Predicting drug polypharmacology from cell morphology readouts using variational autoencoder latent space arithmetic

A variational autoencoder (VAE) is a machine learning algorithm, useful for generating a compressed and interpretable latent space. These representations have been generated from various biomedical data types and can be used to produce realistic-looking simulated data. However, standard vanilla VAEs suffer from entangled and uninformative latent spaces, which can be mitigated using other types of VAEs such as β-VAE and MMD-VAE. In this project, we evaluated the ability of VAEs to learn cell morphology characteristics derived from cell images. We trained and evaluated these three VAE variants—Vanilla VAE, β-VAE, and MMD-VAE—on cell morphology readouts and explored the generative capacity of each model to predict compound polypharmacology (the interactions of a drug with more than one target) using an approach called latent space arithmetic (LSA). To test the generalizability of the strategy, we also trained these VAEs using gene expression data of the same compound perturbations and found that gene expression provides complementary information. We found that the β-VAE and MMD-VAE disentangle morphology signals and reveal a more interpretable latent space. We reliably simulated morphology and gene expression readouts from certain compounds thereby predicting cell states perturbed with compounds of known polypharmacology. Inferring cell state for specific drug mechanisms could aid researchers in developing and identifying targeted therapeutics and categorizing off-target effects in the future.     

Phylogenomic analysis of gene co‐expression networks reveals the evolution of functional modules

Molecular evolutionary studies correlate genomic and phylogenetic information with the emergence of new traits of organisms. These traits are, however, the consequence of dynamic gene networks composed of functional modules, which might not be captured by genomic analyses. Here, we established a method that combines large‐scale genomic and phylogenetic data with gene co‐expression networks to extensively study the evolutionary make‐up of modules in the moss Physcomitrella patens, and in the angiosperms Arabidopsis thaliana and Oryza sativa (rice). We first show that younger genes are less annotated than older genes. By mapping genomic data onto the co‐expression networks, we found that genes from the same evolutionary period tend to be connected, whereas old and young genes tend to be disconnected. Consequently, the analysis revealed modules that emerged at a specific time in plant evolution. To uncover the evolutionary relationships of the modules that are conserved across the plant kingdom, we added phylogenetic information that revealed duplication and speciation events on the module level. This combined analysis revealed an independent duplication of cell wall modules in bryophytes and angiosperms, suggesting a parallel evolution of cell wall pathways in land plants. We provide an online tool allowing plant researchers to perform these analyses at http://www.gene2function.de .     

Integrating diverse genomic data using gene sets

We introduce and evaluate data analysis methods to interpret simultaneous measurement of multiple genomic features made on the same biological samples. Our tools use gene sets to provide an interpretable common scale for diverse genomic information. We show we can detect genetic effects, although they may act through different mechanisms in different samples, and show we can discover and validate important disease-related gene sets that would not be discovered by analyzing each data type individually.     

Discovery and expansion of gene modules by seeking isolated groups in a random graph process

Background

A central problem in systems biology research is the identification and extension of biological modules–groups of genes or proteins participating in a common cellular process or physical complex. As a result, there is a persistent need for practical, principled methods to infer the modular organization of genes from genome-scale data.

Results

We introduce a novel approach for the identification of modules based on the persistence of isolated gene groups within an evolving graph process. First, the underlying genomic data is summarized in the form of ranked gene–gene relationships, thereby accommodating studies that quantify the relevant biological relationship directly or indirectly. Then, the observed gene–gene relationship ranks are viewed as the outcome of a random graph process and candidate modules are given by the identifiable subgraphs that arise during this process. An isolation index is computed for each module, which quantifies the statistical significance of its survival time.

Conclusions

The Miso (module isolation) method predicts gene modules from genomic data and the associated isolation index provides a module-specific measure of confidence. Improving on existing alternative, such as graph clustering and the global pruning of dendrograms, this index offers two intuitively appealing features: (1) the score is module-specific; and (2) different choices of threshold correlate logically with the resulting performance, i.e. a stringent cutoff yields high quality predictions, but low sensitivity. Through the analysis of yeast phenotype data, the Miso method is shown to outperform existing alternatives, in terms of the specificity and sensitivity of its predictions.     

Joint Microbial and Metabolomic Network Estimation with the Censored Gaussian Graphical Model

Joint analysis of microbiome and metabolomic data represents an imperative objective as the field moves beyond basic microbiome association studies and turns towards mechanistic and translational investigations. We present a censored Gaussian graphical model framework, where the metabolomic data are treated as continuous and the microbiome data as censored at zero, to identify direct interactions (defined as conditional dependence relationships) between microbial species and metabolites. Simulated examples show that our method metaMint performs favorably compared to the existing ones. metaMint also provides interpretable microbe-metabolite interactions when applied to a bacterial vaginosis data set. R implementation of metaMint is available on GitHub.

     

Testing gene set enrichment for subset of genes: Sub-GSE

Background  

Many methods have been developed to test the enrichment of genes related to certain phenotypes or cell states in gene sets. These approaches usually combine gene expression data with functionally related gene sets as defined in databases such as GeneOntology (GO), KEGG, or BioCarta. The results based on gene set analysis are generally more biologically interpretable, accurate and robust than the results based on individual gene analysis. However, while most available methods for gene set enrichment analysis test the enrichment of the entire gene set, it is more likely that only a subset of the genes in the gene set may be related to the phenotypes of interest.     

Multiscale Embedded Gene Co-expression Network Analysis

Gene co-expression network analysis has been shown effective in identifying functional co-expressed gene modules associated with complex human diseases. However, existing techniques to construct co-expression networks require some critical prior information such as predefined number of clusters, numerical thresholds for defining co-expression/interaction, or do not naturally reproduce the hallmarks of complex systems such as the scale-free degree distribution of small-worldness. Previously, a graph filtering technique called Planar Maximally Filtered Graph (PMFG) has been applied to many real-world data sets such as financial stock prices and gene expression to extract meaningful and relevant interactions. However, PMFG is not suitable for large-scale genomic data due to several drawbacks, such as the high computation complexity O(|V|³), the presence of false-positives due to the maximal planarity constraint, and the inadequacy of the clustering framework. Here, we developed a new co-expression network analysis framework called Multiscale Embedded Gene Co-expression Network Analysis (MEGENA) by: i) introducing quality control of co-expression similarities, ii) parallelizing embedded network construction, and iii) developing a novel clustering technique to identify multi-scale clustering structures in Planar Filtered Networks (PFNs). We applied MEGENA to a series of simulated data and the gene expression data in breast carcinoma and lung adenocarcinoma from The Cancer Genome Atlas (TCGA). MEGENA showed improved performance over well-established clustering methods and co-expression network construction approaches. MEGENA revealed not only meaningful multi-scale organizations of co-expressed gene clusters but also novel targets in breast carcinoma and lung adenocarcinoma.     

Genomic duplication problems for unrooted gene trees

Background

Discovering the location of gene duplications and multiple gene duplication episodes is a fundamental issue in evolutionary molecular biology. The problem introduced by Guigó et al. in 1996 is to map gene duplication events from a collection of rooted, binary gene family trees onto theirs corresponding rooted binary species tree in such a way that the total number of multiple gene duplication episodes is minimized. There are several models in the literature that specify how gene duplications from gene families can be interpreted as one duplication episode. However, in all duplication episode problems gene trees are rooted. This restriction limits the applicability, since unrooted gene family trees are frequently inferred by phylogenetic methods.

Results

In this article we show the first solution to the open problem of episode clustering where the input gene family trees are unrooted. In particular, by using theoretical properties of unrooted reconciliation, we show an efficient algorithm that reduces this problem into the episode clustering problems defined for rooted trees. We show theoretical properties of the reduction algorithm and evaluation of empirical datasets.

Conclusions

We provided algorithms and tools that were successfully applied to several empirical datasets. In particular, our comparative study shows that we can improve known results on genomic duplication inference from real datasets.

     

When Is Hub Gene Selection Better than Standard Meta-Analysis?

Since hub nodes have been found to play important roles in many networks, highly connected hub genes are expected to play an important role in biology as well. However, the empirical evidence remains ambiguous. An open question is whether (or when) hub gene selection leads to more meaningful gene lists than a standard statistical analysis based on significance testing when analyzing genomic data sets (e.g., gene expression or DNA methylation data). Here we address this question for the special case when multiple genomic data sets are available. This is of great practical importance since for many research questions multiple data sets are publicly available. In this case, the data analyst can decide between a standard statistical approach (e.g., based on meta-analysis) and a co-expression network analysis approach that selects intramodular hubs in consensus modules. We assess the performance of these two types of approaches according to two criteria. The first criterion evaluates the biological insights gained and is relevant in basic research. The second criterion evaluates the validation success (reproducibility) in independent data sets and often applies in clinical diagnostic or prognostic applications. We compare meta-analysis with consensus network analysis based on weighted correlation network analysis (WGCNA) in three comprehensive and unbiased empirical studies: (1) Finding genes predictive of lung cancer survival, (2) finding methylation markers related to age, and (3) finding mouse genes related to total cholesterol. The results demonstrate that intramodular hub gene status with respect to consensus modules is more useful than a meta-analysis p-value when identifying biologically meaningful gene lists (reflecting criterion 1). However, standard meta-analysis methods perform as good as (if not better than) a consensus network approach in terms of validation success (criterion 2). The article also reports a comparison of meta-analysis techniques applied to gene expression data and presents novel R functions for carrying out consensus network analysis, network based screening, and meta analysis.     

Discovery of multi-dimensional modules by integrative analysis of cancer genomic data

Recent technology has made it possible to simultaneously perform multi-platform genomic profiling (e.g. DNA methylation (DM) and gene expression (GE)) of biological samples, resulting in so-called ‘multi-dimensional genomic data’. Such data provide unique opportunities to study the coordination between regulatory mechanisms on multiple levels. However, integrative analysis of multi-dimensional genomics data for the discovery of combinatorial patterns is currently lacking. Here, we adopt a joint matrix factorization technique to address this challenge. This method projects multiple types of genomic data onto a common coordinate system, in which heterogeneous variables weighted highly in the same projected direction form a multi-dimensional module (md-module). Genomic variables in such modules are characterized by significant correlations and likely functional associations. We applied this method to the DM, GE, and microRNA expression data of 385 ovarian cancer samples from the The Cancer Genome Atlas project. These md-modules revealed perturbed pathways that would have been overlooked with only a single type of data, uncovered associations between different layers of cellular activities and allowed the identification of clinically distinct patient subgroups. Our study provides an useful protocol for uncovering hidden patterns and their biological implications in multi-dimensional ‘omic’ data.     

Enhanced Antitumor Efficacy of an Oncolytic Herpes Simplex Virus Expressing an Endostatin–Angiostatin Fusion Gene in Human Glioblastoma Stem Cell Xenografts

Viruses have demonstrated strong potential for the therapeutic targeting of glioblastoma stem cells (GSCs). In this study, the use of a herpes simplex virus carrying endostatin–angiostatin (VAE) as a novel therapeutic targeting strategy for glioblastoma-derived cancer stem cells was investigated. We isolated six stable GSC-enriched cultures from 36 human glioblastoma specimens and selected one of the stable GSCs lines for establishing GSC-carrying orthotopic nude mouse models. The following results were obtained: (a) VAE rapidly proliferated in GSCs and expressed endo–angio in vitro and in vivo 48 h and 10 d after infection, respectively; (b) compared with the control gliomas treated with rHSV or Endostar, the subcutaneous gliomas derived from the GSCs showed a significant reduction in microvessel density after VAE treatment; (c) compared with the control, a significant improvement was observed in the length of the survival of mice with intracranial and subcutaneous gliomas treated with VAE; (d) MRI analysis showed that the tumor volumes of the intracranial gliomas generated by GSCs remarkably decreased after 10 d of VAE treatment compared with the controls. In conclusion, VAE demonstrated oncolytic therapeutic efficacy in animal models of human GSCs and expressed an endostatin–angiostatin fusion gene, which enhanced antitumor efficacy most likely by restricting tumor microvasculature development.