Forum:How do you make an original contribution to knowledge in applied bioinformatics?

Question

[Edna]

Hello,

I am in a molecular biology PhD program. I am interested in epigenetics and am in discussions to join a developmental epigenetics lab. I have openly discussed with the PI that I would like to choose a computational project, since my goal is a career in bioinformatics. However, she is concerned (understandably) about what exactly this project would look like for someone with no computer science training, and how I would generate enough original knowledge to publish good work and eventually graduate.

I could not really give her an answer. All my experience in the field so far has been more applied bioinformatics (e.g. using existing tools to mine/analyze data), and I'm not sure how feasible it would be for me to catch up on all the computer science required to actually develop new, useful tools.

Is it possible to graduate from a PhD program by just using various data science and statistics methods to test a hypothesis in existing data, or do you really need to be creating tools? I would appreciate any perspective to help me understand my position (and hopefully convince my PI)!

Answer 1

[i.sudbery]

If we modify the question to be "How do we make a contribution in applied molecular biology?" we would think this was a very odd question. If you are an applied bioinformatician, then the contribution you make is the biology you discover using bioinformatics, much the same way as if you are a molecular biologist, the contribution you make is the biology you find with your western blots and qPCRs. You do not need to create a new biochemical assay to make a contribution to knowledge as a molecular biologist, and you don't need to create new algorithms to make a contribution to knowledge as a bioinformatician.

Working as a applied bioinformatician in this situation involves all the same parts of science as anyone else - you must find a research question, form hypotheses, design analyses to test thest hypothesis using either existing or to be collected data and interpret the results.

However, you will need a supervisor that understands both this, and is able to train you to do really good applied bioinformatics, which is quite different from "wet" work - there are far fewer "protocols", and much more thinking on your feet while being aware of common pitfalls. If you go with this supervisor, I strongly recommend you find yourself a co-supervisor to is qualified to supervise bioinformatics properly.

[Edna]

Thank you, this was really helpful and clarifying for me. While I couldn't find anybody comfortable mentoring me in bioinformatics in the biology departments, I found a couple people in the statistics department, and I am going to talk to them next week.

Answer 2

[GenoMax]

Is it possible to graduate from a PhD program by just using various data science and statistics methods to test a hypothesis in existing data

To get clarity about this statement. You don't expect/want to generate any new data of your own. You only want to analyze existing/public data (that you have no control over) to test an as yet undetermined hypothesis.

I have openly discussed with the PI that I would like to choose a computational project, since my goal is a career in bioinformatics.

If your mentor has no experience in computational biology/bioinformatics how can they feel comfortable with this idea. A responsible PI would always be interested in making sure their student succeeds in their project. If they are unable to provide meaningful guidance for subject then this is bound to not work. Perhaps you need to be looking for a different mentor or find a computational expert who will be willing to act as a co-advisor with your PI.

I'm not sure how feasible it would be for me to catch up on all the computer science required to actually develop new, useful tools.

This may only be limited by your ambition. It will take a few additional months/courses in programming but you should be able to acquire enough programming knowledge to put a useful tool together. You will need an appropriate scientific problem/algorithm/logic to address that problem for sure. Your code may not win awards but it will be able to do the job.

[Edna]

Thank you for your response.

To clarify, I am not completely against generating my own data (although I'd really rather avoid lab work if I can), but I suppose I am under the impression that a purely computational PhD would best situate me for a career in bioinformatics. However, I really don't have anyone to ask and I feel quite lost about this.

I agree that a big part of the problem is that the PI is not a perfect fit. I am in a difficult situation—I joined the program based on biological interest, and subsequently discovered my desire to move into bioinformatics. There really is no bioinformatic expertise (or any purely computational lab working on biological questions) at my university. This PI is one of the strongest, since the lab generates huge amounts of sequencing data. I very much appreciate any advice you might have.

[GenoMax]

I am under the impression that a purely computational PhD would best situate me for a career in bioinformatics.

Not quite. Plenty of experimental biologists switch over time and work as bioinformaticians. Their bench expertise is critical in understanding nuances about experimental techniques that generate data and helping biologists with the hypothesis generation/testing.

Answer 3

[noodle]

I'm 10 years in the field of 'molecular biology', which could be considered early-mid career. My opinion is that these days you need proficiency in both, and it makes the most sense to start at the bench. The exception to this is if you're super heavy in math/computer science (for bioinformatics,) or if you're doing something very technical like advanced microscopy (for benchwork).

Is it possible to graduate from a PhD program by just using various data science and statistics methods to test a hypothesis in existing data, or do you really need to be creating tools?

Depends on the program. If you do both benchwork and the bioinformatic analysis, likely yes. If only bioinformatics, maybe...? IMO there are many tools which are just copy-paste of existing tools with minor (insignificant) tweaks, and don't contribute to the field as a whole, but might check some box towards a PhD.