Hi, i made a RT-qPCR to acrB gene in two strains to compare their expression, and I used like a refernce gene: gapA. I obtained expression of genes in each strain and then we quantifed relative gene expression to gene acrB .This was calculated using the following formula (= gene expression acrB / reference gene expression). Once obtained these values in both strains, were divided to calculate the difference between acrB's expression between these two strains, it was 0.967. Now I have a dilemma, because my chief said that:
- It was determined that there wasn't a significant difference in gene expression between these two strains , because that is in the case where is ≥20, that indicates an overexpression. -
But I searched information about that, for example in: The MIQE Guidelines: Minimum Information for Publication of Quantitative Real-Time PCR Experiments (http://www.clinchem.org/content/55/4/611.full.pdf+html) and I didn't find any explanation to resolve my dilemma.
I found too that some people say that 2-fold is a significant difference.
Then......whats the value?
Unfortunately there are at least two different meanings of the word "significant".
It may be used to indicate that the finding is "relevant" to some aspect. The judgement of relevance surely depends on the field of research, the particular process under study and its relation to our understanding or to practical consideartions (therapeutic implication, economic implications, etc...).
The other meaning is a statistical thing. Statistical significance is a measure to express the expectation of data that is at least as "extreme" as the one observed, and this expectation is based on some assumptions (a statistical model and a particular hypothesis within this model). One may see the statistical significance as some kind of signal-to-noise ratio (SNR)*. When this SNR is large enough, one may become convinced that one observed a trustable "signal". Some people call such findings "significant", when the observed SNR is "rather high" (how high is again a matter of subject-specific judgement). Relying only on looking at such SNRs can be misleading, because one misses to see the reason for a high SNR: it could be a large signal with some noise or a tiny signal with even lower noise.
---
*practically, the stat. significance is usually given a a "p-value" expressing the probability of more extreme data than the one observed given a statistical model and a defined null-hypothesis. Low p-values indicate that the observed data is really unexpected under the null-hypothesis, and this is taken as evidence against the null-hypothesis.
Unfortunately there are at least two different meanings of the word "significant".
It may be used to indicate that the finding is "relevant" to some aspect. The judgement of relevance surely depends on the field of research, the particular process under study and its relation to our understanding or to practical consideartions (therapeutic implication, economic implications, etc...).
The other meaning is a statistical thing. Statistical significance is a measure to express the expectation of data that is at least as "extreme" as the one observed, and this expectation is based on some assumptions (a statistical model and a particular hypothesis within this model). One may see the statistical significance as some kind of signal-to-noise ratio (SNR)*. When this SNR is large enough, one may become convinced that one observed a trustable "signal". Some people call such findings "significant", when the observed SNR is "rather high" (how high is again a matter of subject-specific judgement). Relying only on looking at such SNRs can be misleading, because one misses to see the reason for a high SNR: it could be a large signal with some noise or a tiny signal with even lower noise.
---
*practically, the stat. significance is usually given a a "p-value" expressing the probability of more extreme data than the one observed given a statistical model and a defined null-hypothesis. Low p-values indicate that the observed data is really unexpected under the null-hypothesis, and this is taken as evidence against the null-hypothesis.
Hi Marc, I am always suspicious of only 2-fold increases because this is well within the experimental error. The thing to do is repeat, repeat, repeat. Each time with a fresh RNA and cDNA sample. This way at least if it is always 2-fold, it is more significant than if you only did it once. So, back to the lab!
I will add to Jochen's answer that, as a journal editor, I beg you (and everyone else) to only use the word "significant" in a manuscript when indicating mathematical significance.
- Badges
- Science topic