2
$\begingroup$

I am a PhD student working on developing ideas for my dissertation papers. One of my planned papers will be working with some (human) gene expression data.

I have had a class on working with gene expression data, but no practical experience. From what I remember from my project in that class, you do a bunch of pre-checks and QC, then you remove batch effects, then you summarize expression across probes and normalize. I thank that part of my class project took me about three days.

But when I talk to other people about it, they say that cleaning/QCing/normalizing gene expression data is really hard, that it will take forever to get it working, and that if I do this dissertation paper I am going to end up taking extra years to graduate.

Am I missing something here? Why do the people I talk to seem to think this is so hard and time consuming -- is it really that bad? Should I avoid having a dissertation paper that uses gene expression data if I want to graduate on time?

(The data will be from Affymetrix microarrays. There are enough samples to use two 96-well plates, so batch effects will be more of a concern than they were in my class project.)

$\endgroup$
  • 1
    $\begingroup$ Most of the problems with normalizing data are due to a lack of proper design. How is the experiment designed? What do you plan to do if your QC fails? Will it take months to repeat the same experiment? Or it will take 1k$ more to repeat it and 2 weeks? (Who is running with the costs?) What do you expect to learn from this experiment? Aren't there already public datasets that answer your question (I don't know what do you need to do for your dissertation paper)? Why do you do NGS instead of microarrays? (Sorry for the questions, but they will give context about why your idea) $\endgroup$ – llrs Feb 1 '18 at 22:29
  • $\begingroup$ Thanks for commenting! The samples are all collected and frozen, and haven't been run yet. There is a correlation structure in the data (we sampled the same people repeatedly over time) that might complicate things. Because the samples haven't been run yet, we're hoping we can organize our plating so that all samples from the same person end up in the same column on the same plate to minimize batch effects. We are using microarrays for cost reasons. If lots of samples fail QC, the project dies, we won't have money to do it again. $\endgroup$ – bluemouse Feb 1 '18 at 23:19
  • 1
    $\begingroup$ If you put all the samples of an individual on the same plate you will not be able to distinguish between batch effects and individual effects!! That's the thing that's tricky when performing those analysis (Realizing the implications and later problems you will face when analyzing the data). How many samples do you have? Is there any other condition involved (origin of the sample/hospital/date/treatment/age/sex...) that could affect your plates? If it is your first time check your design with more experienced people of your lab before running anything (or spending the samples)! $\endgroup$ – llrs Feb 1 '18 at 23:23
  • $\begingroup$ More info: We'll be looking at within-individual changes over time, and between-individual differences at the same timepoint. $\endgroup$ – bluemouse Feb 1 '18 at 23:23
  • 1
    $\begingroup$ It sounds like you're facing an academic, rather than a technical issue. I suggest you edit your question here to focus only on the technical aspects. Describe the experiment you want to run in detail and ask how long you should expect the QC step to take. The question of whether to do it and whether it should be a part of your dissertation can then be asked on Academia. $\endgroup$ – terdon Feb 2 '18 at 9:56
2
$\begingroup$

I am not sure how much you know about bioinformatics already, can you use R? For a bioinformatician looking at QC for microarrays should not be a big deal, at least for me it would take maybe a day (or two) to get this done. However, if you never used R and want to start from scratch, it depends on how quickly you learn how to deal with arrays and QC. There are many tutorials and review papers about normalization and QC of arrays. Try there and see if you can easily understand what these papers are telling you.

Read the limma user's guide, and try to follow the tutorials therein. If you understand that quickly you can try a real experiment of your own. If you don't understand what's described in there, you might want to look for another project for your thesis.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.