Here, University of Pennsylvania researchers develop a probability model to weigh a given RNA-Seq sample as a representative of an experimental condition when performing alternative splicing analysis. They demonstrate that this model detects outlier samples which are consistently and significantly different compared to other samples from the same condition. Moreover, they show that instead of discarding such samples the proposed weighting scheme can be used to downweight samples and specific splicing variations suspected as outliers, gaining statistical power. These weights can then be used for differential splicing (DS) analysis, where the resulting algorithm offers a generalization of the MAJIQ algorithm. Using both synthetic and real-life data the researchers perform an extensive evaluation of the improved MAJIQ algorithm in different scenarios involving perturbed samples, mislabeled samples, same condition groups, and different levels of coverage, showing it compares favorably to other tools. Overall, this work offers an outlier detection algorithm that can be combined with any splicing pipeline, a generalized and improved version of MAJIQ for differential splicing detection, and evaluation metrics with matching code and data for DS algorithms.

**Illustration of local splicing variations (LSV)**

**Availability**: Software and data are accessible via majiq.biociphers.org/norton_et_al_2017/.

Norton S, Vaquero-Garcia J, Lahens NF, Grant GR, Barash Y. (2017) **Outlier detection for improved differential splicing quantification from RNA-Seq experiments with replicates**. *Bioinformatics* [Epub ahead of print]. [abstract]

**Availability** – The new extensions and exemplar code for their usage are freely available online at: https://github.com/agaye/proper_extension

Gaye A. (2017) **Extending the R Library PROPER to Enable Power Calculations for Isoform-Level Analysis with EBSeq**. *Front Genet* 7:225. [article]

Researcher from the University of Dundee performed an RNA-seq experiment with 48 biological replicates in each of two conditions to answer these questions and provide guidelines for experimental design. With three biological replicates, eight of the 11 tools evaluated found only 20%-40% of the significantly differentially expressed (SDE) genes identified with the full set of 42 clean replicates. This rises to >85% for the subset of SDE genes changing in expression by more than fourfold.

To achieve >85% for all SDE genes regardless of fold change requires more than 20 biological replicates.

The same eight tools successfully control their false discovery rate at ≲5% for all numbers of replicates, while the remaining three tools fail to control their FDR adequately, particularly for low numbers of replicates.

*Hierarchical clustering of eleven RNA-seq DGE tools and five standard statistical tests using all of the full clean data set comprising 42 WT and 44 Δsnf2 replicates. For each tool, or test, a 7126-element long vector of 1’s and 0’s was constructed representing whether each gene in the annotation was called as SDE (adjusted P-value or FDR threshold ≤0.05) by the tool or not. The vectors for each tool and test were then ordered by the gene id and hierarchically clustered by Euclidian distance with complete linkage using the R package pvclust. Approximately unbiased P-value percentages (bracketed values) calculated for each branch in the clustering represent the support in the data for the observed sub-tree clustering. AU% > 95% are strongly supported by the data. AU% values are not shown for branch points where AU% = 100 for clarity. The outlier clustering of baySeq, DEGSeq, edgeR (GLM), and NOISeq suggest that these tools are clearly distinct from the other tools. Combined with the tool performance data shown in Figure 2, this suggests that, given a large number of replicates, the tools and tests in Cluster 1 are reliably and reproducibly converging on a similar answer, and are likely to be correctly capturing the SDE signal in the data.*

For future RNA-seq experiments, these results suggest that more than six biological replicates should be used, rising to more than 12 when it is important to identify SDE genes for all fold changes. If less than 12 replicates are used, a superior combination of true positive and false positive performances makesedgeRthe leading tool. For higher replicate numbers, minimizing false positives is more important and DESeq marginally outperforms the other tools.

Schurch NJ, Schofield P, Gierliński M, Cole C, Sherstnev A, Singh V, Wrobel N, Gharbi K, Simpson GG, Owen-Hughes T, Blaxter M, Barton GJ. (2016) **How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use?** *RNA* [Epub ahead of print]. [article]

Power calculation is a critical component of RNA-seq experimental design. The flexibility of RNA-seq experiment and the wide dynamic range of transcription it measures make it an attractive technology for whole transcriptome analysis. These features, in addition to the high dimensionality of RNA-seq data, bring complexity in experimental design, making an analytical power calculation no longer realistic.

Classical power calculation that deals with a single hypothesis takes a few simple assumptions. These include:

- the effect size, representing the minimum difference that is scientiﬁcally meaningful between groups in comparison;
- within-group variation, representing natural variation in observations regardless of between-group difference;
- an acceptable type I error rate, usually in the form of p-value; and
- the sample size.

With these values set, one can calculate statistical power, the probability of rejecting the null hypothesis when the effect is as large as assumed. If a certain power level is desired, one can also do a reverse calculation to determine the minimum sample size to achieve the desired power while controlling the type I error rate.

In DE analysis for RNA-seq experiments, we consider similar factors with more complexity since it is a high throughput experiment querying all transcripts simultaneously, and these transcripts are not exchangeable (read more…)

**Comprehensive visualization of stratified power, generated by the function plotAll**

**Availability** – PROspective Power Evaluation for RNAseq (PROPER) is avialable at: https://www.bioconductor.org/packages/release/bioc/html/PROPER.html

Wu Z, Wu H (2016) **Experimental Design and Power Calculation for RNA-seq Experiments**. *Methods Mol Bi*ol 1418:379-90. [abstract]

In this article, a team led by researchers at Emory University provide a maximum-likelihood framework for cis-eQTL mapping with RNA-seq data. Their approach integrates the inference of haplotypes and the association analysis into a single stage, and is thus unbiased and statistically powerful. The team also developed a pipeline for performing a comprehensive scan of all local eQTLs for all genes in the genome by controlling for false discovery rate, and implemented the methods in a computationally efficient software program. The advantages of the proposed methods over the existing ones are demonstrated through realistic simulation studies and an application to empirical breast cancer data from The Cancer Genome Atlas project.

*Power of the ASE model for testing a cis-eQTL in the rst simulation study. The nominal signicance level is 0.05. MLE, IMP, and TRUE are the maximum-likelihood method, the two-stage method, and the method using the true phase, respectively. Each power estimate is based on 10,000 replicates.*

**Availability** – TRECASE_MLE is available at: http://web1.sph.emory.edu/users/yhu30/software.html

Hu YJ, Sun W, Tzeng JY, Perou CM. (2015) **Proper Use of Allele-Specific Expression Improves Statistical Power for cis-eQTL Mapping with RNA-Seq Data**. *J Am Stat Assoc* 110(511):962-974. [abstract]