I am currently collecting biological specimen for microarray study. However, the availability of specimen is very scarce and I have to wait sometimes 2-3 months for new sample to be available. I am afraid my sample size will not be enough for my research. But my colleague said it's okay for sample size to be small for microarray study. Is this true? Help me

This is fine, there is plenty of biomedical research with just a few specimens. Maybe check with your supervisor?

Oh thanks. I am just afraid my sample size will be too small.

It's 'ok' in the sense that if your work is carried out well then it will probably be sufficient for a PhD.

It may, however, limit the ability of your work to provide real insights into a phenomenon. This in turn may affect the likelihood of publication.

I would advise consulting a (bio)statistician for advice on this. Sample size is partly contingent on the nature of your hypotheses/studies.

You don't mention what kind of samples you are seeking, but it may be that it would be possible to obtain a larger number through collaboration. If, for example, you are collecting patient specimens from a single hospital, you could approach potential collaborators at other hospitals to increase your collection rate. But to do this you should firstly check with your supervisor and secondly ensure that doing so is within any study/ethics approvals you have obtained.

Sample I am collecting is tumor tissue. It's very hard to obtain this tissue in my country since the prevalence is low.

In which case it might be worth seeing whether you can make overseas contacts who will let you have access to samples. Or see if they can be purchased from cell/tissue banks. Of course with permission from your supervisor.

It may feel intimidating to approach collaborators but it's actually an important skill for an academic career. There's a knack to it though, so seek advice.

Whether it is worth it kind of depends on whether you want this work to provide meaning just for the thesis, or have a wider impact. A study based on 10-15 samples has limited interpretability.

As I said, you ought to discuss this with a biostatistician. Whether or not 20 samples is adequate depends on the hypotheses you're testing.

Well, it could also depend on whether you're doing hypothesis testing or hypothesis generation.Different study designs have different requirements with regard to sample size. Do you also have control subjects for your study?

If you google "how to determine adequate sample size for micro array" there are some helpful looking papers. A biostatistician with a lab focus will be more helpful than an epidemiologist in this scenario.

I used to work with microarrays so you have my sympathies on the cost aspect! Realistically it is very difficult for a PhD student to do 'enough' experiments on the funding available.

Whether or not you get to improve your sample size, being aware of the limitations the size places on the interpretations is very important for when you write up (whether for papers or your thesis). No statistically significant findings may be due to under-powering the study. In contrast, detection of homogeneity or heterogeneity between your samples may be difficult to interpret without controls - how do you know that similarities/differences are related to the tumour (or tumour subtypes) and don't simply represent naturally occurring homo/heterogeneity across the general population?

Sorry if I am giving you too much to think about. I am familiar with your conundrum as I used to do similar work when I was a lab scientist and it frustrated me the way that I was told very small sample sizes were ok when I could see potential flaws with the approach. Either way, your thesis will be fine, but it may affect publications leading from it.

In your case a control could possibly be comparing a normal (non-cancerous) sample to another normal sample as a test for just regular variation within the same organ.

I feel your pain regarding the collection of samples. I have a similar (sequencing) project where I need to collect tumor+normal samples from patients meeting certain criteria, which is really slow. I will also not have as many patients as I'd want to but I'm "only" an MSc student, so for me it's okay. No pressure to publish, my project has a fixed duration and my supervisor can build on it either way.

However, because I am also not experienced enough in biostatistics, I just contacted someone from overseas who is. When researching I found a lot of papers from two researchers which were directly related to my problem, so I just sent them a friendly email explaining what I do and if they could help me. It has given me a lot of insight into what amount to data I (or my supervisor) should shoot for, and how to proceed with it.
Maybe you could find someone who could discuss your statistical issues with you too?

What are your supervisor's expectations of how you should get the samples by the way? If he does not want collaborators from elsewhere then I assume he wants to get samples from a local hospital?
Also, how did you phrase it when asking about collaboration? I mean, there is also a difference between collaboration and requesting tissue. I am also using tissues I got from other national hospitals but I am not actually collaborating with them (in that they will not have any co-authorship whatsoever once there is a publication). Maybe there was a bit of a misunderstanding and your supervisor just didn't want to share the glory?