-seq has a far larger fraction of missing data reported as zeros (dropouts) than traditional bulk RNA-seq, and unsupervised clustering combined with Principal Component Analysis (PCA) can be used to overcome this limitation. After clustering, however, one has to interpret the average expression of markers on each cluster to identify the corresponding cell types, and this is normally done by hand by an expert curator. Results We present a computational tool for processing single cell RNA-seq data that uses a voting algorithm to automatically identify cells based on approval votes received by known molecular markers. Using a stochastic procedure that accounts for imbalances in the number of known molecular signatures for different cell types, the method computes the statistical significance of the final approval score and automatically assigns a cell type to clusters without an expert curator. We demonstrate the utility of the tool in the analysis of eight samples of bone marrow from the Human Cell Atlas. The tool provides a systematic identification of cell types in bone marrow based on a list of markers of immune cell types, and incorporates a suite of visualization tools that can be overlaid on a t-SNE representation. The software is freely available as a Python package at https://github.com/sdomanskyi/DigitalCellSorter . Conclusions This methodology assures that extensive marker to cell type matching information is taken into account in a systematic way when assigning cell clusters to cell types. Moreover, the method allows for a high throughput processing of multiple scRNA-seq datasets, since it does not involve an expert curator, and it can be applied recursively to obtain cell sub-types. The software is designed to allow the user to substitute the marker to cell type matching information and apply the methodology to different cellular environments.

Associative memories in Hopfield's neural networks are mapped to gene expression pattern to model different paths of disease progression towards Multiple Myeloma (MM). The model is built using single cell RNA-seq data from bone marrow aspirates of MM patients as well as patients diagnosed with Monoclonal Gammopathy of Undetermined Significance (MGUS) and Smoldering Multiple Myeloma (SMM), two medical conditions that often progress to full MM. Results: We identify different clusters of MGUS, SMM, and MM cells, map them to Hopfield associative memory patterns, and model the dynamics of transition between the different patterns. The model is then used to identify genes that are differentialy expressed across different MM stages and whose simultaneous inhibition is associated to a delayed disease progression.

Modern time series gene expression and other omics data sets have enabled unprecedented resolution of the dynamics of cellular processes such as cell cycle and response to pharmaceutical compounds. In anticipation of the proliferation of time series data sets in the near future, we use the Hopfield model, a recurrent neural network based on spin glasses, to model the dynamics of cell cycle in HeLa (human cervical cancer) and S. cerevisiae cells. We study some of the rich dynamical properties of these cyclic Hopfield systems, including the ability of populations of simulated cells to recreate experimental expression data and the effects of noise on the dynamics. Next, we use a genetic algorithm to identify sets of genes which, when selectively inhibited by local external fields representing gene silencing compounds such as kinase inhibitors, disrupt the encoded cell cycle. We find, for example, that inhibiting the set of four kinases AURKB, NEK1, TTK, and WEE1 causes simulated HeLa cells to accumulate in the M phase. Finally, we suggest possible improvements and extensions to our model.

Pancreatic β-cell lipotoxicity is a central feature of the pathogenesis of type 2 diabetes. To study the mechanism by which fatty acids cause β-cell death and develop novel approaches to prevent it, a high-throughput screen on the β-cell line INS1 was carried out. The cells were exposed to palmitate to induce cell death and compounds that reversed palmitate-induced cytotoxicity were ascertained. Hits from the screen were analyzed by an increasingly more stringent testing funnel, ending with studies on primary human islets treated with palmitate. MAP4K4 inhibitors, which were not part of the screening libraries but were ascertained by a bioinformatics analysis, and the endocannabinoid anandamide were effective at inhibiting palmitate-induced apoptosis in INS1 cells as well as primary rat and human islets. These targets could serve as the starting point for the development of therapeutics for type 2 diabetes.

Abstract Introduction: Acute myeloid leukemia (AML) remains a challenging malignancy to treat, with high mortality despite recent advances in cancer care. The mainstay of therapy is intensive chemotherapy including anthracyclines and antimetabolites, with or without allogeneic bone marrow transplant. The standard of care has not changed significantly in decades, and treatment options are limited for patients who do not respond to induction, or who relapse. Although the genetic and molecular diversity of AML is well recognized, the effective integration of targeted therapies into treatment regimens has been difficult. The current study evaluates the in vitro activity of an array of antineoplastic agents in AML, with the goal of identifying drug sensitivity patterns that may help guide therapies based on mutational and cytogenetic profiling. Methods: 51 patients with AML were enrolled in the study from September 2013 to July 2016 under an IRB approved consenting process at Scripps Health. Samples were collected at both diagnosis and relapse if available. Both bone marrow and peripheral blood were accepted, with a requirement of significant circulating blasts if peripheral blood was used. Samples were tested for common biomarkers and cytogenetic abnormalities. 2,500 cells per well were transferred onto tissue culture treated plates. Drugs with potential antileukemic activity were added to each well in concentrations of 0.1, 1.0, and 10.0 μM. The cells were incubated with the drugs for 96 hours at 37oC. Cell viability was measured and reported as a percentage of plate-specific controls incubated with dimethylsulfoxide alone. Results: Drugs were grouped by therapeutic class. In vitro responses to anthracyclines and antimetabolites were noted across all mutational subtypes of AML. BCL-2 inhibitors and the histone deacetylase inhibitor romidepsin showed significant in vitro antileukemic activity across all subtypes. Proteasome inhibitors almost universally showed robust in vitro activity, even at the lowest drug concentration. The drug pevonedistat, a selective small-molecule inhibitor of NEDD8-activating enzyme, had significant differential activity depending on mutational status. FLT3-ITD mutations conferred sensitivity to the molecule, while mutations in NPM1 appeared to confer resistance. For the 10.0 μM concentration of pevonedistat, the average cell viability was 149.8% vs 37.3% (P=0.02) for the NPM1 mutated samples vs the FLT3-ITD mutated samples respectively. Conclusion: This in vitro assay demonstrates the ability to rapidly determine sensitivity of human AML cells to a wide variety of antileukemic drugs. Limitations include the fixed concentrations used across all medications which allowed comparisons between patients, but limits comparison of drug efficacy for an individual patient. The mechanism of some medications, such as hypomethylating agents and tretinoin, may require longer duration of exposure, thus confounding interpretation of the 96-hour viability results. Despite these limitations, we were able to find interesting patterns of responses across a wide spectrum of AML types. Anticipated applications of the assay include experimentation with novel drug combinations, directing in vivo clinical studies, and informing individualized treatment decisions in AML. Disclosures No relevant conflicts of interest to declare.

The diverse, specialized genes present in today's lifeforms evolved from a common core of ancient, elementary genes. However, these genes did not evolve individually: gene expression is controlled by a complex network of interactions, and alterations in one gene may drive reciprocal changes in its proteins' binding partners. Like many complex networks, these gene regulatory networks (GRNs) are composed of communities, or clusters of genes with relatively high connectivity. A deep understanding of the relationship between the evolutionary history of single genes and the topological properties of the underlying GRN is integral to evolutionary genetics. Here, we show that the topological properties of an acute myeloid leukemia GRN and a general human GRN are strongly coupled with its genes' evolutionary properties. Slowly evolving ("cold"), old genes tend to interact with each other, as do rapidly evolving ("hot"), young genes. This naturally causes genes to segregate into community structures with relatively homogeneous evolutionary histories. We argue that gene duplication placed old, cold genes and communities at the center of the networks, and young, hot genes and communities at the periphery. We demonstrate this with single-node centrality measures and two new measures of efficiency, the set efficiency and the interset efficiency. We conclude that these methods for studying the relationships between a GRN's community structures and its genes' evolutionary properties provide new perspectives for understanding evolutionary genetics.

Contemporary cancer treatment is advancing toward individualized therapy and recent developments have utilized treatment based on each patient's distinct tumor biology. The distinct biology of each patient's individual malignancy has begun to affect therapeutic decisions: i.e., the use of biomarker data such as FLT-3 and NPM1 mutational status in order to identify high and low risk patients in acute myeloid leukemia; or the use of genetic expression assays in early stage invasive breast cancer. However, the treatment of acute myeloid leukemia, as in most malignancies, remains focused on therapy derived from outcome data from large scale clinical trials and not individual tumor biology.

Abstract Introduction and purpose of the study: Significant progress has been made in biological network reconstruction methods in recent years, with much emphasis placed on investigating the topology of gene regulatory networks (GRNs). While studying a network's topology can provide useful biological information, a dynamical model for how genes and proteins exchange information is needed in order to understand and predict a cell's response to stimuli such as drugs which inhibit the activity of a protein. Two thirds of patients with acute myeloid leukemia (AML) have an unfavorable prognosis. This is in part due to the high tumor cell heterogeneity in AML, which is often the ultimate cause of drug resistance and relapse in patients. Mathematical models of signaling dynamics in AML which account for the heterogeneity of clonally derived cells in a tumor could be valuable new tools for designing effective, original therapies. Heterogeneity can be described in signaling by defining nonlinear models with multiple attractor states. Novel computational methods: We present two mathematical signaling models that encode real gene expression measurements as attractors in a directed AML GRN. Gene expression profiles obtained from RNA-seq in normal progenitor and AML cells data are used to define a set of robust gene expression profiles corresponding to different clonal states. The first model is Boolean, and is based on an asymmetric Hopfield model with multiple memory patterns (Szedlak et al. 2014). The second model uses continuous expression values in which the elements of the GRN interact like oscillators characterized by multi-stability. The equations of motion for the oscillators are defined in such a way that the expression profiles for the RNA-seq data match the multiple steady states of the signaling network. Summary of new data: We examine the sensitivity of normal and cancer attractors to single gene perturbations and to real gene-inhibiting drugs in both models. For the network topology we use a recently developed network that is specific for AML, specifically AML 2.3 (Ong et al. 2014). Attractors are defined using RNA-seq data from AML cells and hematopoietic controls (Macrae et al. 2013). According to the Hopfield model, we found that ELAVL1, GATA1, IRX5, MYOG, RXRA, and TFEB are genes in AML which, when inhibited, strongly destabilize the cancer attractor. In the continuum model, we explicitly included interactions between drugs currently in AML clinical trials and their targets. Focusing on known targets of lenalidomide and sorafenib, we found that FLT1, KDR, and PDGFRB are associated with the strongest sensitivity to perturbations. In addition, we found that besides the direct targets of these drugs (BRAF, RAF1, FLT4, KDR, FLT3, PDGFRB, KIT, FGFR1, RET, FLT1 for sorafenib, and TNF, TNFSF11, CDH5, PTGS2, CRBN for lenalidomide), RPS18, RPS11, RPS3, and TPT1 are also strongly indirectly down-regulated. Conclusions: We developed two novel methods of network signaling that encode gene expression patterns of cells under varying conditions as attractor states. The methods were applied to AML to identify a set of genes whose perturbation is associated with strong deviations from cancer conditions. Citation Format: Anthony D. Szedlak, Giovanni Paternostro, Carlo Piermarocchi. Many-attractor models of signaling dynamics and heterogeneity in acute myeloid leukemia gene regulatory networks. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr B2-40.

The diverse, specialized genes in today's lifeforms evolved from a common core of ancient, elementary genes. However, these genes did not evolve individually: gene expression is controlled by a complex network of interactions, and alterations in one gene may drive reciprocal changes in its proteins' binding partners. We show that the topology of a leukemia gene regulatory network is strongly coupled with evolutionary properties. Slowly-evolving ("cold"), old genes tend to interact with each other, as do rapidly-evolving ("hot"), young genes, causing genes to evolve in clusters. We argue that gene duplication placed old, cold genes at the center of the network, and young, hot genes on the periphery, and demonstrate this with single-node centrality measures and two new measures of efficiency. Integrating centrality measures with evolutionary information, we define a medically-relevant "cancer network core," strongly enriched for common cancer mutations (p=2× 10^-14). This could aid in identifying driver mutations and therapeutic targets.

Cell-based therapies to treat skeletal muscle disease are limited by the poor survival of donor myoblasts, due in part to acute hypoxic stress. After confirming that the microenvironment of transplanted myoblasts is hypoxic, we screened a kinase inhibitor library in vitro and identified five kinase inhibitors that protected myoblasts from cell death or growth arrest in hypoxic conditions. A systematic, combinatorial study of these compounds further improved myoblast viability, showing both synergistic and additive effects. Pathway and target analysis revealed CDK5, CDK2, CDC2, WEE1, and GSK3β as the main target kinases. In particular, CDK5 was the center of the target kinase network. Using our recently developed statistical method based on elastic net regression we computationally validated the key role of CDK5 in cell protection against hypoxia. This method provided a list of potential kinase targets with a quantitative measure of their optimal amount of relative inhibition. A modified version of the method was also able to predict the effect of combinations using single-drug response data. This work is the first step towards a broadly applicable system-level strategy for the pharmacology of hypoxic damage.

Gene regulatory network inference uses genome-wide transcriptome measurements in response to genetic, environmental, or dynamic perturbations to predict causal regulatory influences between genes. We hypothesized that evolution also acts as a suitable network perturbation and that integration of data from multiple closely related species can lead to improved reconstruction of gene regulatory networks. To test this hypothesis, we predicted networks from temporal gene expression data for 3,610 genes measured during early embryonic development in six Drosophila species and compared predicted networks to gold standard networks of ChIP-chip and ChIP-seq interactions for developmental transcription factors in five species. We found that (i) the performance of single-species networks was independent of the species where the gold standard was measured; (ii) differences between predicted networks reflected the known phylogeny and differences in biology between the species; (iii) an integrative consensus network that minimized the total number of edge gains and losses with respect to all single-species networks performed better than any individual network. Our results show that in an evolutionarily conserved system, integration of data from comparable experiments in multiple species improves the inference of gene regulatory networks. They provide a basis for future studies on the numerous multispecies gene expression datasets for other biological processes available in the literature.

BACKGROUND:Many kinase inhibitors have been approved as cancer therapies. Recently, libraries of kinase inhibitors have been extensively profiled, thus providing a map of the strength of action of each compound on a large number of its targets. These profiled libraries define drug-kinase networks that can predict the effectiveness of untested drugs and elucidate the roles of specific kinases in different cellular systems. Predictions of drug effectiveness based on a comprehensive network model of cellular signalling are difficult, due to our partial knowledge of the complex biological processes downstream of the targeted kinases.RESULTS:We have developed the Kinase Inhibitors Elastic Net (KIEN) method, which integrates information contained in drug-kinase networks with in vitro screening. The method uses the in vitro cell response of single drugs and drug pair combinations as a training set to build linear and nonlinear regression models. Besides predicting the effectiveness of untested drugs, the KIEN method identifies sets of kinases that are statistically associated to drug sensitivity in a given cell line. We compared different versions of the method, which is based on a regression technique known as elastic net. Data from two-drug combinations led to predictive models, and we found that predictivity can be improved by applying logarithmic transformation to the data. The method was applied to the A549 lung cancer cell line, and we identified specific kinases known to have an important role in this type of cancer (TGFBR2, EGFR, PHKG1 and CDK4). A pathway enrichment analysis of the set of kinases identified by the method showed that axon guidance, activation of Rac, and semaphorin interactions pathways are associated to a selective response to therapeutic intervention in this cell line.CONCLUSIONS:We have proposed an integrated experimental and computational methodology, called KIEN, that identifies the role of specific kinases in the drug response of a given cell line. The method will facilitate the design of new kinase inhibitors and the development of therapeutic interventions with combinations of many inhibitors.

Acute myeloid leukemia (AML) is a highly heterogeneous disorder characterized by the rapid clonal proliferation of blasts derived from hematopoietic progenitor cells, leading to failure of normal hematopoiesis. Although standard therapy, usually including idarubicin and cytarabine, has been used to achieve remission, the long-term survival rates remain low.