Robust principal component analysis for accurate outlier sample detection in RNA-Seq data

Introduction: Colon cancer is the fourth most common cancer in the United States and the third leading cause of cancer-related death. High throughput genomic sequencing has led to a number of significant advances in tumor biology and in identifying novel signaling molecules such as long non-coding RNA (lncRNA). The aim of this study was to identify differentially expressed lncRNA in colon cancer from a large RNA-sequencing (RNA-seq) data set. Methods: The raw RNA-seq files of 398 patients with colon cancer were downloaded from The Cancer Genome Atlas (TCGA). Sequencing files were aligned using STAR (Spliced Transcripts Alignment to a Reference) to the most recent genome annotation. Using a subset of patients with paired colon cancer and normal colon epithelium RNA-seq data (n=40), an exploratory binomial regression model was used to calculate differential RNA expression. The most differentially expressed lncRNA were identified from the exploratory analysis and verified by comparing the larger colon cancer RNA-seq data set (n=358) with the normal colon epithelium RNA-seq data set (n=40). Results: 33,514 genes were identified from the comparative analysis, and using differential expression cut off values of > +1.5 or < -1.5 log fold change and a false discovery rate of <0.05; 543 were upregulated and 1822 were downregulated. Within this dysregulated group, 60 lncRNAs were identified. Using the larger data set (n=358 vs n=40), 41/60 lncRNA remained differentially expressed, of which 15 were downregulated and 26 were upregulated. Twenty-four of these lncRNAs have not previously been described in colon cancer, 12 of which are upregulated and 12 of which are downregulated (Table 1). Conclusions: This analysis of RNA-seq data from TCGA has identified dysregulated lncRNAs which have not previously been described in human colon cancer. These lncRNAs may have significant roles in colon cancer tumor signaling. Table 1.Differential expression of lncRNA not described in colon cancerlncRNALog Fold ChangeP-valueDescription in other cancersKRT16P14.1041.36E-21Increased in lung squamous cell carcinomaBOK-AS13.7784.98E-09Increased in oral squamous cell carcinomaCELP3.1151.30E-09Not availableCLDN10-AS12.3944.61E-12Not availablePOU6F2-AS12.2678.25E-09Not availableSLC7A11-AS12.1931.60E-13Decreased in gastric adenocarcinomaCSAG42.1083.59E-10Not availableLUCAT12.0224.21E-11Increased in Ovarian cancer, renal clear cell carcinoma, head and neck squamous cell carcinoma, Non-small cell cancer, glioma, osteosarcoma, esophageal squamous cell carcinomaVPS9D1-AS11.9795.27E-21Decreased in gastric adenocarcinoma, Increased in non-small cell lung cancerLEF1-AS11.8464.48E-13Increased in glioblastomaCASC81.6485.64E-06Single nucleotide polymorphisms in colorectal adenocarcinomaLINC004911.5903.11E-04Not availableLINC00648-1.5021.02E-15Not availableLINC00163-1.6751.03E-22Not availableSATB2-AS1-1.8074.18E-09Increased in osteosarcomaLINC00702-2.1424.81E-18Not availableTRHDE-AS1-2.2623.44E-24Not availableLINC00908-2.3082.76E-103Not availableLINC00483-2.4012.87E-39Increased in gastric adenocarcinomaLINC00461-2.4401.42E-41Increased in glioma, multiple myelomaFENDRR-2.6872.12E-58Decreased in breast cancer, prostate cancer, osteosarcoma, gastric adenocarcinoma,ARHGEF26-AS1-2.7148.55E-32Not availableADAMTS9-AS2-2.9041.78E-74Increased in lung cancer salivary adenoid cystic carcinoma, Decreased in glioma, gastric adenocarcinomaMT1JP-4.4601.08E-114Decreased in retinoblastoma, Decreased in gastric adenocarcinoma Citation Format: Stephen J. O'Brien, Theodore Kalbfleisch, Sudhir Srivastava, Shesh Rai, Susan Galandiuk. Differential expression of long non-coding RNA in colon adenocarcinoma RNA-sequence data set [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1817.

Transcriptome deep sequencing (RNA-seq) data analysis is often affected by reference bias introduced by the use of a single linear reference (SLR) genome. Graph-based pangenomes can mitigate this bias by integrating the SLR genome with complex genetic variations within a species; however, their application remains limited owing to a lack of dedicated analytical tools. Here, we present PanGraphRNA, an integrated bioinformatics platform for RNA-seq data analysis using a graph pangenome as reference. Built on the Galaxy web-based framework, PanGraphRNA provides functional modules for constructing, evaluating, and applying graph pangenomes across different population scales, thus enabling accessibility, traceability, and reproducibility throughout the analysis. Applied to both real and simulated RNA-seq data sets from Arabidopsis (Arabidopsis thaliana), PanGraphRNA outperformed the SLR approach, achieving higher read alignment accuracy and more precise gene expression quantification. PanGraphRNA enabled the identification of drought stress-induced genes and flowering time-related quantitative trait loci that were previously missed with the conventional SLR approach. Furthermore, we successfully applied PanGraphRNA to process RNA-seq data sets from rice (Oryza sativa) and maize (Zea mays). By providing standardized, containerized workflows, PanGraphRNA will facilitate transcriptomic research in key plant species, including Arabidopsis, rice, and maize.

With the development of high-throughput techniques, RNA-sequencing (RNA-seq) is becoming increasingly popular as an alternative for gene expression analysis, such as RNAs profiling and classification. Identifying which type of diseases a new patient belongs to with RNA-seq data has been recognized as a vital problem in medical research. As RNA-seq data are discrete, statistical methods developed for classifying microarray data cannot be readily applied for RNA-seq data classification. Witten proposed a Poisson linear discriminant analysis (PLDA) to classify the RNA-seq data in 2011. Note, however, that the count datasets are frequently characterized by excess zeros in real RNA-seq or microRNA sequence data (i.e. when the sequence depth is not enough or small RNAs with the length of 18-30 nucleotides). Therefore, it is desired to develop a new model to analyze RNA-seq data with an excess of zeros. In this paper, we propose a Zero-Inflated Poisson Logistic Discriminant Analysis (ZIPLDA) for RNA-seq data with an excess of zeros. The new method assumes that the data are from a mixture of two distributions: one is a point mass at zero, and the other follows a Poisson distribution. We then consider a logistic relation between the probability of observing zeros and the mean of the genes and the sequencing depth in the model. Simulation studies show that the proposed method performs better than, or at least as well as, the existing methods in a wide range of settings. Two real datasets including a breast cancer RNA-seq dataset and a microRNA-seq dataset are also analyzed, and they coincide with the simulation results that our proposed method outperforms the existing competitors. The software is available at http://www.math.hkbu.edu.hk/∼tongt. xwan@comp.hkbu.edu.hk or tongt@hkbu.edu.hk. Supplementary data are available at Bioinformatics online.

BackgroundRNA sequencing (RNA-seq) is a powerful tool for genome-wide expression profiling of biological samples with the advantage of high-throughput and high resolution. There are many existing algorithms nowadays for quantifying expression levels and detecting differential gene expression, but none of them takes the misaligned reads that are mapped to non-exonic regions into account. We developed a novel algorithm, XBSeq, where a statistical model was established based on the assumption that observed signals are the convolution of true expression signals and sequencing noises. The mapped reads in non-exonic regions are considered as sequencing noises, which follows a Poisson distribution. Given measureable observed and noise signals from RNA-seq data, true expression signals, assuming governed by the negative binomial distribution, can be delineated and thus the accurate detection of differential expressed genes.ResultsWe implemented our novel XBSeq algorithm and evaluated it by using a set of simulated expression datasets under different conditions, using a combination of negative binomial and Poisson distributions with parameters derived from real RNA-seq data. We compared the performance of our method with other commonly used differential expression analysis algorithms. We also evaluated the changes in true and false positive rates with variations in biological replicates, differential fold changes, and expression levels in non-exonic regions. We also tested the algorithm on a set of real RNA-seq data where the common and different detection results from different algorithms were reported.ConclusionsIn this paper, we proposed a novel XBSeq, a differential expression analysis algorithm for RNA-seq data that takes non-exonic mapped reads into consideration. When background noise is at baseline level, the performance of XBSeq and DESeq are mostly equivalent. However, our method surpasses DESeq and other algorithms with the increase of non-exonic mapped reads. Only in very low read count condition XBSeq had a slightly higher false discovery rate, which may be improved by adjusting the background noise effect in this situation. Taken together, by considering non-exonic mapped reads, XBSeq can provide accurate expression measurement and thus detect differential expressed genes even in noisy conditions.

RNA sequencing (RNA-seq) is a highly sensitive method for transcriptome analysis that allows simultaneous assessment of expression of thousands of genes and identification of expression patterns under various conditions. The existing variety of RNA-seq data formats, normalization methods, and approaches to statistical processing of results complicates comparison of data from different studies and reduces reproducibility of the analysis. This study presents an automated pipeline PipeSeq that combines standard steps of RNA-seq data processing: loading (SRA Toolkit), read alignment to the reference genome (HISAT2), transcript assembly (StringTie), transcript counting (FeatureCounts) and statistical analysis of differential gene expression under various experimental conditions (DESeq2). PipeSeq has a simple visual interface, supports multithreading, and generates ready-to-analyze gene expression heat maps, tables and graphs. The functionality of the pipeline is demonstrated on three sets of raw RNA-seq data from the green alga Chlamydomonas reinhardtii cells available in the NCBI SRA database. The data from these experiments were used to analyze the differential expression of C. reinhardtii genes encoding the GATA family transcription factors under different light cultivation conditions. The data obtained by in silico methods were verified by real-time reverse transcription polymerase chain reaction (RT-qPCR) for 12 GATA genes, which allowed us to hypothesize their functions and evaluate the correlation between the bulk (RNA-seq) and targeted (RT-qPCR) approaches. Our results showed that RNA-seq and RT-qPCR methods reveal similar directions of gene expression changes, but demonstrate differences in the effect size and sensitivity, which emphasizes the need for a combined use of the two approaches. Thus, the PipeSeq program is a tool for conducting a full cycle of bioinformatic analysis of RNA-seq data, additionally providing the opportunity to process RT-qPCR data and perform a comparative statistical analysis of the results obtained.

High-quality RNA is crucial for obtaining reliable RNA sequencing (RNA-seq) data, with metrics like RNA Integrity Number (RIN). However, these metrics, while effective for evaluating RNA integrity, do not always correlate with RNA-seq data quality, especially at the transcript level. This gap is particularly evident in total RNA-seq, where existing measures such as the coefficient of variation (CV) for read coverage fail to fully capture data quality and are influenced by confounding factors like read coverage depth. To address this, we developed a novel method to assess RNA-seq data quality by quantifying nonuniformity in read coverage while minimizing the influence of read coverage depth. In this study we invented new matric called windowCV (wCV) and applied to a diverse range of RNA-seq datasets, including fresh frozen (FF) and FFPE total RNA-seq data, as well as poly(A)-enriched mRNA-seq data. In mRNA-seq, our method captured 3' read coverage bias, a hallmark of RNA degradation, particularly in longer transcripts. For total RNA-seq data, we identified noisy coverage patterns associated with poor data quality, even in samples with high RIN values. By fitting regression lines between wCV and mean coverage depth (MCD) and calculating the area under the curve (wCVAUC), we refined the assessment to account for RNA quality variability. Using the TCGA pilot study and our own datasets, we demonstrated that wCVAUC reliably identified low quality RNA-seq data and highlighted its impact on downstream analyses, including gene expression quantification and clustering. Importantly, we observed that low RIN values do not always predict poor RNA-seq data quality, as some samples with RIN values below 7 exhibited high-quality RNA-seq data based on wCVAUC. It means that our analyses showed that wCVAUC effectively distinguished high-quality from low-quality samples, including cases where traditional metrics like RIN and CV were insufficient. Additionally, our investigation into the relationships between nonuniformity of read coverage, exon GC content, and RNA localization revealed that the transcript-level RNA-seq data quality of lncRNA genes in FFPE samples is influenced by low exon GC content and nuclear localization. In conclusion, our method provides robust, transcript-level metrics for assessing RNA-seq data quality across platforms, enabling more accurate identification of low-quality data and minimizing biases in downstream analyses. This approach offers a new standard for integrating RNA-seq data quality with sample variability, particularly for challenging datasets such as FFPE and total RNA-seq. Citation Format: Wonyoung Choi, Miyeon Yeon, Jay Lee, Hyo Young Choi, David Neil Hayes. The new approach for measuring nonuniformity of read coverages reveals the quality of RNA-seq data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2489.

The standard shotgun proteomics data analysis strategy relies on searching MS/MS spectra against a context-independent protein sequence database derived from the complete genome sequence of an organism. Because transcriptome sequence analysis (RNA-Seq) promises an unbiased and comprehensive picture of the transcriptome, we reason that a sample-specific protein database derived from RNA-Seq data can better approximate the real protein pool in the sample and thus improve protein identification. In this study, we have developed a two-step strategy for building sample-specific protein databases from RNA-Seq data. First, the database size is reduced by eliminating unexpressed or lowly expressed genes according to transcript quantification. Second, high-quality nonsynonymous coding single nucleotide variations (SNVs) are identified based on RNA-Seq data, and corresponding protein variants are added to the database. Using RNA-Seq and shotgun proteomics data from two colorectal cancer cell lines SW480 and RKO, we demonstrated that customized protein sequence databases could significantly increase the sensitivity of peptide identification, reduce ambiguity in protein assembly, and enable the detection of known and novel peptide variants. Thus, sample-specific databases from RNA-Seq data can enable more sensitive and comprehensive protein discovery in shotgun proteomics studies.

Chapter 19 - Dimensionality Reduction and Latent Variable Modeling

The smuggling of liquid precursor chemicals greatly harms social public security. In this paper a method based on principal component analysis (PCA) is presented to recognize the EDXRS spectra of liquid precursor chemicals. Scatter profiles of acetone, acetic anhydride,2-butanone and isosafrole are obtained by using EDXRS. As the similarity of the profiles, it is difficult to identify samples, so classical PCA and robust PCA are applied to extract features, then K-nearest neighbor is proposed for pattern classification, the classification results are satisfactory. It is proved that this method can be used to detect liquid precursor chemicals.

Reproducibility is a fundamental expectation in science and enables investigators to have confidence in their research findings and the ability to compare data from disparate sources, but evaluating reproducibility can be elusive. For example, generating RNA sequencing (RNA-seq) data includes multiple steps where variance can be introduced. Thus, it is unclear if RNA-seq data from different sources can be validly compared. While most studies on RNA-seq reproducibility focus on eukaryotes, we evaluate bias in bacteria using Pseudomonas aeruginosa gene expression data from five laboratory models of cystic fibrosis. We leverage a large data set that includes samples prepared in three different laboratories and paired data sets where the same sample was sequenced using at least two different sequencing pipelines. We report here that expression data are highly reproducible across laboratories. In addition, while samples sequenced with different sequencing pipelines showed significantly more variance in expression profiles than between labs, gene expression was still highly reproducible between sequencing pipelines. Further investigation of expression differences between two sequencing pipelines revealed that library preparation methods were the largest source of error, though analyses to identify the source of this variance were inconclusive. Consistent with the reproducibility of expression between sequencing pipelines, we found that different pipelines detected over 80% of the same differentially expressed genes with large expression differences between conditions. Thus, bacterial RNA-seq data from different sources can be validly compared, facilitating the ability to advance understanding of bacterial behavior and physiology using the wide array of publicly available RNA-seq data sets.IMPORTANCERNA sequencing (RNA-seq) has revolutionized biology, but many steps in RNA-seq workflows can introduce variance, potentially compromising reproducibility. While reproducibility in RNA-seq has been thoroughly investigated in eukaryotes, less is known about pipelines and workflows that introduce variance and biases in bacterial RNA-seq data. By leveraging Pseudomonas aeruginosa transcriptomes in cystic fibrosis models from different laboratories and sequenced with different sequencing pipelines, we directly assess sources of bacterial RNA-seq variance. RNA-seq data were highly reproducible, with the largest variance due to sequencing pipelines, specifically library preparation. Different sequencing pipelines detected overlapping differentially expressed genes, especially those with large expression differences between conditions. This study confirms that different approaches to preparing and sequencing bacterial RNA libraries capture comparable transcriptional profiles, supporting investigators' ability to leverage diverse RNA-seq data sets to advance their science.

Background. A common research goal in transcriptome projects is to find genes that are differentially expressed in different phenotype classes. Biologists might wish to validate such gene candidates experimentally, or use them for downstream systems biology analysis. Producing a coherent differential gene expression analysis from RNA-seq count data requires an understanding of how numerous sources of variation such as the replicate size, the hypothesized biological effect size, and the specific method for making differential expression calls interact. We believe an explicit demonstration of such interactions in real RNA-seq data sets is of practical interest to biologists.Results. Using two large public RNA-seq data sets—one representing strong, and another mild, biological effect size—we simulated different replicate size scenarios, and tested the performance of several commonly-used methods for calling differentially expressed genes in each of them. We found that, when biological effect size was mild, RNA-seq experiments should focus on experimental validation of differentially expressed gene candidates. Importantly, at least triplicates must be used, and the differentially expressed genes should be called using methods with high positive predictive value (PPV), such as NOISeq or GFOLD. In contrast, when biological effect size was strong, differentially expressed genes mined from unreplicated experiments using NOISeq, ASC and GFOLD had between 30 to 50% mean PPV, an increase of more than 30-fold compared to the cases of mild biological effect size. Among methods with good PPV performance, having triplicates or more substantially improved mean PPV to over 90% for GFOLD, 60% for DESeq2, 50% for NOISeq, and 30% for edgeR. At a replicate size of six, we found DESeq2 and edgeR to be reasonable methods for calling differentially expressed genes at systems level analysis, as their PPV and sensitivity trade-off were superior to the other methods’.Conclusion. When biological effect size is weak, systems level investigation is not possible using RNAseq data, and no meaningful result can be obtained in unreplicated experiments. Nonetheless, NOISeq or GFOLD may yield limited numbers of gene candidates with good validation potential, when triplicates or more are available. When biological effect size is strong, NOISeq and GFOLD are effective tools for detecting differentially expressed genes in unreplicated RNA-seq experiments for qPCR validation. When triplicates or more are available, GFOLD is a sharp tool for identifying high confidence differentially expressed genes for targeted qPCR validation; for downstream systems level analysis, combined results from DESeq2 and edgeR are useful.

Certain peripheral T-cell lymphomas (PTCLs) have been associated with viral infection, particularly infection with Epstein-Barr virus (EBV). However, a comprehensive virome analysis across PTCLs has not previously been reported. Here we utilized published whole-transcriptome RNA sequencing (RNA-seq) data sets from seven different PTCL studies and new RNA-seq data from our laboratory to screen for virus association, to analyze viral gene expression, and to assess B- and T-cell receptor diversity paradigms across PTCL subtypes. In addition to identifying EBV in angioimmunoblastic T-cell lymphoma (AITL) and extranodal NK/T-cell lymphoma (ENKTL), two PTCL subtypes with well-established EBV associations, we also detected EBV in several cases of anaplastic large-cell lymphoma (ALCL), and we found evidence of infection by the oncogenic viruses Kaposi's sarcoma-associated herpesvirus and human T-cell leukemia virus type 1 in isolated PTCL cases. In AITLs, EBV gene expression analysis showed expression of immediate early, early, and late lytic genes, suggesting either low-level lytic gene expression or productive infection in a subset of EBV-infected B-lymphocyte stromal cells. Deconvolution of immune cell subpopulations demonstrated a greater B-cell signal in AITLs than in other PTCL subtypes, consistent with a larger role for B-cell support in the pathogenesis of AITL. Reconstructed T-cell receptor (TCR) and B-cell receptor (BCR) repertoires demonstrated increased BCR diversity in AITLs, consistent with a possible EBV-driven polyclonal response. These findings indicate potential alternative roles for EBV in PTCLs, in addition to the canonical oncogenic mechanisms associated with EBV latent infection. Our findings also suggest the involvement of other viruses in PTCL pathogenesis and demonstrate immunological alterations associated with these cancers.IMPORTANCE In this study, we utilized next-generation sequencing data from 7 different studies of peripheral T-cell lymphoma (PTCL) patient samples to globally assess viral associations, provide insights into the contributions of EBV gene expression to the tumor phenotype, and assess the unique roles of EBV in modulating the immune cell tumor microenvironment. These studies revealed potential roles for EBV replication genes in some PTCL subtypes, the possible role of additional human tumor viruses in rare cases of PTCLs, and a role for EBV in providing a unique immune microenvironmental niche in one subtype of PTCLs. Together, these studies provide new insights into the understudied role of tumor viruses in PTCLs.

Robust principal component analysis (PCA) has been proved to be an efficient multivariate statistical method for extraction of a multivariate structure of geochemical data containing outlier values. Furthermore, because of the compositional characteristics of geochemical data, logratio transformation approaches also have to be implemented to transform the data from the simplex space into real space. In order to model the extracted principal component of transformed data, we used sequential Gaussian simulation (SGS) to overcome the drawbacks associated with the traditional kriging method, e.g. bias conditionally relevant to underestimation and overestimation, the smoothing effect and so on. The results of implementing robust PCA on both untransformed and logratio transformed data, which have been acquired from an orogenic gold deposit, indicate that robust PCA analysis of untransformed data is not capable of certifying element association of gold occurrences. However, in robust PCA analysis of transformed data, both varieties of gold mineralization, associated with pyrite mineral and quartz veins, have been revealed. The first variety of gold occurrence is detected in an integrated PC obtained from PC1, PC2 and PC3, which are associated with Pb, As, Hg, Mo, S and Sb elements in quartz–sulphide veins. The second one, PC4, could be correlated with quartz veins. Furthermore, the results of the geostatistical SGS approach are compatible with the detected anomaly in the Alut gold deposit.

BackgroundHeart failure (HF) is the end stage of various cardiovascular diseases with a high mortality rate. Novel diagnostic and therapeutic biomarkers for HF are urgently required. Our research aims to identify HF-related hub genes and regulatory networks using bioinformatics and validation assays.MethodsUsing four RNA-seq datasets in the Gene Expression Omnibus (GEO) database, we screened differentially expressed genes (DEGs) of HF using Removal of Unwanted Variation from RNA-seq data (RUVSeq) and the robust rank aggregation (RRA) method. Then, hub genes were recognized using the STRING database and Cytoscape software with cytoHubba plug-in. Furthermore, reliable hub genes were validated by the GEO microarray datasets and quantitative reverse transcription polymerase chain reaction (qRT-PCR) using heart tissues from patients with HF and non-failing donors (NFDs). In addition, R packages “clusterProfiler” and “GSVA” were utilized for enrichment analysis. Moreover, the transcription factor (TF)–DEG regulatory network was constructed by Cytoscape and verified in a microarray dataset.ResultsA total of 201 robust DEGs were identified in patients with HF and NFDs. STRING and Cytoscape analysis recognized six hub genes, among which ASPN, COL1A1, and FMOD were confirmed as reliable hub genes through microarray datasets and qRT-PCR validation. Functional analysis showed that the DEGs and hub genes were enriched in T-cell-mediated immune response and myocardial glucose metabolism, which were closely associated with myocardial fibrosis. In addition, the TF–DEG regulatory network was constructed, and 13 significant TF–DEG pairs were finally identified.ConclusionOur study integrated different RNA-seq datasets using RUVSeq and the RRA method and identified ASPN, COL1A1, and FMOD as potential diagnostic biomarkers for HF. The results provide new insights into the underlying mechanisms and effective treatments of HF.

In this paper, a new robust principal component analysis (RPCA) method which enables us to exploit the main components of a given corrupted data with non Gaussian outliers is proposed. This method is based on the $\alpha-$divergence which is a parametric measure from information geometry. The proposed method is adjustable using a hyperparameter $\alpha$ and reduces to the classical PCA as a particular case. In order to derive the main components, the $\alpha-$divergence between the empirical data distribution and the assumed model for the distribution is minimized with respect to the unknown parameters. The singular value decomposition (SVD) of the estimated covariance matrix is then used to exploit the main direction of the data. The proposed method is applied to some video and signal processing applications and the results show the superiority of the proposed method over classical PCA and other existing robust methods.