Rxivist logo

A general and powerful stage-wise testing procedure for differential expression and differential transcript usage

By Koen Van den Berge, Charlotte Soneson, Mark D Robinson, Lieven Clement

Posted 16 Feb 2017
bioRxiv DOI: 10.1101/109082 (published DOI: 10.1186/s13059-017-1277-0)

Background: Reductions in sequencing cost and innovations in expression quantification have prompted an emergence of RNA-seq studies with complex designs and data analysis at transcript resolution. These applications involve multiple hypotheses per gene, leading to challenging multiple testing problems. Conventional approaches provide separate top-lists for every contrast and false discovery rate (FDR) control at individual hypothesis level. Hence, they fail to establish proper gene-level error control, which compromises downstream validation experiments. Tests that aggregate individual hypotheses are more powerful and provide gene-level FDR control, but in the RNA-seq literature no methods are available for post-hoc analysis of individual hypotheses. Results: We introduce a two-stage procedure that leverages the increased power of aggregated hypothesis tests while maintaining high biological resolution by post-hoc analysis of genes passing the screening hypothesis. Our method is evaluated on simulated and real RNA-seq experiments. It provides gene-level FDR control in studies with complex designs while boosting power for interaction effects without compromising the discovery of main effects. In a differential transcript usage/expression context, stage-wise testing gains power by aggregating hypotheses at the gene level, while providing transcript-level assessment of genes passing the screening stage. Finally, a prostate cancer case study highlights the relevance of combining gene with transcript level results. Conclusion: Stage-wise testing is a general paradigm that can be adopted whenever individual hypotheses can be aggregated. In our context, it achieves an optimal middle ground between biological resolution and statistical power while providing gene-level FDR control, which is beneficial for downstream biological interpretation and validation.

Download data

  • Downloaded 720 times
  • Download rankings, all-time:
    • Site-wide: 41,835
    • In bioinformatics: 4,438
  • Year to date:
    • Site-wide: 146,166
  • Since beginning of last month:
    • Site-wide: 143,469

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide