Dear All, How can I recognize (detect) ups and downs genes where "Limma" genes have been Significant? What threshold do you recommend to use in this matter? I was wondering If you could explain me why this threshold have been chosen.
Thank you so much
People conventionally use an FDR of 0.05, but depending on the experiment may choose different thresholds, or possibly include a fold-change criterion using e.g., the treat function. But what you do with your data is up to you as the analyst.
I generally draw a volcano plot to see where I can put threshold for adjusted pvalue and logFC. I most cases I use padj<0.05 and logFC>=0.58 (i.e. 1.5 fold up) or logFC<=-0.58 (i.e. 1.5 fold down) to identify differentially expressed genes. However, where you can put threshold to the logFC depend on your experiments so that you can capture biological insight. Hope this helps. -Tanay
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
To elaborate on James' answer: the FDR describes the expected proportion of false positives in your set of significantly DE genes. We usually use a 5% threshold because 5% of genes being false positives seems to be a tolerable proportion. However, the exact value should be chosen based on what you want to do with those DE genes. For example, if you're working in a setting where validation and follow-up studies are cheap, you might consider relaxing the FDR threshold to get more discoveries at the cost of a higher proportion of false positives. On the other hand, if you're going to do something expensive with your DE genes (e.g., setting up a knockout mouse strain), you might be inclined to be more stringent and use a lower threshold. So 5% is a good place to start, but as James has said, you will need to think a bit about what is most suitable for your setting.
Thank you for your consideration, but there is something misunderstanding for me, how can I detect UP and DOWN genes after detection the differentially expression genes? (In some sites I have seen they considered the fold change higher than 0 as UP and less than 0 as DOWN. It should be noted that lima use 'log2foldchange' and numbers less than 1 'log2' have taken value (amount) less than 0 that it has a different way than the way you guys used, so could you let me know what is your idea about this?)
I don't really understand your problem. Genes that have positive log-fold changes are going up. Genes with negative log-fold changes are going down. Look at the sign of the log-fold change in the output of
topTable.Thank you for your consideration, but there is something misunderstanding for me, how can I detect UP and DOWN genes after detection the differentially expression genes? (In some sites I have seen they considered the fold change higher than 0 as UP and less than 0 as DOWN. It should be noted that lima use 'log2foldchange' and numbers less than 1 'log2' have taken value (amount) less than 0 that it has a different way than the way you guys used, so could you let me know what is your idea about this?)