FuSe: a tool to move RNA-Seq analyses from chromosomal/gene loci to functional grouping of mRNA transcripts.

SUMMARY: Typical RNA sequencing (RNA-Seq) analyses are performed either at the gene level by summing all reads from the same locus, assuming that all transcripts from a gene make a protein or at the transcript level, assuming that each transcript displays unique function. However, these assumptions are flawed, as a gene can code for different types of transcripts and different transcripts are capable of synthesizing similar, different or no protein. As a consequence, functional changes are not well illustrated by either gene or transcript analyses. We propose to improve RNA-Seq analyses by grouping the transcripts based on their similar functions. We developed FuSe to predict functional similarities using the primary and secondary structure of proteins. To estimate the likelihood of proteins with similar functions, FuSe computes two confidence scores: knowledge (KS) and discovery (DS) for protein pairs. Overlapping protein pairs exhibiting high confidence are grouped to form 'similar function protein groups' and expression is calculated for each functional group. The impact of using FuSe is demonstrated on in vitro cells exposed to paracetamol, which highlight genes responsible for cell adhesion and glycogen regulation which were earlier shown to be not differentially expressed with traditional analysis methods. AVAILABILITY AND IMPLEMENTATION: The source code is available at https://github.com/rajinder4489/FuSe. Data for APAP exposure are available in the BioStudies database (http://www.ebi.ac.uk/biostudies) under accession numbers S-HECA143, S-HECA(158) and S-HECA139. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

R-ODAF: Omics data analysis framework for regulatory application.

Despite the widespread use of transcriptomics technologies in toxicology research, acceptance of the data by regulatory agencies to support the hazard assessment is still limited. Fundamental issues contributing to this are the lack of reproducibility in transcriptomics data analysis arising from variance in the methods used to generate data and differences in the data processing. While research applications are flexible in the way the data are generated and interpreted, this is not the case for regulatory applications where an unambiguous answer, possibly later subject to legal scrutiny, is required. A reference analysis framework would give greater credibility to the data and allow the practitioners to justify their use of an alternative bioinformatic process by referring to a standard. In this publication, we propose a method called omics data analysis framework for regulatory application (R-ODAF), which has been built as a user-friendly pipeline to analyze raw transcriptomics data from microarray and next-generation sequencing. In the R-ODAF, we also propose additional statistical steps to remove the number of false positives obtained from standard data analysis pipelines for RNA-sequencing. We illustrate the added value of R-ODAF, compared to a standard workflow, using a typical toxicogenomics dataset of hepatocytes exposed to paracetamol.

Towards the development of an omics data analysis framework.

The use of various omics techniques for scientific research is increasing. While toxicogenomics studies have already produced substantial data on diverse omics platforms, to date there has been little routine application in regulatory toxicology. This is despite the promises and excitement of 20 years ago when it was widely speculated that omics methods would reduce or even replace animal use and allow a much enhanced understanding of hazard and susceptibility. One of the reasons for this has been a trepidation about relying on the produced data. It has been argued that omics outputs might not be sufficiently reliable for regulatory application because the techniques, bioinformatics and interpretation can vary. For these reasons the robustness of the obtained results is questioned. This reticence to trust omics data is further magnified by the lack of internationally agreed upon guidelines and protocols for both the generation and processing of omics data. One way forward would be to reach a consensus on an omics data analysis framework (ODAF) for regulatory application (R-ODAF) based on rigorous data analysis. The authors of this article are involved in a Long-Range Research Initiative (LRI) project that will propose an R-ODAF for transcriptomics data. The R-ODAF will then be reviewed and evaluated by the main regulatory agencies and consensus forums such as the Organization for Economic Co-operation and Development (OECD). This work builds on The MicroArray Quality Control work that developed standards for the generation of data from microarrays and sequencing but not for reporting or analysis.

Predicting missing proteomics values using machine learning: Filling the gap using transcriptomics and other biological features.

Proteins are often considered the main biological element in charge of the different functions and structures of a cell. However, proteomics, the global study of all expressed proteins, often performed by mass spectrometry, is limited by its stochastic sampling and can only quantify a limited amount of protein per sample. Transcriptomics, which allows an exhaustive analysis of all expressed transcripts, is often used as a surrogate. However, the transcript level does not present a high level of correlation with the corresponding protein level, notably due to the existence of several post-transcriptional regulatory mechanisms. In this publication, we hypothesize that the missing protein values in proteomics could be predicted using machine learning regression methods, trained with many features extracted from transcriptomics, including known translational regulatory elements such as microRNAs and circular RNAs. After considering different machine learning algorithms applied on two different splitting strategies, we report that random forest can predict proteins in new samples out of transcriptomics data with good accuracy. The proposed pre-processing and model building scripts can be accessed on GitHub: https://github.com/jochotecoa/ml_proteomics.

Multi-omics HeCaToS dataset of repeated dose toxicity for cardiotoxic & hepatotoxic compounds.

The data currently described was generated within the EU/FP7 HeCaToS project (Hepatic and Cardiac Toxicity Systems modeling). The project aimed to develop an in silico prediction system to contribute to drug safety assessment for humans. For this purpose, multi-omics data of repeated dose toxicity were obtained for 10 hepatotoxic and 10 cardiotoxic compounds. Most data were gained from in vitro experiments in which 3D microtissues (either hepatic or cardiac) were exposed to a therapeutic (physiologically relevant concentrations calculated through PBPK-modeling) or a toxic dosing profile (IC20 after 7 days). Exposures lasted for 14 days and samples were obtained at 7 time points (therapeutic doses: 2-8-24-72-168-240-336 h; toxic doses 0-2-8-24-72-168-240 h). Transcriptomics (RNA sequencing & microRNA sequencing), proteomics (LC-MS), epigenomics (MeDIP sequencing) and metabolomics (LC-MS & NMR) data were obtained from these samples. Furthermore, functional endpoints (ATP content, Caspase3/7 and O2 consumption) were measured in exposed microtissues. Additionally, multi-omics data from human biopsies from patients are available. This data is now being released to the scientific community through the BioStudies data repository ( https://www.ebi.ac.uk/biostudies/ ).

Lead-exposure associated miRNAs in humans and Alzheimer's disease as potential biomarkers of the disease and disease processes.

Alzheimer's disease (AD) is a neurodegenerative disease that eventually affects memory and behavior. The identification of biomarkers based on risk factors for AD provides insight into the disease since the exact cause of AD remains unknown. Several studies have proposed microRNAs (miRNAs) in blood as potential biomarkers for AD. Exposure to heavy metals is a potential risk factor for onset and development of AD. Blood cells of subjects that are exposed to lead detected in the circulatory system, potentially reflect molecular responses to this exposure that are similar to the response of neurons. In this study we analyzed blood cell-derived miRNAs derived from a general population as proxies of potentially AD-related mechanisms triggered by lead exposure. Subsequently, we analyzed these mechanisms in the brain tissue of AD subjects and controls. A total of four miRNAs were identified as lead exposure-associated with hsa-miR-3651, hsa-miR-150-5p and hsa-miR-664b-3p being negatively and hsa-miR-627 positively associated. In human brain derived from AD and AD control subjects all four miRNAs were detected. Moreover, two miRNAs (miR-3651, miR-664b-3p) showed significant differential expression in AD brains versus controls, in accordance with the change direction of lead exposure. The miRNAs' gene targets were validated for expression in the human brain and were found enriched in AD-relevant pathways such as axon guidance. Moreover, we identified several AD relevant transcription factors such as CREB1 associated with the identified miRNAs. These findings suggest that the identified miRNAs are involved in the development of AD and might be useful in the development of new, less invasive biomarkers for monitoring of novel therapies or of processes involved in AD development.

Translational Proteomics Analysis of Anthracycline-Induced Cardiotoxicity From Cardiac Microtissues to Human Heart Biopsies.

Anthracyclines, including doxorubicin, idarubicin, and epirubicin, are common antitumor drugs as well as well-known cardiotoxic agents. This study analyzed the proteomics alteration in cardiac tissues caused by these 3 anthracyclines analogs. The in vitro human cardiac microtissues were exposed to drugs in 2 weeks; the proteomic data were measured at 7 time points. The heart biopsy data were collected from heart failure patients, in which some patients underwent anthracycline treatment. The anthracyclines-affected proteins were separately identified in the in vitro and in vivo dataset using the WGCNA method. These proteins engage in different cellular pathways including translation, metabolism, mitochondrial function, muscle contraction, and signaling pathways. From proteins detected in 2 datasets, a protein-protein network was established with 4 hub proteins, and 7 weighted proteins from both cardiac microtissue and human biopsies data. These 11 proteins, which involve in mitochondrial functions and the NF-κB signaling pathway, could provide insights into the anthracycline toxic mechanism. Some of them, such as HSPA5, BAG3, and SH3BGRL, are cardiac therapy targets or cardiotoxicity biomarkers. Other proteins, such as ATP5F1B and EEF1D, showed similar responses in both the in vitro and in vivo data. This suggests that the in vitro outcomes could link to clinical phenomena in proteomic analysis.

Network integration and modelling of dynamic drug responses at multi-omics levels.

Uncovering cellular responses from heterogeneous genomic data is crucial for molecular medicine in particular for drug safety. This can be realized by integrating the molecular activities in networks of interacting proteins. As proof-of-concept we challenge network modeling with time-resolved proteome, transcriptome and methylome measurements in iPSC-derived human 3D cardiac microtissues to elucidate adverse mechanisms of anthracycline cardiotoxicity measured with four different drugs (doxorubicin, epirubicin, idarubicin and daunorubicin). Dynamic molecular analysis at in vivo drug exposure levels reveal a network of 175 disease-associated proteins and identify common modules of anthracycline cardiotoxicity in vitro, related to mitochondrial and sarcomere function as well as remodeling of extracellular matrix. These in vitro-identified modules are transferable and are evaluated with biopsies of cardiomyopathy patients. This to our knowledge most comprehensive study on anthracycline cardiotoxicity demonstrates a reproducible workflow for molecular medicine and serves as a template for detecting adverse drug responses from complex omics data.

Bringing in vitro analysis closer to in vivo: Studying doxorubicin toxicity and associated mechanisms in 3D human microtissues with PBPK-based dose modelling.

Doxorubicin (DOX) is a chemotherapeutic agent of which the medical use is limited due to cardiotoxicity. While acute cardiotoxicity is reversible, chronic cardiotoxicity is persistent or progressive, dose-dependent and irreversible. While DOX mechanisms of action are not fully understood yet, 3 toxicity processes are known to occur in vivo: cardiomyocyte dysfunction, mitochondrial dysfunction and cell death. We present an in vitro experimental design aimed at detecting DOX-induced cardiotoxicity by obtaining a global view of the induced molecular mechanisms through RNA-sequencing. To better reflect the in vivo situation, human 3D cardiac microtissues were exposed to physiologically-based pharmacokinetic (PBPK) relevant doses of DOX for 2 weeks. We analysed a therapeutic and a toxic dosing profile. Transcriptomics analysis revealed significant gene expression changes in pathways related to "striated muscle contraction" and "respiratory electron transport", thus suggesting mitochondrial dysfunction as an underlying mechanism for cardiotoxicity. Furthermore, expression changes in mitochondrial processes differed significantly between the doses. Therapeutic dose reflects processes resembling the phenotype of delayed chronic cardiotoxicity, while toxic doses resembled acute cardiotoxicity. Overall, these results demonstrate the capability of our innovative in vitro approach to detect the three known mechanisms of DOX leading to toxicity, thus suggesting its potential relevance for reflecting the patient situation. Our study also demonstrated the importance of applying physiologically relevant doses during toxicological research, since mechanisms of acute and chronic toxicity differ.

Development and regulatory application of microRNA biomarkers.

MicroRNAs, a class of regulatory small non-coding RNAs, are emerging as promising biomarkers for different health outcomes. Due to their tissue specificity, stability in extracellular space and high conservation between preclinical test species, applications of novel miRNA-based biomarkers for drug safety testing regarding hepatotoxicity and cardiotoxicity are investigated. Furthermore, miRNA expression is altered by environmental exposure such as cigarette smoke or polychlorinated biphenyls. As a consequence, miRNAs potentially influence tumor suppressor genes and oncogenes and may influence carcinogenesis. This has raised the interest in the use of miRNA profiles for health risk assessment. This review summarizes the recent developments in miRNA research with focus on biomarkers for drug safety testing and biomarkers for health outcomes related to environmental exposures.