dearseq: a variance component score test for RNA-seq differential analysis that effectively controls the false discovery rate

Gauthier, Marine; Agniel, Denis; Thiébaut, Rodolphe; Hejblum, Boris P.

doi:10.1093/nargab/lqaa093

Cited by 21 publications

(26 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In particular, on GTEx datasets, differential expression analysis can be performed between two tissues or cell types; on TCGA datasets, differential expression analysis can be performed between two disease statuses or biological conditions. The four representative methods include two popular methods limma-voom [ 14 , 15 ] and NOISeq [ 16 ], a new method dearseq [ 11 ] (which claimed to overcome the FDR inflation issue of DESeq2 and edgeR on large-sample-size data), and the classic Wilcoxon rank-sum test [ 17 ]. Note that DESeq2, edgeR, and limma-voom are parametric methods that assume parametric models for data distribution, while NOISeq, dearseq, and the Wilcoxon rank-sum test are non-parametric methods that are less restrictive but require large sample sizes to have good power.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Exaggerated false positives by popular differential expression methods when analyzing human population samples

et al. 2022

View full text Add to dashboard Cite

When identifying differentially expressed genes between two conditions using human population RNA-seq samples, we found a phenomenon by permutation analysis: two popular bioinformatics methods, DESeq2 and edgeR, have unexpectedly high false discovery rates. Expanding the analysis to limma-voom, NOISeq, dearseq, and Wilcoxon rank-sum test, we found that FDR control is often failed except for the Wilcoxon rank-sum test. Particularly, the actual FDRs of DESeq2 and edgeR sometimes exceed 20% when the target FDR is 5%. Based on these results, for population-level RNA-seq studies with large sample sizes, we recommend the Wilcoxon rank-sum test.

show abstract

Section: Resultsmentioning

confidence: 99%

“…From the literature, we found that several studies had reported the anticonservative behavior of DESeq2 and edgeR [9][10][11]; however, they were restricted to using simulated datasets with small sample sizes. Hence, for our large-sample-size scenario, their findings did not provide a direct answer to our question.…”

mentioning

confidence: 99%

Exaggerated false positives by popular differential expression methods when analyzing human population samples

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The adjusted P- value and [logFC] were calculated. The Benjamini & Hochberg false discovery rate method was used as a correction factor for the adjusted P-value in DESeq2 [41]. The statistically significant DEGs were identified according to adjusted P < .05, and [logFC] > 2.576 for up regulated genes and [logFC] < -2.813 for down regulated genes.…”

Section: Methodsmentioning

confidence: 99%

Comprehensive analysis of next-generation sequencing data in COVID-19 and its secondary complications

Vastrad¹,

Vastrad²

2022

Preprint

View full text Add to dashboard Cite

The ongoing pandemic of coronavirus disease 2019 (COVID-19) has made a serious public health threat globally. To discover key molecular changes in COVID-19 and its secondary complications, we analyzed next-generation sequencing (NGS) data of COVID-19. NGS data (GSE163151) was screened and downloaded from the Gene Expression Omnibus database (GEO). Differentially expressed genes (DEGs) were identified in the present study, using DESeq2 package in R programming software. Gene ontology (GO) and pathway enrichment analysis were performed, and the protein-protein interaction (PPI) network, module analysis, miRNA-hub gene regulatory network and TF-hub gene regulatory network were established. Subsequently, receiver operating characteristic curve (ROC) analysis was used to validate the diagonostics valuesof the hub genes. Firstly, 954 DEGs (477 up regulated and 477 down regulated) were identified from the four NGS dataset. GO enrichment analysis revealed enrichment of DEGs in genes related to the immune system process and multicellular organismal process, and REACTOME pathway enrichment analysis showed enrichment of DEGs in the immune system and formation of the cornified envelope. Hub genes were identified from the PPI network, module analysis, miRNA-hub gene regulatory network and TF-hub gene regulatory network. Furthermore, the ROC analysis indicate that COVID-19 and its secondary complications with following hub genes, namely, RPL10, FYN, FLNA, EEF1A1, UBA52, BMI1, ACTN2, CRMP1, TRIM42 and PTCH1, had good diagnostics values. This study identified several genes associated with COVID-19 and its secondary complications, which improves our knowledge of the disease mechanism.

show abstract

“…Differentially expressed genes (DEGs) between T1DM samples and normal control samples were identified by analyzing NGS data with DESeq2 package of R software [24]. The Benjamini and Hochberg (BH) method was accomplished to adjust P value to reduce the false discover rate [25]. According to the standard, we used FC > 3.835 as the screening criterion for up regulation of DEGs, FC < < 0 for down regulation of DEGs and adjust P value was < 0.05.…”

Section: Methodsmentioning

confidence: 99%

Bioinformatics Analysis of next generation sequencing data for Risk Prediction in Patients with Type 1 diabetes mellitus

Vastrad¹,

Vastrad²

2022

Preprint

View full text Add to dashboard Cite

Type 1 diabetes mellitus (T1DM) comprise the most common forms of autoimmune disease. The aim of this investigation was to apply a bioinformatics approach to reveal related pathways or genes involved in the development of T1DM. The next generation sequencing (NGS) dataset GSE182870 was downloaded from the gene expression omnibus (GEO) database. Differentially expressed gene (DEG) analysis was performed using DESeq2. The g:Profiler was utilized to analyze the functional enrichment, gene ontology (GO) and REACTOME pathway of the differentially expressed genes. Protein-protein interaction (PPI) network, modules, miRNA-hub gene regulatory network, and TF-hub gene regulatory network were conducted via comprehensive target prediction and network analyses. Finally, hub genes were validated by using receiver operating characteristic curve analysis. A total of 860 DEGs were screened out from NGS dataset, among which 477 genes were up regulated and 383 genes were down regulated. GO enrichment analysis indicated that up regulated genes were mainly involved in cellular metabolic process, intracellular anatomical structure and catalytic activity, and the down regulated genes were significantly enriched in cellular nitrogen compound biosynthetic process, protein containing complex and heterocyclic compound binding. REACTOME pathway enrichment analysis showed that the up regulated genes were mainly enriched in metabolism of carbohydrates and the down regulated genes were significantly enriched in metabolism of RNA. A PPI network, modules, miRNA-hub gene regulatory network and TF-hub gene regulatory network were constructed, and hub genes (MAPK14, RHOC, MAD2L1, TAF1, TRAF2, HSP90AA1, TP53, HSP90AB1, UBA52 and RACK1) were identified. This study provides further insights into the underlying pathogenesis of T1DM.

show abstract

dearseq: a variance component score test for RNA-seq differential analysis that effectively controls the false discovery rate

Cited by 21 publications

References 33 publications

Exaggerated false positives by popular differential expression methods when analyzing human population samples

Exaggerated false positives by popular differential expression methods when analyzing human population samples

Comprehensive analysis of next-generation sequencing data in COVID-19 and its secondary complications

Bioinformatics Analysis of next generation sequencing data for Risk Prediction in Patients with Type 1 diabetes mellitus

Contact Info

Product

Resources

About