This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ PrePrints) and either DOI or URL of the article must be cited.
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 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, 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 the biologist. Results: Using two large public RNA-seq data sets - one representing strong, and another mild, biological response, we simulated different replicate size scenarios and tested the performance of several commonly-used methods for calling differentially expressed genes in each of them. Our results suggest that if the biological response of interest in the different phenotype classes is expected to be mild, then RNA-seq experiments should focus on validation of differentially expressed gene candidates. At least triplicates must be used, and the differentially expressed genes should be called using methods with high positive predictive value such as NOISeq or GFOLD. In contrast, for strong biological response, differentially expressed genes mined from unreplicated experiments using NOISeq, ASC and GFOLD had between 30 to 50% mean positive predictive value, an increase of more than 30-fold compared to the case of mild biological response. Among methods with good positive predictive value performance, having triplicates or more substantially improved mean positive predictive value to over 90% for GFOLD, 60% for DESeq2, 50% for NOISeq, and 30% for edgeR. We found DESeq2 to be the most reasonable method to call differentially expressed genes for systems level analysis as it showed the best PPV and sensitivity trade-off (mean PPV and mean sensitivity ∼ 65% at replicate size of six). Conclusion: When biological effect size is strong, NOISeq and GFOLD are effective tools for detecting differentially expressed genes in unreplicated RNA-seq experiments for validation work. Having triplicates or more enables DESeq2 to detect sufficiently large numbers of reliable gene candidates for downstream systems level analysis. When biological effect size is weak, systems level investigation is not possible, and no meaningful result can be obtained in unreplicated experiments. Nonetheless, NOISeq or GFOLD may yield limited numbers of candidates with good validation potential when triplicates or more are available.
This is version 1.0 of a submission to PeerJ for review.
"Following" is like subscribing to any updates related to a preprint.
These updates will appear in your home dashboard each time you visit PeerJ.
You can also choose to receive updates via daily or weekly email digests.
If you are following multiple preprints then we will send you
no more than one email per day or week based on your preferences.
Note: You are now also subscribed to the subject areas of this preprint
and will receive updates in the daily or weekly email digests if turned on.
You can add specific subject areas through your profile settings.