As a common practice, we use the rule of a 1 cycle difference as a meaningful difference in quantitative PCR assays. That would equate to a 2-fold difference and is a useful rule of thumb across assays. Still, I would use formal statistical comparison: We use the measured cycle threshold number rather than the expression levels calculated from the cycle number. Thus, depending on the number of repeats, the level of expression, the biologic variation across samples and the experimental precision you can run the appropriate stats tests (t-test or ANOVA) to get a sense of statistical significance of differences. However, for a cycle difference of less than 1 it would be meaningful to use an independent set of primers for confirmation.

Thanks Dr. Anton for your useful suggestions. Would you please provide me a reference article for why 2-fold change is meaningful biologically.

Another important issue is the use of replicates in the assay to get correct statistical significance. For example I am doing qPCR arrays in 4x96-well plates for comprehensive miRNA expression in cancer models. It is not cost effective when you run more than three replicates. But while running three replicates, the p-value cut-off .o5 or not statistically significant. If the analysis is biased to stringent statistical significance it often dampens the high fold change interpretation. What would you suggest in this situation. Thanks again.

The +/- 1 cycle is a rule of thumb based on experimental variation one sees with the threshold method of qPCR. Whether a twofold difference is BIOLOGICALLY meaningful or not will depend on the gene studied. We also study miRs in tissues and in the circulation and use that as a rule. I attach two papers we published where you can see the way we analyzed the data.

The most important thing is the biological relevance of the difference you observe.The more your difference is low, the more you need to confirm the result: first by doing at least three independant experiments and second by using another method if possible, such as transcriptional fusions for example or analysis at the protein level when possible. When you have a 100 fold change with no statistical significance, this is probably due to an important standard deviation in your experiments. This can be due to your starting material that is not sufficiently standardized.

the threshold can also depend on how many replicates you have, i.e., the more (consistent) replicates, the smaller the difference needs to be to be statistically significant.

for obvious reasons it's impractical to run 15 replicates therefore, exactly as Anton mentioned, you need to correlate a certain change with a biological meaning. I'd say that 1.5FC can be enough if there is good statistics behind it and a biological meaning to it.