Dear colleagues,
I would like to perform Gene Set Enrichment Analysis (or some alternative analysis) on the data that contain only two samples, which represent two phenotypes (one sample in one phenotype). I already performed regular GSEA with tTest and permutation regarding gene sets, however due to small size groups the results are insufficient. I would appreciate if you suggest me any other solution or alternative analysis or maybe some other settings of GSEA that considers such small groups as well. I do not mind R in case.
Thank you in advance!
1 vs 1 comparison is unlikely to yield reliable results if replicates are not present. If your data are high-throughput this is even more true.
Thanks for your response.
I know about it, however there is no other method to generate those data. Probably, GSEA is not the method for it and we should find some compromise-like alternative.
Best regards!
Dear Magdalena, I do not understand the problem. Is it that you do not find any enriched process or that you don't have any differentially expressed gene? if the former is true, than you may try relaxing your thresholds, or reduce the input size such that less tests are performed (which affect how the FDR is calculated). If the latter is instead true, then you have to think why at first (maybe there are not important changes) then you may play a little bit with your thresholds (e.g. if you use min expression level, or minimum log2 fold change or a threshold on the fdr of being differentially expressed) to see if you can change something.
The single experiment case is not a problem in gsea unless you have differentially expressed genes. In case you don't have replicatesthis can create indeed problems but not to gsea, to the tool you use to get the differentially expressed genes. Replicates are in general used to get a measure of the variability of the gene expression level of your gene in a certain condition, such that then you can assess the significance of the different among the expression level of that gene in two different conditions. If you don't have any replicates, the variability is calculated for instance in cuffdiff, by considering the experiments as replicates to infer the dispersion parameters. this means that it will be much more difficult to identify differentially expressed genes.
- Badges
- Science topic
- Similar topics
- Biological Science
- Genetics
- Genetic Analysis