Forum: How To Help Biologists To Ask More Specific Questions?
5
gravatar for jobinv
5.8 years ago by
jobinv1.1k
Bergen, Norway
jobinv1.1k wrote:

I don't mean to sound arrogant with this, and hope that it will not be taken this way. This phenomenon just happens to be a source of some frustration on my behalf, so I'm hoping that some of you who are more experienced in the field than myself perhaps are familiar with the same thing and might have some helpful tips for me.

For the last year, I have been picking up bioinformatics skills while my original professional training is as a medical doctor. I've learnt the basics of how to handle a number of tools and how to interpret the results. I've gotten to the point where a biologist can approach me with their data and their biological question, and I am to some extent able to translate this into a particular bioinformatics/data mining pipeline. This is if they have formulated a biological question in the first place.

Often, though, I've experienced that a biologist brings me the data and places the burden of finding a specific question onto me. This might be formulated as "what can you tell me about this data" or "what are the differences between condition A and B", which are very broad questions indeed.

My ersatz solution has been to for example follow a standard pipeline to differential expression and pathway analysis, followed by returning the list of overrepresented pathways to the biologist, explaining the limitations of bioinformatics pipelines in going further with this on my own, and asking them for their feedback on 1) which particular pathways they are interested in pursuing further, and 2) putting the interpretation of how these pathways are dysregulated, into a biological (and publishable) context. This seems to work quite nicely, and creates quite a dynamic workflow with input from both them and me.

Do others have any further tips in this context?

forum • 1.9k views
ADD COMMENTlink modified 5.8 years ago by KCC3.9k • written 5.8 years ago by jobinv1.1k

Choose better collaborators? Seriously - if they see you as a black box that takes in data and magically produces results, then they aren't doing their part to understand what's going on.

ADD REPLYlink modified 5.8 years ago • written 5.8 years ago by Chris Miller20k

Well, sure, but I would much rather be able to help them understand the proper uses of bioinformatics. If people have tried and tested ways of managing this, I'd be happy to hear tips..

ADD REPLYlink written 5.8 years ago by jobinv1.1k
6
gravatar for KCC
5.8 years ago by
KCC3.9k
Cambridge, MA
KCC3.9k wrote:

If I understand your question correctly, you are wondering how to get biologists to ask specific questions in the language of bioinformatics. However, if they don't know bioinformatics very well, then more than likely they have absolutely no specific bioinformatics questions. (All their specific questions will be biological.)

Speaking based on my own experiences, the question a biologist will typically want answered is, "is there anything you can pull out of my data, that will help me answer my very precise biological question?" This is not a very precise question at all. Quite often they have expended a large amount of effort collecting the data they present to you, and they have some sense that there is valuable insight to be had if the data can be mined the right way.

It's hard to do much beyond the standard analyses unless you understand their biological question in a sufficiently precise way. As you probably know, quite often, subtle features of the question being asked, how the data was collected, the available databases etc, have a huge impact on interpreting the statistics involved in the analysis. You need to be able to hold their problem in your head and translate it into a series of precise bioinformatics questions, then translate the results into biology.

Bioinformatics is often just one tool among many that the biologists are using to get at the heart of their question. I find the biologists I work with are often having to learn to use very complicated and relatively unfamiliar tools, not because they are interested in the tool, but because they want to apply the tool. They are in a constant cycle of learn the tool, use the tool, analyze the results, move on to the next tool. They don't usually have time for a deep understanding of the ways in which the tool can fail to give the right answer.

It seems to me that the proper home for deep understanding of how bioinformatic tools work, nuanced interpretation of their results and translation of those results into biology is bionformatics; and the right people to do this job are bioinformaticians.

So, actually I am skeptical the idea of trying to get biologists to ask specific (bioinformatics) questions. I would suggest that you learn the biology well enough, that you can come up with the precise bioinformatics questions based on your knowledge of what the non-computational biologist is trying to answer. If you can't do this, then I would say the project is probably not worth much of your time. If you have a script that will churn out a standard analysis, then you should do that and let the biologist know that you have just done a basic, standard analysis.

I try to take on projects where I can fully engage with the questions the biologist is asking. This usually means involvement at the authorship level of commitment. Or, I take their data, run a standard analysis with the standard parameters and return the results. I am very selective of projects that fall between these extremes. They can waste a lot of time.

I have definitely been in situations where every time I talk to the biologists, I hear a new detail that I didn't know about before, after I have been working on their project for weeks or even months, and then I have to redo my analysis. Also, I am sure it's painful for the non-computational people when the interpretation of their results keep changing from week to week.

ADD COMMENTlink modified 5.8 years ago • written 5.8 years ago by KCC3.9k

This here is a very good answer. You're completely right, of course - if I get better knowledge about the kind of questions that they in reality want answers for, then I will be better equipped to ask and find out the details that I need to know. Thank you.

ADD REPLYlink written 5.8 years ago by jobinv1.1k
2
gravatar for kanwarjag
5.8 years ago by
kanwarjag950
United States
kanwarjag950 wrote:

There is nothing wrong the way biologists are asking question. No one is going to come and say that can you tell me if A gene is differentially expressed in two conditions and so on so forth. A person (if he/she is no a a sole computer expert) will ask what maximum one can get from the data so the obvious question as they are presented to you. Researchers have ideas and bioinformaticians are going to help them to find those solutions to certain extent and they (biologist) will make story out of it and either publish or write grants after confirming findings in wet lab. I am surrounded by people who think pipelines can do everything; I dont think these pipelines give you a good start and then obviously one has to tweak in lot of parameters/ use several complimentary tools if one has to reach a biologically significant conclusion. So for that one also need to have little more understanding of the focus/ aims of the experiment.

ADD COMMENTlink modified 5.8 years ago • written 5.8 years ago by kanwarjag950
6

You seem to spend the entire post commenting on how I am wrong about demanding more from the biologists, but then in your last sentence you say the same thing as I do. What I want is to get the biologists to explain their focus or aims, rather than blindly expecting bioinformatics to "tell" them what paper to publish.

Basically, I feel that they have an unrealistic expectation of what bioinformatics can do for them. Correct me if I'm wrong, but bioinformatics is not about getting two datasets, running differential expression analysis on these, picking out the genes that come out on the top ten spots and validating them. Nor is it about running a pathway analysis, and assuming that the most statistically significant pathway also has to be the most biologically interesting.

I hope my question is clearer now.

ADD REPLYlink written 5.8 years ago by jobinv1.1k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1779 users visited in the last hour