Bench Scientist To Computational Biologist - Can It Be Done?

Question

[zephryin]

Hello everyone,

This spring I will be a PhD in Human Genetics:-). As the title states, I have a background that is mainly in lab work and bioinformatic work but I want to move to computational biology.

The problem: My thesis research consists of using RNA-seq and methylation bead array analysis in a small sample of human subjects to find differences in a classic disease vs. control experiment. I also did sample prep and verification studies. During my thesis I obtained a certificate in Perl programming and I have taken some coursera courses on Machine Learning, Network Analysis, Data Analysis, etc. I have even applied some of approaches I learned from the coursera course to my thesis research. Unfortunately, I do not have much more "computational" experience beyond this point.

Why? While differences in DNA, expression, methylation are informative, I want to be able to integrate several 'OMIC datasets (i.e.-transcriptomics, methylomics, metabolomics, proteomics, etc. ) to better understand the processes perturbed by the disease. The idea is that in the future, I want to take a blood sample from a patient, look at their "OMIC data" and predict their chances of developing a disease. Or in a more realistic world, I would predict the drug regiment the patient should undergo to keep their disease(s) at bay.

Questions I am asking you: Does this seem crazy to you? When I am reading papers in the field I want to move into, the methods sections generally includes more intensive math than my current skill set allows me to understand. What could I do to realistically develop these skills in an effective (meaning someone will believe my results) manner ? I am currently going through several (use R!) books in an attempt to improve my statistical and data mining skills. Is this sufficient or would you suggest something else? While I am a hard worker, I do not want to do a 9 year post-doc. I'd prefer to keep it to five-- two years to fully develop skills and then three years publishing. Is this realistic? There is plenty of freely available biological data to mine and analyze but I do not want to go off doing this without any guidance. Does anyone know of a forum that discusses data analysis of GEO data sets?

If anyone can give me any type of guidance, that would be awesome! Thanks for reading until the end. -Z

Answer 1

[Damian Kao]

It's possible, but you have to stick with it. I've seen plenty of people who want to make this transition but just couldn't stick with it. There is a difficult initial learning curve if you don't have a programming/IT background. But once you get over that initial hump, everything can be pretty straightforward. You'll find that the community is extremely helpful and there are plenty of resources online.

This initial hump is mostly learning how to use the tools available to you. You have to learn how to "think" like a computer if you really want to utilize these tools well.

My biggest advice is to learn by doing. You don't learn these skills by just reading. You can easily read and understand code samples or algorithms, but you have to actually implement it by writing the code from scratch. You'll find that there is no room for fuzzy thinking in programming. I've met "bioinformatics" undergrads who have taken perl/python/java programming classes but can't write a single line of code. They can probably read code and give you an abstract representation of what the code means, but they can't implement anything.

You have the luxury of running/testing a bunch of different methods/analysis on your data as a bioinformatician; unlike benchwork where you are limited by reagents, samples, time. Take advantage of it by doing as much as you can.

Answer 2

[Dan D]

Yes. I've done it. I basically totally agree with Damian, and like him I want to emphasize that you shouldn't try to learn passively. Learn by doing. There are plenty of sites with challenging problems, including Project Euler and Rosalind. Use these as learning tools. You'll learn by wrestling with problems, occasionally flirting with despair, and then riding the euphoric eureka wave to the answer. It's incredibly challenging and rewarding.

I haven't had any formal education in programming, and I don't feel like I've missed anything. There is an abundance of great books out there for pretty much every language that will provide you with all the foundation you need. The classroom environment is too slow. If you know the science, you'll figure out how to apply the language you choose to your domain. Every language is a toolset, and there are huge amounts of overlap between the various languages. Learn about a few you're interested in to see if it fits with your style of thinking, and then GO! PM me anytime if you want my opinion on something.