Data-driven human transcriptomic modules determined by independent component analysis

Exploring combinations of dimensionality reduction, transfer learning, and regularization methods for predicting binary phenotypes with transcriptomic data

BACKGROUND

Numerous transcriptomic-based models have been developed to predict or understand the fundamental mechanisms driving biological phenotypes. However, few models have successfully transitioned into clinical practice due to challenges associated with generalizability and interpretability. To address these issues, researchers have turned to dimensionality reduction methods and have begun implementing transfer learning approaches.

METHODS

In this study, we aimed to determine the optimal combination of dimensionality reduction and regularization methods for predictive modeling. We applied seven dimensionality reduction methods to various datasets, including two supervised methods (linear optimal low-rank projection and low-rank canonical correlation analysis), two unsupervised methods [principal component analysis and consensus independent component analysis (c-ICA)], and three methods [autoencoder (AE), adversarial variational autoencoder, and c-ICA] within a transfer learning framework, trained on > 140,000 transcriptomic profiles. To assess the performance of the different combinations, we used a cross-validation setup encapsulated within a permutation testing framework, analyzing 30 different transcriptomic datasets with binary phenotypes. Furthermore, we included datasets with small sample sizes and phenotypes of varying degrees of predictability, and we employed independent datasets for validation.

RESULTS

Our findings revealed that regularized models without dimensionality reduction achieved the highest predictive performance, challenging the necessity of dimensionality reduction when the primary goal is to achieve optimal predictive performance. However, models using AE and c-ICA with transfer learning for dimensionality reduction showed comparable performance, with enhanced interpretability and robustness of predictors, compared to models using non-dimensionality-reduced data.

CONCLUSION

These findings offer valuable insights into the optimal combination of strategies for enhancing the predictive performance, interpretability, and generalizability of transcriptomic-based models.

CoRegNet: unraveling gene co-regulation networks from public RNA-Seq repositories using a beta-binomial statistical model

Millions of RNA sequencing samples have been deposited into public databases, providing a rich resource for biological research. These datasets encompass tens of thousands of experiments and offer comprehensive insights into human cellular regulation. However, a major challenge is how to integrate these experiments that acquired at different conditions. We propose a new statistical tool based on beta-binomial distributions that can construct robust gene co-regulation network (CoRegNet) across tens of thousands of experiments. Our analysis of over 12 000 experiments involving human tissues and cells shows that CoRegNet significantly outperforms existing gene co-expression-based methods. Although the majority of the genes are linearly co-regulated, we did discover an interesting set of genes that are non-linearly co-regulated; half of the time they change in the same direction and the other half they change in the opposite direction. Additionally, we identified a set of gene pairs that follows the Simpson's paradox. By utilizing public domain data, CoRegNet offers a powerful approach for identifying functionally related gene pairs, thereby revealing new biological insights.

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.

Meta-Analysis of Esophageal Cancer Transcriptomes Using Independent Component Analysis

Independent Component Analysis is a matrix factorization method for data dimension reduction. ICA has been widely applied for the analysis of transcriptomic data for blind separation of biological, environmental, and technical factors affecting gene expression. The study aimed to analyze the publicly available esophageal cancer data using the ICA for identification and comprehensive analysis of reproducible signaling pathways and molecular signatures involved in this cancer type. In this study, four independent esophageal cancer transcriptomic datasets from GEO databases were used. A bioinformatics tool « BiODICA-Independent Component Analysis of Big Omics Data» was applied to compute independent components (ICs). Gene Set Enrichment Analysis (GSEA) and ToppGene uncovered the most significantly enriched pathways. Construction and visualization of gene networks and graphs were performed using the Cytoscape, and HPRD database. The correlation graph between decompositions into 30 ICs was built with absolute correlation values exceeding 0.3. Clusters of components-pseudocliques were observed in the structure of the correlation graph. The top 1,000 most contributing genes of each ICs in the pseudocliques were mapped to the PPI network to construct associated signaling pathways. Some cliques were composed of densely interconnected nodes and included components common to most cancer types (such as cell cycle and extracellular matrix signals), while others were specific to EC. The results of this investigation may reveal potential biomarkers of esophageal carcinogenesis, functional subsystems dysregulated in the tumor cells, and be helpful in predicting the early development of a tumor.

Development of a fixed module repertoire for the analysis and interpretation of blood transcriptome data

As the capacity for generating large-scale molecular profiling data continues to grow, the ability to extract meaningful biological knowledge from it remains a limitation. Here, we describe the development of a new fixed repertoire of transcriptional modules, BloodGen3, that is designed to serve as a stable reusable framework for the analysis and interpretation of blood transcriptome data. The construction of this repertoire is based on co-clustering patterns observed across sixteen immunological and physiological states encompassing 985 blood transcriptome profiles. Interpretation is supported by customized resources, including module-level analysis workflows, fingerprint grid plot visualizations, interactive web applications and an extensive annotation framework comprising functional profiling reports and reference transcriptional profiles. Taken together, this well-characterized and well-supported transcriptional module repertoire can be employed for the interpretation and benchmarking of blood transcriptome profiles within and across patient cohorts. Blood transcriptome fingerprints for the 16 reference cohorts can be accessed interactively via:  https://drinchai.shinyapps.io/BloodGen3Module/ .

BloodGen3Module: Blood transcriptional module repertoire analysis and visualization using R

Motivation

We previously described the construction and characterization of fixed reusable blood transcriptional module repertoires. More recently we released a third iteration (‘BloodGen3’ module repertoire) that comprises 382 functionally annotated modules and encompasses 14 168 transcripts. Custom bioinformatic tools are needed to support downstream analysis, visualization and interpretation relying on such fixed module repertoires.

Results

We have developed and describe here an R package, BloodGen3Module. The functions of our package permit group comparison analyses to be performed at the module-level, and to display the results as annotated fingerprint grid plots. A parallel workflow for computing module repertoire changes for individual samples rather than groups of samples is also available; these results are displayed as fingerprint heatmaps. An illustrative case is used to demonstrate the steps involved in generating blood transcriptome repertoire fingerprints of septic patients. Taken together, this resource could facilitate the analysis and interpretation of changes in blood transcript abundance observed across a wide range of pathological and physiological states.

Availability and implementation

The BloodGen3Module package and documentation are freely available from Github: https://github.com/Drinchai/BloodGen3Module.

Supplementary information

Supplementary data are available at Bioinformatics online.

Correcting for experiment-specific variability in expression compendia can remove underlying signals

MOTIVATION

In the past two decades, scientists in different laboratories have assayed gene expression from millions of samples. These experiments can be combined into compendia and analyzed collectively to extract novel biological patterns. Technical variability, or "batch effects," may result from combining samples collected and processed at different times and in different settings. Such variability may distort our ability to extract true underlying biological patterns. As more integrative analysis methods arise and data collections get bigger, we must determine how technical variability affects our ability to detect desired patterns when many experiments are combined.

OBJECTIVE

We sought to determine the extent to which an underlying signal was masked by technical variability by simulating compendia comprising data aggregated across multiple experiments.

METHOD

We developed a generative multi-layer neural network to simulate compendia of gene expression experiments from large-scale microbial and human datasets. We compared simulated compendia before and after introducing varying numbers of sources of undesired variability.

RESULTS

The signal from a baseline compendium was obscured when the number of added sources of variability was small. Applying statistical correction methods rescued the underlying signal in these cases. However, as the number of sources of variability increased, it became easier to detect the original signal even without correction. In fact, statistical correction reduced our power to detect the underlying signal.

CONCLUSION

When combining a modest number of experiments, it is best to correct for experiment-specific noise. However, when many experiments are combined, statistical correction reduces our ability to extract underlying patterns.

Application of Transcriptional Gene Modules to Analysis of Caenorhabditis elegans' Gene Expression Data

Identification of co-expressed sets of genes (gene modules) is used widely for grouping functionally related genes during transcriptomic data analysis. An organism-wide atlas of high-quality gene modules would provide a powerful tool for unbiased detection of biological signals from gene expression data. Here, using a method based on independent component analysis we call DEXICA, we have defined and optimized 209 modules that broadly represent transcriptional wiring of the key experimental organism C. elegans. These modules represent responses to changes in the environment (e.g., starvation, exposure to xenobiotics), genes regulated by transcriptions factors (e.g., ATFS-1, DAF-16), genes specific to tissues (e.g., neurons, muscle), genes that change during development, and other complex transcriptional responses to genetic, environmental and temporal perturbations. Interrogation of these modules reveals processes that are activated in long-lived mutants in cases where traditional analyses of differentially expressed genes fail to do so. Additionally, we show that modules can inform the strength of the association between a gene and an annotation (e.g., GO term). Analysis of “module-weighted annotations” improves on several aspects of traditional annotation-enrichment tests and can aid in functional interpretation of poorly annotated genes. We provide an online interactive resource with tutorials at http://genemodules.org/, in which users can find detailed information on each module, check genes for module-weighted annotations, and use both of these to analyze their own gene expression data (generated using any platform) or gene sets of interest.

Responsible, practical genomic data sharing that accelerates research

Data sharing anchors reproducible science, but expectations and best practices are often nebulous. Communities of funders, researchers and publishers continue to grapple with what should be required or encouraged. To illuminate the rationales for sharing data, the technical challenges and the social and cultural challenges, we consider the stakeholders in the scientific enterprise. In biomedical research, participants are key among those stakeholders. Ethical sharing requires considering both the value of research efforts and the privacy costs for participants. We discuss current best practices for various types of genomic data, as well as opportunities to promote ethical data sharing that accelerates science by aligning incentives.

A modular framework for the development of targeted Covid-19 blood transcript profiling panels

BACKGROUND

Covid-19 morbidity and mortality are associated with a dysregulated immune response. Tools are needed to enhance existing immune profiling capabilities in affected patients. Here we aimed to develop an approach to support the design of targeted blood transcriptome panels for profiling the immune response to SARS-CoV-2 infection.

METHODS

We designed a pool of candidates based on a pre-existing and well-characterized repertoire of blood transcriptional modules. Available Covid-19 blood transcriptome data was also used to guide this process. Further selection steps relied on expert curation. Additionally, we developed several custom web applications to support the evaluation of candidates.

RESULTS

As a proof of principle, we designed three targeted blood transcript panels, each with a different translational connotation: immunological relevance, therapeutic development relevance and SARS biology relevance.

CONCLUSION

Altogether the work presented here may contribute to the future expansion of immune profiling capabilities via targeted profiling of blood transcript abundance in Covid-19 patients.

Data Integration for Immunology

Over the last several years, next-generation sequencing and its recent push toward single-cell resolution have transformed the landscape of immunology research by revealing novel complexities about all components of the immune system. With the vast amounts of diverse data currently being generated, and with the methods of analyzing and combining diverse data improving as well, integrative systems approaches are becoming more powerful. Previous integrative approaches have combined multiple data types and revealed ways that the immune system, both as a whole and as individual parts, is affected by genetics, the microbiome, and other factors. In this review, we explore the data types that are available for studying immunology with an integrative systems approach, as well as the current strategies and challenges for conducting such analyses.

Compressing gene expression data using multiple latent space dimensionalities learns complementary biological representations

BACKGROUND

Unsupervised compression algorithms applied to gene expression data extract latent or hidden signals representing technical and biological sources of variation. However, these algorithms require a user to select a biologically appropriate latent space dimensionality. In practice, most researchers fit a single algorithm and latent dimensionality. We sought to determine the extent by which selecting only one fit limits the biological features captured in the latent representations and, consequently, limits what can be discovered with subsequent analyses.

RESULTS

We compress gene expression data from three large datasets consisting of adult normal tissue, adult cancer tissue, and pediatric cancer tissue. We train many different models across a large range of latent space dimensionalities and observe various performance differences. We identify more curated pathway gene sets significantly associated with individual dimensions in denoising autoencoder and variational autoencoder models trained using an intermediate number of latent dimensionalities. Combining compressed features across algorithms and dimensionalities captures the most pathway-associated representations. When trained with different latent dimensionalities, models learn strongly associated and generalizable biological representations including sex, neuroblastoma MYCN amplification, and cell types. Stronger signals, such as tumor type, are best captured in models trained at lower dimensionalities, while more subtle signals such as pathway activity are best identified in models trained with more latent dimensionalities.

CONCLUSIONS

There is no single best latent dimensionality or compression algorithm for analyzing gene expression data. Instead, using features derived from different compression models across multiple latent space dimensionalities enhances biological representations.

Independent Component Analysis for Unraveling the Complexity of Cancer Omics Datasets

Independent component analysis (ICA) is a matrix factorization approach where the signals captured by each individual matrix factors are optimized to become as mutually independent as possible. Initially suggested for solving source blind separation problems in various fields, ICA was shown to be successful in analyzing functional magnetic resonance imaging (fMRI) and other types of biomedical data. In the last twenty years, ICA became a part of the standard machine learning toolbox, together with other matrix factorization methods such as principal component analysis (PCA) and non-negative matrix factorization (NMF). Here, we review a number of recent works where ICA was shown to be a useful tool for unraveling the complexity of cancer biology from the analysis of different types of omics data, mainly collected for tumoral samples. Such works highlight the use of ICA in dimensionality reduction, deconvolution, data pre-processing, meta-analysis, and others applied to different data types (transcriptome, methylome, proteome, single-cell data). We particularly focus on the technical aspects of ICA application in omics studies such as using different protocols, determining the optimal number of components, assessing and improving reproducibility of the ICA results, and comparison with other popular matrix factorization techniques. We discuss the emerging ICA applications to the integrative analysis of multi-level omics datasets and introduce a conceptual view on ICA as a tool for defining functional subsystems of a complex biological system and their interactions under various conditions. Our review is accompanied by a Jupyter notebook which illustrates the discussed concepts and provides a practical tool for applying ICA to the analysis of cancer omics datasets.

Discovering Pathway and Cell Type Signatures in Transcriptomic Compendia with Machine Learning

Pathway and cell type signatures are patterns present in transcriptome data that are associated with biological processes or phenotypic consequences. These signatures result from specific cell type and pathway expression but can require large transcriptomic compendia to detect. Machine learning techniques can be powerful tools for signature discovery through their ability to provide accurate and interpretable results. In this review, we discuss various machine learning applications to extract pathway and cell type signatures from transcriptomic compendia. We focus on the biological motivations and interpretation for both supervised and unsupervised learning approaches in this setting. We consider recent advances, including deep learning, and their applications to expanding bulk and single-cell RNA data. As data and computational resources increase, there will be more opportunities for machine learning to aid in revealing biological signatures.

Genetics, epigenetics and redox homeostasis in rhabdomyosarcoma: Emerging targets and therapeutics

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma accounting for 5-8% of malignant tumours in children and adolescents. Children with high risk disease have poor prognosis. Anti-RMS therapies include surgery, radiation and combination chemotherapy. While these strategies improved survival rates, they have plateaued since 1990s as drugs that target differentiation and self-renewal of tumours cells have not been identified. Moreover, prevailing treatments are aggressive with drug resistance and metastasis causing failure of several treatment regimes. Significant advances have been made recently in understanding the genetic and epigenetic landscape in RMS. These studies have identified novel diagnostic and prognostic markers and opened new avenues for treatment. An important target identified in high throughput drug screening studies is reactive oxygen species (ROS). Indeed, many drugs in clinical trials for RMS impact tumour progression through ROS. In light of such emerging evidence, we discuss recent findings highlighting key pathways, epigenetic alterations and their impacts on ROS that form the basis of developing novel molecularly targeted therapies in RMS. Such targeted therapies in combination with conventional therapy could reduce adverse side effects in young survivors and lead to a decline in long-term morbidity.