| | 24 | |
| | 25 | == How many replicates are needed for a RNA-seq experiment? == |
| | 26 | |
| | 27 | * One resource is the [https://bioconductor.org/packages/release/bioc/html/RNASeqPower.html RNASeqPower] R package described in [https://www.ncbi.nlm.nih.gov/pubmed/23961961 Hart et al., 2019] |
| | 28 | * A good resource to help with this analysis is [https://www.ncbi.nlm.nih.gov/pubmed/25246651 Ching et al.] |
| | 29 | * The main command from RNASeqPower is '''rnapower(depth, n, cv, effect, alpha, power)''', where one can enter all values but one, and the algorithm will calculate the missing value |
| | 30 | * The inputs are |
| | 31 | * depth = the number of reads (of a median gene, which according to Ching et al. is abou 16-32 per million mapped reads) |
| | 32 | * n = sample size (number of biological replicates per group) |
| | 33 | * cv = coefficient of variation (sqrt(disperson)), which can vary between 0.2 (cell lines) and 0.5 (animals or human subjects); see the above references for a discussion of typical values for different types of experiments |
| | 34 | * effect = effect size that one wants to identify (such as 2 for a 2-fold difference) |
| | 35 | * alpha = test statistic (such as FDR) threshold, such as 0.05 |
| | 36 | * power = power of the test (the fraction of true positives that will be detected); can be calculated or set to a value like 0.8 |
