Improved prediction of smoking status via isoform-aware RNA-seq deep learning models

Most predictive models based on gene expression data do not leverage information related to gene splicing, despite the fact that splicing is a fundamental feature of eukaryotic gene expression. Cigarette smoking is an important environmental risk factor for many diseases, and it has profound effects on gene expression. Using smoking status as a prediction target, we developed deep neural network predictive models using gene, exon, and isoform level quantifications from RNA sequencing data in 2,557 subjects in the COPDGene Study. We observed that models using exon and isoform quantifications clearly outperformed gene-level models when using data from 5 genes from a previously published prediction model. Whereas the test set performance of the previously published model was 0.82 in the original publication, our exon-based models including an exon-to-isoform mapping layer achieved a test set AUC (area under the receiver operating characteristic) of 0.88, which improved to an AUC of 0.94 using exon quantifications from a larger set of genes. Isoform variability is an important source of latent information in RNA-seq data that can be used to improve clinical prediction models.

Download Full-text

An Objective Search for Unrecognized Bias in Validated COVID-19 Prediction Models

10.1101/2021.10.28.21265629 ◽

2021 ◽

Author(s):

Hossein Estiri ◽

Zachary Strasser ◽

Sina Rashidian ◽

Jeffrey Klann ◽

Kavishwar Wagholikar ◽

...

Keyword(s):

Predictive Models ◽

Medical Records ◽

Prediction Models ◽

Evaluation Studies ◽

Protected Group ◽

Increased Risk ◽

Using Data ◽

Bias Evaluation ◽

Algorithmic Bias

The growing recognition of algorithmic bias has spurred discussions about fairness in artificial intelligence (AI) / machine learning (ML) algorithms. The increasing translation of predictive models into clinical practice brings an increased risk of direct harm from algorithmic bias; however, bias remains incompletely measured in many medical AI applications. Using data from over 56 thousand Mass General Brigham (MGB) patients with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), we evaluate unrecognized bias in four AI models developed during the early months of the pandemic in Boston, Massachusetts that predict risks of hospital admission, ICU admission, mechanical ventilation, and death after a SARS-CoV-2 infection purely based on their pre-infection longitudinal medical records. We discuss that while a model can be biased against certain protected groups (i.e., perform worse) in certain tasks, it can be at the same time biased towards another protected group (i.e., perform better). As such, current bias evaluation studies may lack a full depiction of the variable effects of a model on its subpopulations. If the goal is to make a change in a positive way, the underlying roots of bias need to be fully explored in medical AI. Only a holistic evaluation, a diligent search for unrecognized bias, can provide enough information for an unbiased judgment of AI bias that can invigorate follow-up investigations on identifying the underlying roots of bias and ultimately make a change.

Download Full-text

Side-by-side analysis of alternative approaches on multi-level RNA-seq data

10.1101/131862 ◽

2017 ◽

Author(s):

Irina Mohorianu

Keyword(s):

Differential Expression ◽

Rna Seq ◽

Sequencing Data ◽

Essential Components ◽

Sequencing Quality ◽

Quality Checks ◽

Multi Level ◽

Using Data ◽

Key Steps ◽

Robust Prediction

AbstractBackgroundRNA sequencing (RNA-seq) is widely used for RNA quantification across environmental, biological and medical sciences; it enables the description of genome-wide patterns of expression and the deduction of regulatory interactions and networks. The aim of computational analyses is to achieve an accurate output, i.e. rigorous quantification of genes/transcripts to allow a reliable prediction of differential expression (DE), despite the variable levels of noise and biases present in sequencing data. The evaluation of sequencing quality and normalization are essential components of this process.ResultsWe investigate the discriminative power of existing approaches for the quality checking of mRNA-seq data and also propose additional, quantitative, quality checks. To accommodate the analysis of a nested, multi-level design using data on D. melanogaster, we incorporated the sample layout into the analysis. We describe a “subsampling without replacement”-based normalization and identification of DE that accounts for the experimental design i.e. the hierarchy and amplitude of effect sizes within samples. We also evaluate the differential expression call in comparison to existing approaches. To assess the broader applicability of these methods, we applied this series of steps to a published set of H. sapiens mRNA-seq samples.ConclusionsThe dataset-tailored methods improved sample comparability and delivered a robust prediction of subtle gene expression changes. Overall, the proposed approach offers the potential to improve key steps in the analysis of RNA-seq data by incorporating the structure and characteristics of biological experiments into the data analysis. 38

Download Full-text

Ak-mer-based method for the identification of phenotype-associated genomic biomarkers and predicting phenotypes of sequenced bacteria

10.1101/302026 ◽

2018 ◽

Author(s):

Erki Aun ◽

Age Brauer ◽

Veljo Kisand ◽

Tanel Tenson ◽

Maido Remm

Keyword(s):

Statistical Model ◽

Programming Language ◽

Prediction Models ◽

Genomic Sequence ◽

Medical Diagnostics ◽

Sequencing Data ◽

Test Set ◽

Phenotype Prediction ◽

Python Programming Language ◽

Python Programming

AbstractWe have developed an easy-to-use and memory-efficient method called PhenotypeSeeker that (a) generates ak-mer-based statistical model for predicting a given phenotype and (b) predicts the phenotype from the sequencing data of a given bacterial isolate. The method was validated on 167Klebsiella pneumoniaeisolates (virulence), 200Pseudomonas aeruginosaisolates (ciprofloxacin resistance) and 460Clostridium difficileisolates (azithromycin resistance). The phenotype prediction models trained from these datasets performed with 88% accuracy on theK. pneumoniaetest set, 88% on theP. aeruginosatest set and 96.5% on theC. difficiletest set. Prediction accuracy was the same for assembled sequences and raw sequencing data; however, building the model from assembled genomes is significantly faster. On these datasets, the model building on a mid-range Linux server takes approximately 3 to 5 hours per phenotype if assembled genomes are used and 10 hours per phenotype if raw sequencing data are used. The phenotype prediction from assembled genomes takes less than one second per isolate. Thus, PhenotypeSeeker should be well-suited for predicting phenotypes from large sequencing datasets.PhenotypeSeeker is implemented in Python programming language, is open-source software and is available at GitHub (https://github.com/bioinfo-ut/PhenotypeSeeker/).SummaryPredicting phenotypic properties of bacterial isolates from their genomic sequences has numerous potential applications. A good example would be prediction of antimicrobial resistance and virulence phenotypes for use in medical diagnostics. We have developed a method that is able to predict phenotypes of interest from the genomic sequence of the isolate within seconds. The method uses statistical model that can be trained automatically on isolates with known phenotype. The method is implemented in Python programming language and can be run on low-end Linux server and/or on laptop computers.

Download Full-text

Impact of therapy on gene expression in high-risk prostate cancer (PCA) treated with neoadjuvant docetaxel and androgen deprivation therapy.

Journal of Clinical Oncology ◽

10.1200/jco.2016.34.2_suppl.8 ◽

2016 ◽

Vol 34 (2_suppl) ◽

pp. 8-8

Author(s):

Himisha Beltran ◽

Alexander Wyatt ◽

Edmund Chedgy ◽

Ladan Fazli ◽

Andrea Sboner ◽

...

Keyword(s):

Gene Expression ◽

High Risk ◽

Hormone Receptors ◽

Dna Analysis ◽

Rna Extraction ◽

Post Treatment ◽

Rna Seq ◽

Sequencing Data ◽

Quantitative Expression ◽

High Risk Prostate Cancer

8 Background: Molecular analyses of neoadjuvant post-treatment radical prostatectomy (RP) specimens has been challenging as often times only microscopic foci remain present at time of RP precluding RNA-seq. DNA analysis alone in the absence of expression may be suboptimal in elucidating complex mechanisms of resistance and/or prognostic risk stratification. We therefore set out to develop an assay that could quantify mRNA expression in treated and untreated PCA using formalin fixed paraffin embedded (FFPE) tissues. Methods: We evaluated 40 untreated and post-treatment FFPE specimens as well as patient-matched pre-treated needle biopsies and baseline clinical data from patients enrolled on CALGB 90203: a randomized phase 3 trial comparing noeadjuvant docetaxel and ADT followed by RP vs RP alone for men with high risk localized PCA. High-density tumor areas were selected for RNA extraction (min 50ng RNA). We used NanoString nCounter to quantify gene expression of a custom panel of 75 genes including AR and androgen regulated, neural/neuroendocrine (NE), EMT, cell cycle, hormone receptors, TMPRSS-ERG, ARv7 splice variant, and housekeeper genes. mRNA data was integrated with matched whole exome sequencing data. Frozen specimens and RNA-Seq (n = 7) were used for QC and comparative analysis. Results: Quantitative expression using Nanostring showed high correlation with RNA-seq of patient-matched frozen tissue (Spearman coefficient 0.9). There was significant upregulation of AR and the ARv7 expression following treatment, as well as a subset of NE and EMT genes; three high chromogranin A outlier cases were identified in the treatment arm. There was an overall higher AR score in treated cases (based on expression of 30 AR signaling genes) compared to untreated, along the spectrum of CRPC. Conclusions: These data support the feasibility of quantifying gene expression in neoadjuvant-treated PCA cases with limited FFPE tissue requirement. Extensive characterization of AR status and NE/EMT genes identifies molecular outliers that can arise post-treatment and provides new insight into the heterogeneity of treatment response and potential early markers of resistance. Clinical trial information: NCT00430183.

Download Full-text

RNA sequencing analysis for profiling activation of cancer-associated molecular pathways.

Journal of Clinical Oncology ◽

10.1200/jco.2019.37.15_suppl.e13032 ◽

2019 ◽

Vol 37 (15_suppl) ◽

pp. e13032-e13032 ◽

Cited By ~ 2

Author(s):

Anton Buzdin ◽

Andrew Garazha ◽

Maxim Sorokin ◽

Alex Glusker ◽

Alexey Aleshin ◽

...

Keyword(s):

Gene Expression ◽

Original Data ◽

Tissue Expression ◽

Molecular Pathways ◽

Sequencing Analysis ◽

Rna Seq ◽

Sequencing Data ◽

Healthy Human ◽

Tissue Samples ◽

Normal Tissues

e13032 Background: Intracellular molecular pathways (IMPs) control all major events in the living cell. They are considered hotspots in contemporary oncology because knowledge of IMPs activation is essential for understanding mechanisms of molecular pathogenesis in oncology. Profiling IMPs requires RNA-seq data for tumors and for a collection of reference normal tissues. However, there is a shortage now in such profiles for normal tissues from healthy human donors, uniformly profiled in a single series of experiments. Access to the largest dataset of normal profiles GTEx is only partly available through the dbGaP. In TCGA database, norms are adjacent to surgically removed tumors and may be affected by tumor-linked growth factors, inflammation and altered vascularization. ENCODE datasets were for the autopsies of normal tissues, but they can’t form statistically significant reference groups. Methods: Tissue samples representing 20 organs were taken from post-mortal human healthy donors killed in road accidents no later than 36 hours after death, blood samples were taken from healthy volunteers. Gene expression was profiled in RNA-seq experiments using the same reagents, equipment and protocols. Bioinformatic algorithms for IMP analysis were developed and validated using experimental and public gene expression datasets. Results: From original sequencing data we constructed the biggest fully open reference expression database of normal human tissues including 465 profiles termed Oncobox Atlas of Normal Tissue Expression (ANTE, original data: GSE120795). We next developed a method termed Oncobox for interrogating activation of IMPs in human cancers. It includes modules of expression data harmonization and comparison and an algorithm for automatic annotation of molecular pathways. The Oncobox system enables accurate scoring of thousands molecular pathways using RNA-seq data. Oncobox pathway analysis is also applicable for quantitative proteomics and microRNA data in oncology. Conclusions: The Oncobox system can be used for a plethora of applications in cancer research including finding differentially regulated genes and IMPs, and for discovery of new pathway-related diagnostic and prognostic biomarkers.

Download Full-text

NASA GeneLab RNA-Seq Consensus Pipeline: Standardized Processing of Short-Read RNA-Seq Data

10.1101/2020.11.06.371724 ◽

2020 ◽

Author(s):

Eliah G. Overbey ◽

Amanda M. Saravia-Butler ◽

Zhe Zhang ◽

Komal S. Rathi ◽

Homer Fogle ◽

...

Keyword(s):

Expression Profiles ◽

Gene Expression Profiles ◽

Rna Seq ◽

Sequencing Data ◽

Analysis Pipeline ◽

Short Read ◽

Gene Quantification ◽

Working Groups ◽

Using Data ◽

Data Analysis Pipeline

SummaryWith the development of transcriptomic technologies, we are able to quantify precise changes in gene expression profiles from astronauts and other organisms exposed to spaceflight. Members of NASA GeneLab and GeneLab-associated analysis working groups (AWGs) have developed a consensus pipeline for analyzing short-read RNA-sequencing data from spaceflight-associated experiments. The pipeline includes quality control, read trimming, mapping, and gene quantification steps, culminating in the detection of differentially expressed genes. This data analysis pipeline and the results of its execution using data submitted to GeneLab are now all publicly available through the GeneLab database. We present here the full details and rationale for the construction of this pipeline in order to promote transparency, reproducibility and reusability of pipeline data, to provide a template for data processing of future spaceflight-relevant datasets, and to encourage cross-analysis of data from other databases with the data available in GeneLab.

Download Full-text

Impact of Gene Annotation Choice on the Quantification of RNA-Seq Data

10.21203/rs.3.rs-421080/v1 ◽

2021 ◽

Author(s):

David Chisanga ◽

Yang Liao ◽

Wei Shi

Keyword(s):

Gene Expression ◽

Gene Annotation ◽

Expression Data ◽

Refseq Gene ◽

Rna Seq ◽

Sequencing Data ◽

Microarray Expression Data ◽

Sequencing Quality ◽

Gene Expression Quantification ◽

Expression Quantification

Abstract Background: RNA sequencing is currently the method of choice for genome-wide profiling of gene expression. A popular approach to quantify expression levels of genes from RNA-seq data is to map reads to a reference genome and then count mapped reads to each gene. Gene annotation data, which include chromosomal coordinates of exons for tens of thousands of genes, are required for this quantification process. There are several major sources of gene annotations that can be used for quantification, such as Ensembl and RefSeq databases. However, there is very little understanding of the effect that the choice of annotation has on the accuracy of gene expression quantification in an RNA-seq analysis.Results: In this paper, we present results from our comparison of Ensembl and RefSeq human annotations on their impact on gene expression quantification using a benchmark RNA-seq dataset generated by the SEquencing Quality Control (SEQC) consortium. We show that the use of RefSeq gene annotation models led to better quantification accuracy, based on the correlation with ground truths including expression data from >800 real-time PCR validated genes, known titration ratios of gene expression and microarray expression data. We also found that the recent expansion of the RefSeq annotation has led to a decrease in its annotation accuracy. Finally, we demonstrated that the RNA-seq quantification differences observed between different annotations were not affected by the use of different normalization methods.Conclusion: In conclusion, our study found that the use of the conservative RefSeq gene annotation yields better RNA-seq quantification results than the more comprehensive Ensembl annotation. We also found that, surprisingly, the recent expansion of the RefSeq database, which was primarily driven by the incorporation of sequencing data into the gene annotation process, resulted in a reduction in the accuracy of RNA-seq quantification.

Download Full-text

SPARTA: Simple Program for Automated reference-based bacterial RNA-seq Transcriptome Analysis

10.1101/021915 ◽

2015 ◽

Author(s):

Benjamin K Johnson ◽

Matthew B Scholz ◽

Tracy K Teal ◽

Robert B Abramovitch

Keyword(s):

Gene Expression ◽

Differential Gene Expression ◽

Quality Analysis ◽

Rna Seq ◽

Sequencing Data ◽

Data Set ◽

Bacterial Rna ◽

Analysis Workflow ◽

Differential Gene ◽

Reference Counting

Summary: SPARTA is a reference-based bacterial RNA-seq analysis workflow application for single-end Illumina reads. SPARTA is turnkey software that simplifies the process of analyzing RNA-seq data sets, making bacterial RNA-seq analysis a routine process that can be undertaken on a personal computer or in the classroom. The easy-to-install, complete workflow processes whole transcriptome shotgun sequencing data files by trimming reads and removing adapters, mapping reads to a reference, counting gene features, calculating differential gene expression, and, importantly, checking for potential batch effects within the data set. SPARTA outputs quality analysis reports, gene feature counts and differential gene expression tables and scatterplots. The workflow is implemented in Python for file management and sequential execution of each analysis step and is available for Mac OS X, Microsoft Windows, and Linux. To promote the use of SPARTA as a teaching platform, a web-based tutorial is available explaining how RNA-seq data are processed and analyzed by the software. Availability and Implementation: Tutorial and workflow can be found at sparta.readthedocs.org. Teaching materials are located at sparta-teaching.readthedocs.org. Source code can be downloaded at www.github.com/abramovitchMSU/, implemented in Python and supported on Mac OS X, Linux, and MS Windows. Contact: Robert B. Abramovitch ([email protected]) Supplemental Information: Supplementary data are available online

Download Full-text

RNA sequencing data: hitchhiker's guide to expression analysis

10.7287/peerj.preprints.27283 ◽

2018 ◽

Author(s):

Koen Van Den Berge ◽

Katharina Hembach ◽

Charlotte Soneson ◽

Simone Tiberi ◽

Lieven Clement ◽

...

Keyword(s):

Gene Expression ◽

Rna Sequencing ◽

Large Scale ◽

Science Studies ◽

Rna Seq ◽

Sequencing Data ◽

Data Types ◽

The Past ◽

Long Read ◽

Statistical Approaches

Gene expression is the fundamental level at which the result of various genetic and regulatory programs are observable. The measurement of transcriptome-wide gene expression has convincingly switched from microarrays to sequencing in a matter of years. RNA sequencing (RNA-seq) provides a quantitative and open system for profiling transcriptional outcomes on a large scale and therefore facilitates a large diversity of applications, including basic science studies, but also agricultural or clinical situations. In the past 10 years or so, much has been learned about the characteristics of the RNA-seq datasets as well as the performance of the myriad of methods developed. In this review, we give an overall view of the developments in RNA-seq data analysis, including experimental design, with an explicit focus on quantification of gene expression and statistical approaches for differential expression. We also highlight emerging data types, such as single-cell RNA-seq and gene expression profiling using long-read technologies.

Download Full-text

GEO2RNAseq: An easy-to-use R pipeline for complete pre-processing of RNA-seq data

10.1101/771063 ◽

2019 ◽

Cited By ~ 2

Author(s):

Bastian Seelbinder ◽

Thomas Wolf ◽

Steffen Priebe ◽

Sylvie McNamara ◽

Silvia Gerber ◽

...

Keyword(s):

Gene Expression ◽

Single Species ◽

Gene Expression Omnibus ◽

Rna Seq ◽

Sequencing Data ◽

Interacting Species ◽

Link Type ◽

Fastq Format ◽

Standard Tool ◽

Processing Steps

ABSTRACTIn transcriptomics, the study of the total set of RNAs transcribed by the cell, RNA sequencing (RNA-seq) has become the standard tool for analysing gene expression. The primary goal is the detection of genes whose expression changes significantly between two or more conditions, either for a single species or for two or more interacting species at the same time (dual RNA-seq, triple RNA-seq and so forth). The analysis of RNA-seq can be simplified as many steps of the data pre-processing can be standardised in a pipeline.In this publication we present the “GEO2RNAseq” pipeline for complete, quick and concurrent pre-processing of single, dual, and triple RNA-seq data. It covers all pre-processing steps starting from raw sequencing data to the analysis of differentially expressed genes, including various tables and figures to report intermediate and final results. Raw data may be provided in FASTQ format or can be downloaded automatically from the Gene Expression Omnibus repository. GEO2RNAseq strongly incorporates experimental as well as computational metadata. GEO2RNAseq is implemented in R, lightweight, easy to install via Conda and easy to use, but still very flexible through using modular programming and offering many extensions and alternative workflows.GEO2RNAseq is publicly available at https://anaconda.org/xentrics/r-geo2rnaseq and https://bitbucket.org/thomas_wolf/geo2rnaseq/overview, including source code, installation instruction, and comprehensive package documentation.

Download Full-text