High-throughput Biological Data Research Articles

Using feature selection and Bayesian network identify cancer subtypes based on proteomic data

The Cancer Proteome Atlas (TCPA) project collects reverse-phase protein arrays (RPPA)-based proteome datasets from nearly 8000 samples across 32 cancer types. This study aims to investigate the pan-cancer proteome signature and identify cancer subtypes of glioma, kidney cancer, and lung cancer based on TCPA data. We first visualized the tumor clustering models using t-distributed stochastic neighbour embedding (t-SNE) and bi-clustering heatmap. Then, three feature selection methods (pyHSICLasso, XGBoost, and Random Forest) were performed to select protein features for classifying cancer subtypes in training dataset, and the LibSVM algorithm was empolyed to test classification accuracy in the validation dataset. Clustering analysis revealed that different kinds of tumors have relatively distinct proteomic profiling based on tissue or origin. We identified 20, 10, and 20 protein features with the highest accuracies in classifying subtypes of glioma, kidney cancer, and lung cancer, respectively. The predictive abilities of the selected proteins were confirmed by receiving operating characteristic (ROC) analysis. Finally, the Bayesian network was utilized to explore the protein biomarkers that have direct causal relationships with cancer subtypes. Overall, we highlight the theoretical and technical applications of machine learning based feature selection approaches in the analysis of high-throughput biological data, particularly for cancer biomarker research. SignificanceFunctional proteomics is a powerful approach for characterizing cell signaling pathways and understanding their phenotypic effects on cancer development. The TCPA database provides a platform to explore and analyze TCGA pan-cancer RPPA-based protein expression. With the advent of the RPPA technology, the availability of high-throughput data in TCPA platform has made it possible to use machine learning methods to identify protein biomarkers and further differentiate subtypes of cancer based on proteomic data. In this study, we highlight the role of feature selection and Bayesian network in discovery protein biomarker for classifying cancer subtypes based on functional proteomic data. The application of machine learning methods in the analysis of high-throughput biological data, particularly for cancer biomarker researches, which have potential clinical values in developing individualized treatment strategies.

Biological Phenotyping in Sepsis and Acute Respiratory Distress Syndrome.

Heterogeneity in sepsis and acute respiratory distress syndrome (ARDS) is increasingly being recognized as one of the principal barriers to finding efficacious targeted therapies. The advent of multiple high-throughput biological data ("omics"), coupled with the widespread access to increased computational power, has led to the emergence of phenotyping in critical care. Phenotyping aims to use a multitude of data to identify homogenous subgroups within an otherwise heterogenous population. Increasingly, phenotyping schemas are being applied to sepsis and ARDS to increase understanding of these clinical conditions and identify potential therapies. Here we present a selective review of the biological phenotyping schemas applied to sepsis and ARDS. Further, we outline some of the challenges involved in translating these conceptual findings to bedside clinical decision-making tools.

NetMix2: A Principled Network Propagation Algorithm for Identifying Altered Subnetworks.

A standard paradigm in computational biology is to leverage interaction networks as prior knowledge in analyzing high-throughput biological data, where the data give a score for each vertex in the network. One classical approach is the identification of altered subnetworks, or subnetworks of the interaction network that have both outlier vertex scores and a defined network topology. One class of algorithms for identifying altered subnetworks search for high-scoring subnetworks in subnetwork families with simple topological constraints, such as connected subnetworks, and have sound statistical guarantees. A second class of algorithms employ network propagation-the smoothing of vertex scores over the network using a random walk or diffusion process-and utilize the global structure of the network. However, network propagation algorithms often rely on ad hoc heuristics that lack a rigorous statistical foundation. In this work, we unify the subnetwork family and network propagation approaches by deriving the propagation family, a subnetwork family that approximates the sets of vertices ranked highly by network propagation approaches. We introduce NetMix2, a principled algorithm for identifying altered subnetworks from a wide range of subnetwork families. When using the propagation family, NetMix2 combines the advantages of the subnetwork family and network propagation approaches. NetMix2 outperforms other methods, including network propagation on simulated data, pan-cancer somatic mutation data, and genome-wide association data from multiple human diseases.

Identification of functional features underlying heat stress response in Sprague–Dawley rats using mixed linear models

Since global temperature is expected to rise by 2 °C in 2050 heat stress may become the most severe environmental factor. In the study, we illustrate the application of mixed linear models for the analysis of whole transcriptome expression in livers and adrenal tissues of Sprague–Dawley rats obtained by a heat stress experiment. By applying those models, we considered four sources of variation in transcript expression, comprising transcripts (1), genes (2), Gene Ontology terms (3), and Reactome pathways (4) and focussed on accounting for the similarity within each source, which was expressed as a covariance matrix. Models based on transcripts or genes levels explained a larger proportion of log2 fold change than models fitting the functional components of Gene Ontology terms or Reactome pathways. In the liver, among the most significant genes were PNKD and TRIP12. In the adrenal tissue, one transcript of the SUCO gene was expressed more strongly in the control group than in the heat-stress group. PLEC had two transcripts, which were significantly overexpressed in the heat-stress group. PER3 was significant only on gene level. Moving to the functional scale, five Gene Ontologies and one Reactome pathway were significant in the liver. They can be grouped into ontologies related to DNA repair, histone ubiquitination, the regulation of embryonic development and cytoplasmic translation. Linear mixed models are valuable tools for the analysis of high-throughput biological data. Their main advantages are the possibility to incorporate information on covariance between observations and circumventing the problem of multiple testing.

Two-stage linked component analysis for joint decomposition of multiple biologically related data sets.

Integrative analysis of multiple data sets has the potential of fully leveraging the vast amount of high throughput biological data being generated. In particular such analysis will be powerful in making inference from publicly available collections of genetic, transcriptomic and epigenetic data sets which are designed to study shared biological processes, but which vary in their target measurements, biological variation, unwanted noise, and batch variation. Thus, methods that enable the joint analysis of multiple data sets are needed to gain insights into shared biological processes that would otherwise be hidden by unwanted intra-data set variation. Here, we propose a method called two-stage linked component analysis (2s-LCA) to jointly decompose multiple biologically related experimental data sets with biological and technological relationships that can be structured into the decomposition. The consistency of the proposed method is established and its empirical performance is evaluated via simulation studies. We apply 2s-LCA to jointly analyze four data sets focused on human brain development and identify meaningful patterns of gene expression in human neurogenesis that have shared structure across these data sets.

Sparse matrix linear models for structured high-throughput data

Recent technological advancements have led to the rapid generation of high-throughput biological data which can be used to address novel scientific questions in broad areas of research. These data can be thought of as a large matrix with covariates annotating both its rows and columns. Matrix linear models provide a convenient way for modeling such data. In many situations, sparse estimation of these models is desired. We present fast, general methods for fitting sparse matrix linear models to structured high-throughput data. We induce model sparsity using an L1 penalty and consider the case when the response matrix and the covariate matrices are large. Due to data size, standard methods for estimation of these penalized regression models fail if the problem is converted to the corresponding univariate regression scenario. By leveraging matrix properties in the structure of our model, we develop several fast estimation algorithms (coordinate descent, FISTA and ADMM) and discuss their trade-offs. We evaluate our method’s performance on simulated data, E. coli chemical genetic screening data and two Arabidopsis genetic datasets with multivariate responses. Our algorithms have been implemented in the Julia programming language and are available at https://github.com/senresearch/MatrixLMnet.jl.

Comparison of Different Label-Free Techniques for the Semi-Absolute Quantification of Protein Abundance.

In proteomics, it is essential to quantify proteins in absolute terms if we wish to compare results among studies and integrate high-throughput biological data into genome-scale metabolic models. While labeling target peptides with stable isotopes allow protein abundance to be accurately quantified, the utility of this technique is constrained by the low number of quantifiable proteins that it yields. Recently, label-free shotgun proteomics has become the “gold standard” for carrying out global assessments of biological samples containing thousands of proteins. However, this tool must be further improved if we wish to accurately quantify absolute levels of proteins. Here, we used different label-free quantification techniques to estimate absolute protein abundance in the model yeast Saccharomyces cerevisiae. More specifically, we evaluated the performance of seven different quantification methods, based either on spectral counting (SC) or extracted-ion chromatogram (XIC), which were applied to samples from five different proteome backgrounds. We also compared the accuracy and reproducibility of two strategies for transforming relative abundance into absolute abundance: a UPS2-based strategy and the total protein approach (TPA). This study mentions technical challenges related to UPS2 use and proposes ways of addressing them, including utilizing a smaller, more highly optimized amount of UPS2. Overall, three SC-based methods (PAI, SAF, and NSAF) yielded the best results because they struck a good balance between experimental performance and protein quantification.

Bayesian multistudy factor analysis for high-throughput biological data

This paper analyzes breast cancer gene expression across seven studies to identify genuine and thus replicable gene patterns shared among these studies. Our premise is that genuine biological signal is more likely to be reproducibly present in multiple studies than spurious signal. Our analysis uses a new modeling strategy for the joint analysis of high-throughput biological studies which simultaneously identifies shared as well as study-specific signal. To this end, we generalize the multi-study factor analysis model to handle high-dimensional data and generalize the sparse Bayesian infinite factor model to this context. We provide strategies for the identification of the loading matrices, common and study-specific. Through extensive simulation analysis, we characterize the performance of the proposed approach in various scenarios and show that it outperforms standard factor analysis in identifying replicable signal in all scenarios considered. The analysis of breast cancer gene expression studies identifies clear replicable gene patterns. These patterns are related to well-known biological pathways involved in breast cancer, such as the ER, cell cycle, immune system, collagen, and metabolic pathways. Some of these patterns are also associated with existing breast cancer subtypes, such as LumA, Her2, and basal subtypes, while other patterns identify novel pathways active across subtypes and missed by hierarchical clustering approaches. The R package MSFA implementing the method is available on GitHub.

Clipper: p-value-free FDR control on high-throughput data from two conditions

High-throughput biological data analysis commonly involves identifying features such as genes, genomic regions, and proteins, whose values differ between two conditions, from numerous features measured simultaneously. The most widely used criterion to ensure the analysis reliability is the false discovery rate (FDR), which is primarily controlled based on p-values. However, obtaining valid p-values relies on either reasonable assumptions of data distribution or large numbers of replicates under both conditions. Clipper is a general statistical framework for FDR control without relying on p-values or specific data distributions. Clipper outperforms existing methods for a broad range of applications in high-throughput data analysis.

Multimodal regularized linear models with flux balance analysis for mechanistic integration of omics data.

High-throughput biological data, thanks to technological advances, have become cheaper to collect, leading to the availability of vast amounts of omic data of different types. In parallel, the in silico reconstruction and modeling of metabolic systems is now acknowledged as a key tool to complement experimental data on a large scale. The integration of these model- and data-driven information is therefore emerging as a new challenge in systems biology, with no clear guidance on how to better take advantage of the inherent multisource and multiomic nature of these data types while preserving mechanistic interpretation. Here, we investigate different regularization techniques for high-dimensional data derived from the integration of gene expression profiles with metabolic flux data, extracted from strain-specific metabolic models, to improve cellular growth rate predictions. To this end, we propose ad-hoc extensions of previous regularization frameworks including group, view-specific and principal component regularization and experimentally compare them using data from 1143 Saccharomyces cerevisiae strains. We observe a divergence between methods in terms of regression accuracy and integration effectiveness based on the type of regularization employed. In multiomic regression tasks, when learning from experimental and model-generated omic data, our results demonstrate the competitiveness and ease of interpretation of multimodal regularized linear models compared to data-hungry methods based on neural networks. All data, models and code produced in this work are available on GitHub at https://github.com/Angione-Lab/HybridGroupIPFLasso_pc2Lasso. Supplementary data are available at Bioinformatics online.

Handling the Cellular Complex Systems in Alzheimer's Disease Through a Graph Mining Approach.

In the last two decades, the medical sciences have changed their approach to pathogenesis as well as to the diagnosis and treatment of complex human diseases. The main reason for this change is the explosive development of biomedical technology and research, which produces a huge amount of information and data which are generated at an increasing rate. Toward this direction is the pathway analysis, a thriving research area of systems biology tools and methodologies which aim to unravel the inherent complexity of high-throughput biological data produced by the advent of omics technologies. Through this graph mining approach, we can deal with the complexity of the cellular systems of various diseases such as Alzheimer's disease. In this work, we developed a subpathway analysis method for single-cell RNA-seq experiments which isolates differentially expressed subpathways indicating potentially perturbed biological processes. The differential expression status of each gene is negotiated among well-established RNA-seq differential expression analysis tools in order to minimize false discoveries. Also, we demonstrate the efficacy of our method on a single-cell RNA-seq dataset for temporal tracking of microglia activation in neurodegeneration. Results suggest that our approach succeeds in isolating several perturbed biological processes known to be associated with neurodegeneration.

Markov chain Monte Carlo for active module identification problem

BackgroundIntegrative network methods are commonly used for interpretation of high-throughput experimental biological data: transcriptomics, proteomics, metabolomics and others. One of the common approaches is finding a connected subnetwork of a global interaction network that best encompasses significant individual changes in the data and represents a so-called active module. Usually methods implementing this approach find a single subnetwork and thus solve a hard classification problem for vertices. This subnetwork inherently contains erroneous vertices, while no instrument is provided to estimate the confidence level of any particular vertex inclusion. To address this issue, in the current study we consider the active module problem as a soft classification problem.ResultsWe propose a method to estimate probabilities of each vertex to belong to the active module based on Markov chain Monte Carlo (MCMC) subnetwork sampling. As an example of the performance of our method on real data, we run it on two gene expression datasets. For the first many-replicate expression dataset we show that the proposed approach is consistent with an existing resampling-based method. On the second dataset the jackknife resampling method is inapplicable due to the small number of biological replicates, but the MCMC method can be run and shows high classification performance.ConclusionsThe proposed method allows to estimate the probability that an individual vertex belongs to the active module as well as the false discovery rate (FDR) for a given set of vertices. Given the estimated probabilities, it becomes possible to provide a connected subgraph in a consistent manner for any given FDR level: no vertex can disappear when the FDR level is relaxed. We show, on both simulated and real datasets, that the proposed method has good computational performance and high classification accuracy.

Systems Analysis of Biliary Atresia Through Integration of High-Throughput Biological Data.

Biliary atresia (BA), blockage of the proper bile flow due to loss of extrahepatic bile ducts, is a rare, complex disease of the liver and the bile ducts with unknown etiology. Despite ongoing investigations to understand its complex pathogenesis, BA remains the most common cause of liver failure requiring liver transplantation in children. To elucidate underlying mechanisms, we analyzed the different types of high-throughput genomic and transcriptomic data collected from the blood and liver tissue samples of children suffering from BA. Through use of a novel integrative approach, we identified potential biomarkers and over-represented biological functions and pathways to derive a comprehensive network showing the dysfunctional mechanisms associated with BA. One of the pathways highlighted in the integrative network was hypoxia signaling. Perturbation with hypoxia inducible factor activator, dimethyloxalylglycine, induced the biliary defects of BA in a zebrafish model, serving as a validation for our studies. Our approach enables a systems-level understanding of human BA biology that is highlighted by the interaction between key biological functions such as fibrosis, inflammation, immunity, hypoxia, and development.

An Experiment on Ab Initio Discovery of Biological Knowledge from scRNA-Seq Data Using Machine Learning

SummaryExpectations of machine learning (ML) are high for discovering new patterns in high-throughput biological data, but most such practices are accustomed to relying on existing knowledge conditions to design experiments. Investigations of the power and limitation of ML in revealing complex patterns from data without the guide of existing knowledge have been lacking. In this study, we conducted systematic experiments on such ab initio knowledge discovery with ML methods on single-cell RNA-sequencing data of early embryonic development. Results showed that a strategy combining unsupervised and supervised ML can reveal major cell lineages with minimum involvement of prior knowledge or manual intervention, and the ab initio mining enabled a new discovery of human early embryonic cell differentiation. The study illustrated the feasibility, significance, and limitation of ab initio ML knowledge discovery on complex biological problems.

Modern deep learning in bioinformatics.

Deep learning (DL) has shown explosive growth in its application to bioinformatics and has demonstrated thrillingly promising power to mine the complex relationship hidden in large-scale biological and biomedical data. A number of comprehensive reviews have been published on such applications, ranging from high-level reviews with future perspectives to those mainly serving as tutorials. These reviews have provided an excellent introduction to and guideline for applications of DL in bioinformatics, covering multiple types of machine learning (ML) problems, different DL architectures, and ranges of biological/biomedical problems. However, most of these reviews have focused on previous research, whereas current trends in the principled DL field and perspectives on their future developments and potential new applications to biology and biomedicine are still scarce. We will focus on modern DL, the ongoing trends and future directions of the principled DL field, and postulate new and major applications in bioinformatics.

Gaussian Embedding for Large-scale Gene Set Analysis.

Gene sets, including protein complexes and signaling pathways, have proliferated greatly, in large part as a result of high-throughput biological data. Leveraging gene sets to gain insight into biological discovery requires computational methods for converting them into a useful form for available machine learning models. Here, we study the problem of embedding gene sets as compact features that are compatible with available machine learning codes. We present Set2Gaussian, a novel network-based gene set embedding approach, which represents each gene set as a multivariate Gaussian distribution rather than a single point in the low-dimensional space, according to the proximity of these genes in a protein-protein interaction network. We demonstrate that Set2Gaussian improves gene set member identification, accurately stratifies tumors, and finds concise gene sets for gene set enrichment analysis. We further show how Set2Gaussian allows us to identify a previously unknown clinical prognostic and predictive subnetwork around NEFM in sarcoma, which we validate in independent cohorts.

CAncer bioMarker Prediction Pipeline (CAMPP)-A standardized framework for the analysis of quantitative biological data.

With the improvement of -omics and next-generation sequencing (NGS) methodologies, along with the lowered cost of generating these types of data, the analysis of high-throughput biological data has become standard both for forming and testing biomedical hypotheses. Our knowledge of how to normalize datasets to remove latent undesirable variances has grown extensively, making for standardized data that are easily compared between studies. Here we present the CAncer bioMarker Prediction Pipeline (CAMPP), an open-source R-based wrapper (https://github.com/ELELAB/CAncer-bioMarker-Prediction-Pipeline -CAMPP) intended to aid bioinformatic software-users with data analyses. CAMPP is called from a terminal command line and is supported by a user-friendly manual. The pipeline may be run on a local computer and requires little or no knowledge of programming. To avoid issues relating to R-package updates, a renv .lock file is provided to ensure R-package stability. Data-management includes missing value imputation, data normalization, and distributional checks. CAMPP performs (I) k-means clustering, (II) differential expression/abundance analysis, (III) elastic-net regression, (IV) correlation and co-expression network analyses, (V) survival analysis, and (VI) protein-protein/miRNA-gene interaction networks. The pipeline returns tabular files and graphical representations of the results. We hope that CAMPP will assist in streamlining bioinformatic analysis of quantitative biological data, whilst ensuring an appropriate bio-statistical framework.

Variability in protein cargo detection in technical and biological replicates of exosome-enriched extracellular vesicles.

Exosomes are extracellular vesicles (EVs) of ~20–200 nm diameter that shuttle DNAs, RNAs, proteins and other biomolecules between cells. The large number of biomolecules present in exosomes demands the frequent use of high-throughput analysis. This, in turn, requires technical replicates (TRs), and biological replicates (BRs) to produce accurate results. As the number and abundance of identified biomolecules varies between replicates (Rs), establishing the replicate variability predicted for the event under study is essential in determining the number of Rs required. Although there have been few reports of replicate variability in high throughput biological data, none of them focused on exosomes. Herein, we determined the replicate variability in protein profiles found in exosomes released from 3 lung adenocarcinoma cell lines, H1993, A549 and H1975. Since exosome isolates are invariably contaminated by a small percentage of ~200–300 nm microvesicles, we refer to our samples as exosome-enriched EVs (EE-EVs). We generated BRs of EE-EVs from each cell line, and divided each group into 3 TRs. All Rs were analyzed by liquid chromatography/mass spectrometry (LC/MS/MS) and customized bioinformatics and biostatistical workflows (raw data available via ProteomeXchange: PXD012798). We found that the variability among TRs as well as BRs, was largely qualitative (protein present or absent) and higher among BRs. By contrast, the quantitative (protein abundance) variability was low, save for the H1975 cell line where the quantitative variability was significant. Importantly, our replicate strategy identified 90% of the most abundant proteins, thereby establishing the utility of our approach.

A probabilistic graphical model for system-wide analysis of gene regulatory networks.

The inference of gene regulatory networks (GRNs) from DNA microarray measurements forms a core element of systems biology-based phenotyping. In the recent past, numerous computational methodologies have been formalized to enable the deduction of reliable and testable predictions in today's biology. However, little focus has been aimed at quantifying how well existing state-of-the-art GRNs correspond to measured gene-expression profiles. Here, we present a computational framework that combines the formulation of probabilistic graphical modeling, standard statistical estimation, and integration of high-throughput biological data to explore the global behavior of biological systems and the global consistency between experimentally verified GRNs and corresponding large microarray compendium data. The model is represented as a probabilistic bipartite graph, which can handle highly complex network systems and accommodates partial measurements of diverse biological entities, e.g. messengerRNAs, proteins, metabolites and various stimulators participating in regulatory networks. This method was tested on microarray expression data from the M3D database, corresponding to sub-networks on one of the best researched model organisms, Escherichia coli. Results show a surprisingly high correlation between the observed states and the inferred system's behavior under various experimental conditions. Processed data and software implementation using Matlab are freely available at https://github.com/kotiang54/PgmGRNs. Full dataset available from the M3D database.

ORSO (Online Resource for Social Omics): A data-driven social network connecting scientists to genomics datasets.

High-throughput sequencing has become ubiquitous in biomedical sciences. As new technologies emerge and sequencing costs decline, the diversity and volume of available data increases exponentially, and successfully navigating the data becomes more challenging. Though datasets are often hosted by public repositories, scientists must rely on inconsistent annotation to identify and interpret meaningful data. Moreover, the experimental heterogeneity and wide-ranging quality of high-throughput biological data means that even data with desired cell lines, tissue types, or molecular targets may not be readily interpretable or integrated. We have developed ORSO (Online Resource for Social Omics) as an easy-to-use web application to connect life scientists with genomics data. In ORSO, users interact within a data-driven social network, where they can favorite datasets and follow other users. In addition to more than 30,000 datasets hosted from major biomedical consortia, users may contribute their own data to ORSO, facilitating its discovery by other users. Leveraging user interactions, ORSO provides a novel recommendation system to automatically connect users with hosted data. In addition to social interactions, the recommendation system considers primary read coverage information and annotated metadata. Similarities used by the recommendation system are presented by ORSO in a graph display, allowing exploration of dataset associations. The topology of the network graph reflects established biology, with samples from related systems grouped together. We tested the recommendation system using an RNA-seq time course dataset from differentiation of embryonic stem cells to cardiomyocytes. The ORSO recommendation system correctly predicted early data point sources as embryonic stem cells and late data point sources as heart and muscle samples, resulting in recommendation of related datasets. By connecting scientists with relevant data, ORSO provides a critical new service that facilitates wide-ranging research interests.