The more differences that exist between your cases and controls, the more confounding factors you have to deal with in your analysis. Ideally, they should be as similar as possible, save for the phenotype that classifies them either as a "case" or a "control".

Good one Sangeet. It depends on how you are deriving the problem.

If in case you wish to compare with and without the preamplification step from say FFPE tissues, you better have a subset of samples that would allow you to compare compare genotypes taken from blood and FFPE tissue samples

Otherwise if you are restricting your studies with identification of say one allele/one genes with CNV analyses, stick to one of those.

I hope this helps.

The more differences that exist between your cases and controls, the more confounding factors you have to deal with in your analysis. Ideally, they should be as similar as possible, save for the phenotype that classifies them either as a "case" or a "control".

All above are valid points. But there is a need to pay more attention to the goal of the experiment here, rather than the traditional case-control situations. It may be important to consider the following in your experimental design:

1. In the end, you want to establish SNPs that are different in the test (disease) DNA as compared to most of the normal individuals' genome.

2. I do not think the DNA from body-tissues (normally available as FFPE samples) will have variations introduced as SNPs due to the processing (you may miss some SNPs though).

3. Hence, it is important to ensure that the differences are with respect to the fresh normal tissues (rather than stored tissues).

But also note that the tissues are likely to have differential mutation rates (e.g., skin is exposed more to environmental factors than brain tissue, and hence there can be non-hereditary mutations in the skin). So, in that way, comparing a blood tissue (fresh preferred) vs. other tissue (FFPE in your case) may be more carefully decided - even though the genomic DNA is almost identical across tissues of an individual.

I agree with Michael Wendl.

When using blood you will get germline mutations, when using tumor samples embedded in FFPE you will get somatic mutations. And how do you want to compare somatic mutations in one case with germ line mutations in the other case? Doesn't make sense.

If you can obtain from patients both FFPE embedded tissue and blood you could check if there are differences in between germ line and somatic mutations in order to find out which relevance a SNP has.

Daniela is absolutely right.

Thank you all for the valuable suggestions and inputs.

Daniela and Jorge, correct me if i am wrong: Blood is not the only true-representative of 'germline' mutations. Also, all other tissues do not have identical somatic mutations. Blood cells are expected to have lesser chances of somatic mutations compared to a few of the 'peripheral' tissues. Blood is usually used as a measure of germline mutations, because it is easily available.

For the current discussion initiated by Sangeet, I still think that using blood is OK because,

a) the best control (the same tissue as the stored sample, but normal one) may not be available;

b) for diagnostics, it is imperative that the SNPs identified by the study as 'associated with a test condition (across multiple test samples)' are different from the SNPs in the 'normal' population (which can be observed using blood tissue).

Yaa. Two conditions mentiond by Dr. kshitish Acharya, is exactly the challenge, I am facing right now. I want to study a specific SNP in NSCLC FFPE tissue procured from Indian patients but I am unable to procure normal lung tissue. The samples I get in my lab, are mainly FNAC sample. I do not get normal lung tissue. That was the main reason, I thought about using normal blood sample as control. I was not sure about it, if using blood samples as control against FFPE tissue would be a good idea.

Sangeet, I would strongly suggest you to ask if appart from your FFPE tissue additional blood is available for your patients - and analyse both blood and FFPE tissue concerning your SNPs. Only if you make this comparison you can be sure if the SNPs are the same or not...

Dear Dr. Daniela Bruennert, In my lab FFPE samples do not come along with the additional blood samples. That is the biggest challenge for me. By the way thank you for your inputs.

Given the situation with long-stored 'test' samples, I think if a study detects SNPs that are 'present in DNA of most or all test samples' but 'absent in all control (blood) DNA' - it will be a valuable outcome.

But I agree with Daniela in that we should preferably analyze both FFPE and blood tissues from test conditions.

May be a few fresh biopsied tissue and blood samples from new patient donors can be obtained in near future to confirm if the identified SNPs occur in DNA from blood as well.

Thank you Dr. Acharya and Daniela. Thats great. So, I can do this work. First i will study the archived samples and control blood samples and I will look for patients with lung biopsy and who can provide blood sample as well. For the later condition, what would be the preferable sample size(10, 50, 100 samples)?

For this kind of study there is some kind of statistics telling that you need a minimum of 80 vs 80 samples. I know from SNP studies that it's possible to publish data with approx. 50 patients (and 100 patients), but for our ovarian cancer study we trying to publish at the moment we were using approx. 250 patients and 400 healthy controls, using DNA extracted from EDTA blood. If you want to introduce SNPs as molecular markers it is always better to go for a big patients cohort (something like 1000 or even more patients), but initially I would start with 10-20 samples to see if the SNP has any drastic difference. For me, a good molecular marker is a SNP where one allele can found in something like 80-90% of patients whereas the other allele can be found in 80-90% of control group.If you have a big patients cohort (several hundred or thousands of patients) you will obtain fine differences (like e.g. 22.4 vs 26.1%)...

If you are using a already known SNP for your study and want to know its prevalence using a case-control study, then you have to calculate your sample size based on previous studies which yielded significant results with that SNP. But, make sure that you cases and controls are matched for age, gender, smoking status (as smoking is the major risk factor for lung cancer) and, if possible, ethnicity. Controls should be free from any other inflammatory disorders and should be disease free, if possible. For normal control samples, you can use blood DNA or even buccal swab which can provide you epithelial cells for DNA isolation. Taking buccal swab is non-invasive and easy method for collecting cells for DNA isolation.

All the best.