If I only have 1 sample of microarray study (n=1) and based on the results of microarray, I will increase samples size become (ie n=3 or n=5). Is it acceptable for publications?
I don't understand? You have a single array, or a single study (i.e. a single study using multiple samples and multiple microarrays)?
Microarray data is purely relative data - you need some context to compare arrays to draw inferences from. You can compare single arrays, one to another - for example, a single disease patient versus a healthy patient. In such a situation where samples are highly limited, you may wish to compare a single pair of arrays, and there are some tools to do that - for example, the SScore distribution can be use to compare individual Affymetrix cartridge arrays (i.e. arrays with both perfect match AND mis-match probe sets) with reasonable statistical power. This sort of comparison is not uncommon in clinical situations, where replication is simply not possible due to highly limited samples (e.g. tissues from surgical patients or some such clinical sampling study).
It is always best, however to have replication so you can compute robust statistical results. If looking purely at differential gene expression, then biological replication is truly essential, as you need to be able to distinguish population level differences from individual levels of variation in expression. You can only do that if you have reasonably sampled the inherent individual biological variation with replicates.
But a single array is pretty much useless - all it contains is relative expression signal, and without some other array or set of arrays to contrast it to, you really know nothing.
If the intent here is to publish, what was the question your research was attempting to answer in the first place? That will tell you what you minimally need in terms of samples, replicates and so forth.
if you want to do statistical analysis, at least n should >1 for every sample or you will never know the deviation of your data.
n=3 or n=5 is a normal choice for animal experiment study, such as rats. However if you want to do clinical study, the sample size always should be much larger.
I agree with both the comments above, particularly the statements about significance of biological questions. But, if you do more hybridizations with new arrays of the same type you used before, you should be able to combine the data from all arrays. Of course, higher the sample size the better it is. If there is public data available for the same condition and tissue, I would always look at meta-analysis as an additional option to add value to my findings, even though it is better done before my own experiments. We have always done this (analysis of microarray, EST, RNA-seq and proteomic data before deciding on new experiments) and often avoided the need to do our own experiments. I can give you more inputs for meta analysis, if you need. You can also look into some of our publications.
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology