I want to replicate my Microarray (methylation) results in a larger population using a different technique. Should the population chosen for Microarray be the same for the replication population? As in, the inclusion and exclusion criteria kept can be changed or should I adhere to the same for replicating my results ?
Dear Shyamashree,
Meta-analyses practiced on microarray studies related with a same pathology usually yield very different results even when the same analytical platform were use. The picture is worsen, if in addition, different analytical platfotms are compared. So, microarrays are far to produce comparable results due to multiple factors, such as analytical variability, biological variablity at the sampled individues, and differences due to data processing. Asking your specific question, if you introduce different criteria of inclusión (or exclusion) you are introducing even more uncontrolled variability in the data to be analyzed. Ideally, you have to control your sources of variability. A principle of such control technique is to use the same experimental units subjected to several methods processing. In this way, each subject can be treated as a block in wich all the treatments are represented allowing to be compared independently of the differences among the individues.
Thank you Sir. So, basically you mean that to reduce the variability, best is ti keep both the population as similar as possible. Also, considering the fact that I will be working with atherosclerosis which is a multifactorial disease I should keep both the population homogenous and similar. eg. if I take non-diabetic population both microarray and replication polultion needs to be non-diabetic?
Please correct me if I am wrong
Yes, the same type of populations should be used for both methods. If you wish to include a diabetic population, you should also include as additional covariate in all the methods. So, you could manage the assay as a full factorial design. If you not go in this way, you will "confound" the statiscal effects of the variables, and you will attribute variations to causes that are not the true causes.
A haphazard sample selection is not equivalent to random sampling. You need to know what are you analyzing.
I agree with the other answers. You should use the same inclusion and exclusion criteria for your second population in order to keep variability as low as possible.
Hi MS Banerjee,
I understand that you would like to validade the results got by array in real time pcr/
if so, the sensibility is quite different so you should be very criterious to keep the same criteria to select your casuist, I suggest also to make a sub group by age, gender ethinic origem, because metilation can be modify by menopause, hormones levels, and others ambiental sources of interference. I understand that you also should be carefull in to add hypertension, hypotiroidism, dislipidemia, metabolic syndrome and chronic disease in your criteria of exclusion /inclusion
Dear Shemashree,
I agree with Mario Hirata. However, a form to simplify your experimental design should be first to define if your interest is to select the best analytical method, or to investigate a biological problema or both aspects.
If your motivation is the first case, you should compare the methods using the same experimental untis, but selecting a unique biological conditions (e.g., control treatment). Once you have selected the method according to the best analytical performance, you can made a second round of experiments using several biological conditions.
Stratification according to hormonal stage or other metabolic diseases is also posible and desirable, but you should take in consideration that higher the number of variables, higher the number of microarrays needed to run, and higher the complexity of the data processing. If the number of biological conditions were higher tan two, "incomplete factorial" design could be recomendable.
- Badges
- Science method