No----depending on the comparator design (paired vs unpaired), the overall range of expression (is it near the threshold for detection) and the numbers of subjects. I would look for a 1.5 to 2 fold increase at a minimum to consider it biologically significant (if it is also statistically significant including a correction for FDR [false discovery rate]). It is important to consider where the raw intensity values fall compared to the entire dynamic range of expression you are measuring. A decrease to 50% of the control would be equivalent to a 2-fold decline in expression. But be careful with low expressers because you can be misled by an expression level near threshold which is essentially "off"; and then it appears to have a "huge" fold change value when the transcript essentially becomes detectable above threshold. In fact, the biological effect may be insignificant if you compare the "increased" values to the actual dynamic range of expression for the entire transcriptome. Consult with a stats pro for normalization approaches (log2 or not/ median array normalization/quartile normalization etc.) as this is a critical step in comparing subjects.

No----depending on the comparator design (paired vs unpaired), the overall range of expression (is it near the threshold for detection) and the numbers of subjects. I would look for a 1.5 to 2 fold increase at a minimum to consider it biologically significant (if it is also statistically significant including a correction for FDR [false discovery rate]). It is important to consider where the raw intensity values fall compared to the entire dynamic range of expression you are measuring. A decrease to 50% of the control would be equivalent to a 2-fold decline in expression. But be careful with low expressers because you can be misled by an expression level near threshold which is essentially "off"; and then it appears to have a "huge" fold change value when the transcript essentially becomes detectable above threshold. In fact, the biological effect may be insignificant if you compare the "increased" values to the actual dynamic range of expression for the entire transcriptome. Consult with a stats pro for normalization approaches (log2 or not/ median array normalization/quartile normalization etc.) as this is a critical step in comparing subjects.

Have you tried calculating z-scores for your analysis? That might help you gauge as a starting point on what a good fold-change to start using. We calculate z-scores for every analysis and set our cut-offs as a start using 1 or 2 s.d. from the sample mean.

Try using Q significant hits rather all DEGs, in terms of LogFC generally a value greater than 2 is acceptable in most of the experiments. Q-Significant DEGs are statistical very significant and logFC >2 is very in terms of differential expression while comparing control and test cases.

Hope this helps.

Best Wishes

Pankaj

It totally depends on the experimental study. For general study, where overall gene expression level will be high, we keep |logFc| >2 as threshold but some in cases, for instance, in case of,Japanese encephalitis, where overall gene expression level will be minimal, we keep | logFc | >1 .

You can follow our two papers, if you want,

https://www.researchgate.net/publication/318700627_Identification_and_characterization_of_differentially_expressed_genes_from_human_microglial_cell_samples_infected_with_Japanese_encephalitis_virus

https://www.researchgate.net/publication/309366063_In_silico_analysis_of_differential_gene_expressions_in_biliary_stricture_and_hepatic_carcinoma