Forum: How to switch from bench research to computational biology?
6
gravatar for yuggnurug
4.6 years ago by
yuggnurug70
United States
yuggnurug70 wrote:

I am currently a 3rd year phd student doing my research on angiogenesis/viral immunology. But I am becoming very interested in computational biology and would like to make a switch from bench research.  How can I make that switch after my phd? Or is it even possible to integrate some computational tools in my current research? (I have no idea). What can i do now to find a post-doc position or research scientist position in computational biology after my phd? Should I take computational biology certificate course after my phd? I am currently learning some programming languages such as python but I am very confused what to do next. Any suggestions would be greatly appreciated.

forum career • 6.1k views
ADD COMMENTlink modified 4.5 years ago • written 4.6 years ago by yuggnurug70

What exactly is your line of bench work and what methods do you use?

ADD REPLYlink written 4.6 years ago by pld4.8k
9
gravatar for Michael Dondrup
4.6 years ago by
Bergen, Norway
Michael Dondrup46k wrote:

Giving  my very clear and honest opinion, as always, please don't feel offended.

Now is not the right time to change everything! It is not the right moment to change your field completely. Don't panic. You need to focus finishing your phd. What you can do: try to integrate computational analyses into your thesis, after consultation with your supervisor. You need to focus on finishing what you invested so much time in already or make a drastic decision about your topic. The rest of this answer builds on that you want to finish your thesis.

Computational biology is of course very interesting, but the CS skills it requires are something very different from wet-lab biology skills. 

So when is the right time to go 'computational'. In my opinion computational biology analyses, statistics, and basic programming should become integral part courses in the curriculum of biology, because biology is becoming more and more a computational field.

However, the CS part of it is something completely different from wet-lab and also programming is only a small part of CS. 

You noticed that CS, programming and information processing and biology work well together and are very interesting, so what to do now?

  • Integrate small pieces of computational  analyses into your thesis if still possible after talking to your supervisor, this will increase your credibility for knowing computation
  • If you still need credits for your phd curriculum you can try to take as much computational courses as possible
  • everything else can be learned after your thesis is finished, if you want to do studies of CS/CB even starting from scratch. E.g. a BSc in pure CS or software engineering (another 3 years!) might serve you better than some programming courses. 
  • Take course in programming and get certificates after your thesis is finished, that increases credibility about your programming skills.
  • Try to take a computationally oriented post-doc, you need to be aware of the fact that you will have a certain disadvantage over applicants with long-standing CS/CB background when applying for heavily computational positions, so possibly choose positions that require both.

Hope this helps!

ADD COMMENTlink modified 4.6 years ago • written 4.6 years ago by Michael Dondrup46k
4

Bioinformatics isn't computer science! This drives me nuts!

Knowing how to program is a very useful skill in the field of bioinformatics but it is in no way an absolute requirement, especially if one has zero desire to develop any software. Even when you get into the basic levels of programming in bioinformatics, you don't need proper CS training. Knowing you way around a scripting language or two is more than enough to allow for a ton of basic stuff.

If you've got a few dozen brain cells you can learn to work in a CLI/*nix environment and that's the only initial roadblock before you get to learning how to use the software. The most important part is that you understand what your software is doing, not the mechanics of how it does it.

I'm not saying that a strong CS background doesn't help and that there's an advantage in having a "computational" mindset but I know of plenty of labs that are able to carry out their bioinformatics without a massive compsci background. You don't need to be a mechanical engineer to drive a car.

With that said, Michael makes some good points about how competitive you might be for post-docs. It might be tough to get into a strictly dry lab, you might be interested in labs that utilize both wet and dry. Your wet skills would be useful and you could get exposure to the dry side and see how they can be used to build projects.

ADD REPLYlink written 4.6 years ago by pld4.8k

Honestly, I think the mind set you are describing is a major problem of the field at the moment. There is no equal footing to achieve in computation without proper CS training in algorithms, complexity analysis etc.

Ofc bioinformatics is not equal to CS but it bridges between fields and for me it is likely to me that a formal CS training with some added biology and genomics knowledge can work as well, or complement research, as can a main focus in biology education with added computational training.

"If you've got a few dozen brain cells you can learn to work in a CLI/*nix environment and that's the only initial roadblock before you get to learning how to use the software." 

I have seen a lot of people with biology background with this task, not saying anything about brain cells or death of those.... Then, you are educating scientists, they need to be enabled to understand the inner workings, not of everything but of the most essential pieces of software they are using, they also need to know statistics to design an

I am not the one to decide what bioinformatics is or not, but I foresee that computation and biology will become even more intermingled, in a few years or decades the differentiation between computational biology, bioinformatics and biology will most like disappear, in favor of different research focus or courses. 

ADD REPLYlink modified 4.6 years ago • written 4.6 years ago by Michael Dondrup46k

Part of the lack of equal footing is because the experience required isn't there, part of it is because what experience is assumed to be required for anyone doing bioinformatics. In the end it is all about access to experience

I agree the statistics matter but I really don't see how understanding algorithms at theoretical level or understanding the ins and outs of complexity analysis (and other higher level CS topics) does anything to help me pick the right software. If you can understand an ROC plot and your other typical benchmark figures you're fine.

The details can be important, but for the most part having a qualitative understanding of the algorithm is what matters. All you really need to know is why x is best for your usage case and what drawbacks x has. Most reviews are easy enough to allow the average person to get enough of an understanding to make decisions. Do you really need to understand what a prefix trie is to know how to use BWA or read a review and see what advantages and disadvantages it may have?

I've seen biologists and trained students on CLI, it is daunting but I think there's a strong cultural component. It takes some time to learn and if someone heads in going "I'm not a programmer so this won't make any sense" they're going to have a really hard time. I think it can be easy for beginners to get discouraged because there's a sense elitism from those with more experience (real and/or perceived) that leads them to believe they're not able to perform on the same level because they're not smart enough and not simply because they're beginners.

The major advantage of having CS experience is that you've already spent time working in some of the same environments and you've had time to develop (and probably started with) the required mindset. It ends up being a head start over a traditionally trained wet biologist.

I agree that they're merging and I think it is a good thing. We're all biologists in the end.

ADD REPLYlink written 4.6 years ago by pld4.8k
2

Just to chime in some of my own observations after teaching computational courses for life scientists with little or no background in computation.

I have come to believe that computational thinking is similar to arts and other hard to quantify talents. Education can help develop those talents and identify those talents in those that did know had them.  What it cannot do is guarantee that these skills can be developed to a given level.  Many students surprise me in how easily they make the transition, others cant make any progress whatsoever and not for the lack of trying.

It is a bit like singing, drawing, playing a musical instrument, playing basketball etc. It absolutely not like "reading" - I think everyone can learn to read well. Consequently the concept of general "computing literacy" is misguided. Not everyone can learn how to use computational tools. Many people can but those that cannot - nothing wrong with that either. There are great many jobs out there that do not need it. 

Here is a neat post that I found most illuminating, it makes the point that Modeling and not Coding is what people need to learn

http://www.chris-granger.com/2015/01/26/coding-is-not-the-new-literacy/
 

ADD REPLYlink modified 4.6 years ago • written 4.6 years ago by Istvan Albert ♦♦ 81k

This is a great point, there's a general mindset that is needed to totally excel, and everyone will have some shade of a computational mindset ranging from "none" to "genius". Just like everything else, given some desired level of skill, you have to have enough of a inherent talent and the interest.

The problem that I see is people confusing the need for a more "computational" view on things with "I have to be a programmer". I also think people don't realize that building your computational skills up to a point based on their desire/skill/circumstance is perfectly fine. You don't have to completely and totally master computer science and then master all aspects of bioinformatics.

Maybe you just want to be able to use a few more tools than BLAST or want to do some of the analysis of expression or SNP data. Maybe you just want to learn enough about the technologies and practices so you can find/build dry components that will best help your research. You'll never do the work yourself, but you'll know enough to steer things. Maybe you are really bored of wet work and want to change completely.

 

ADD REPLYlink written 4.6 years ago by pld4.8k
1

Thanks for the honest opinion Micheal. Just a small clarification, it seems you are answering the question "Should I change my PhD/career now", while the OP is asking "How can I make that switch after my phd". (Apart from this I agree with you).

ADD REPLYlink written 4.6 years ago by dariober10k
1

Yes, you are right, op asked about after the phd. However, it looks like that this shift in interest has had consequences already, like starting to learn programming and being confused. Maybe I was reading too much between the lines?

ADD REPLYlink written 4.6 years ago by Michael Dondrup46k
4
gravatar for Emily_Ensembl
4.6 years ago by
Emily_Ensembl19k
EMBL-EBI
Emily_Ensembl19k wrote:

Is there a question that you can ask, that is relevant to your thesis, that you could do computationally? You could add a whole chapter to your thesis without doing a single wet experiment, just slot it in while you're letting something incubate/react/run etc. That's enough to give you an idea if that's something you could enjoy, and to give you something to talk about in interviews and things.

A thing to note is that a lot of experimental post-docs involve a lot of bioinformatics these days. If you've got a bit of bioinformatics under your belt, then you're more likely to get the post-docs where you're doing the wet science then analysing the data yourself and you can build from there. I have friends who started out on a wet-lab project and by the end of it were entirely computational.

ADD COMMENTlink written 4.6 years ago by Emily_Ensembl19k

I think this is a great approach. This also give OP a chance to test the water before looking to make a complete commitment.

OP could also explore if there are bioinformatic labs that would want to collaborate.

ADD REPLYlink written 4.6 years ago by pld4.8k
2
gravatar for A. Domingues
4.6 years ago by
A. Domingues2.1k
Dresden, Germany
A. Domingues2.1k wrote:

I would like to add my personal perspective which is not too dissimilar from your experience. Some background:

I am a trained wet-lab biologist and used live-cell imaging, molecular biology and cell culture in the context of neurobiology up to my 1st posdoc. Then, due to some results during my PhD, I decided to make a change to study regulation of gene expression in a cell biology lab. The project was mostly wet-lab, but it already had the promise of large datasets and the opportunity to learn how to analyse that data. For a number of reasons, at some point I was mostly sitting on the computer mapping NGS data and plotting in R. And I liked it a lot. So much so, that my supervisor supported me in taking some courses and leaving the wet-lab for others while I did the analysis. Fast forward a couple of years and now I do bioinformatics.

Bioinformatics or computational biology? Even though there is not established distinction between what constitutes What Is Bioinformatics?, in my mind one is more about data analysis and the other more about developing algorithms. The latter has stronger CS component in my view, and the two often go hand-in-hand of course. So do you want to develop algorithms or are you more interested in teasing out answers from data asap? Because, IMO, the path is different. Also bear in mind that bioinformatics and computational biology are very broad churches. You could be doing analysis NGS data from *-seq, doing some protein network analysis, running simulations of evolution, or even developing algorithms for data extraction from images. Even though Biostar focus more on the gen* side, there are many more things out there.

Now, I will answer your questions assuming that you are more interested in bioinformatics.

 

How can I make that switch after my phd?

For your post-doc choose a lab with a project that you like (very important), and that has both components (wet+dry), or is looking for someone willing to learn the bioinformatics bit. There are more projects/labs like that than you think, specially if you are willing to work with NGS. It is important that if you are going to learn "on the job", as a lot of us do, there will be people next to you to help you in the beginning. This could be a lab colleague or a collaborator. Failing that, the lab should allow you take the courses needed to learn.

In the meantime, you can start by taking a few courses even during your PhD to learn the basics. Let's say how to work with the command-line or programming in phyton/perl/java and R. Your university might have some courses available that you can take, or you can attend a (free) software-carpentry workshop. Learning how to use the command-line is extremely important and useful regardless of what you do. There are also two books that I earthly recommend (getting one of them his enough though - unless there is grant money available):    

- Practical Computing for Biologists, a bit older but still current.

Bioinformatics Data Skills - O'Reilly Media, as of writing only available as ebook, but well worth it.

Both are written from a very practical perspective and will get you started almost immediately. The 1st also contains chapters on python, and the second spends more time evangelizing on the the power of reproducible research. They sort of complement each other, but the first is better to get things done from page 1 and the second is better if you are serious about bioinformatics in the long run. Both will teach you how to use the command-line and it will save you time as soon as you start!

If you don't use Linux or a MacOS, set-up a computer with Linux (Ubuntu*) dual-boot or a virtualBox.  Also pay more attention to statistics. Take a course at Uni if possible, or do one online.

 

Or is it even possible to integrate some computational tools in my current research? (I have no idea).

Without knowing *exactly* what your project is I can't really answer, but chances are that you can. For e.g. during my PhD, which involved cloning receptors, I had to do homology searches, domain predictions, and analysis of imaging data. I did not knew it at the time, but learning some bioinformatics (and R) would have helped a lot.

 

What can i do now to find a post-doc position or research scientist position in computational biology after my phd?

To be honest,for you to get a completely dry lab position after your PhD you would have to learn quite a bit in the meantime. That said, if you can learn enough to write a chapter your thesis from analysis that you performed, it will go a long way. Remember that most labs/positions will know that a pos-doc does not know everything, but brings something. So say, you go to a computational lab that works in your current field, but from a different perspective, you will bring them biological know-how, and in turn they will teach you the computational side specially if you already shown an active interest in learning.

 

Should I take computational biology certificate course after my phd?

When I was planning for the switch I considered taking a step back and doing an MSc in bioinformatics. At the time I asked a few bionformaticians that knew what I was doing and they suggested it was a bit of a waste of time. They believed that an MSc would not teach me enough to be worth the time - besides I am more interested in answering question than developing algorithms/software. In subsequent job applications/interviews a formal education in CB was not never an issue. Showing that I could do the analysis was more relevant. That said, depending on what your goal is, some formal education might be worth it. If I could go back I would have probably taken a class in CS and statistics while doing my PhD.

 

Given the lack of detail in your post it hard to give more specific advice, but I hope this helps. If there something I learned from working with people that came to science at different stages, and changed fields relatively late in the career, is that it is never too late. 

 

*other flavours are available

 

[edited for grammar]

ADD COMMENTlink modified 4.6 years ago • written 4.6 years ago by A. Domingues2.1k
1
gravatar for Jean-Karim Heriche
4.6 years ago by
EMBL Heidelberg, Germany
Jean-Karim Heriche21k wrote:

I agree with what Michael Dondrup wrote above. First, finish your PhD and don't try to go computational just for the sake of it. If the project/question doesn't need computational answers don't try to go down this road. You'll just lower the quality of your PhD. Keep focused on the question(s) at hand and on finding and applying the most suitable answers. The best way to switch fields is to start a postdoc in the new field. If you have never programmed before, have no notions of statistics then you may have a hard time but you can prepare for it by taking some classes.  Another way to prepare may be to take on a small programming project on the side (showing some code is better than a certificate from my point of view) but don't let this eat into your PhD. You might also find a postdoc supervisor willing to train you from the ground up.

ADD COMMENTlink written 4.6 years ago by Jean-Karim Heriche21k
1
gravatar for TriS
4.6 years ago by
TriS3.9k
United States, Buffalo
TriS3.9k wrote:

first, I totally reinforce the idea of finishing your PhD as priority. then, I add that the computational component is much better if done gradually. don't panic about having to know everything in the shortest time possible, give yourself time to familiarize not only with "how" to do bioinformatics stuff but also "why" you might need to do them and which questions you can answer with it.

furthemore, as mentioned above, completely "dry" labs might be hard to get in but do not underestimate the importance of being able to bridge between wet and dry lab. I talk to many bioinformaticians/former engineers that know next to zero about what the data actually mean in a biological context and, on the other hand, I see tons of biologist that get a headache by spelling the word "algorithm" and think that "time complexity" is equal to "time travel". it appears to me that you have the chance to get into a unique niche that will allow you to learn the skills from both side of the coin and work in both environments by allowing ppl to understand each other. it will take time, effort, energy (and lots of coffee) but if you get your mind into it, as I'm sure you will from we can see from what you wrote, you will be able to make yourself indispensable for the lab(s) that will hire you.

at the end, finish your PhD, start getting your feet wet with bioinformatics and, most importantly, keep learning. 

ADD COMMENTlink modified 4.6 years ago • written 4.6 years ago by TriS3.9k
0
gravatar for yuggnurug
4.5 years ago by
yuggnurug70
United States
yuggnurug70 wrote:

Thank you all for your comments and I apologize for delaying my response. Currently I am working on elucidating mechanisms for corneal angiogenesis following HSV-1 infection using mouse model. To be specific I have done microarray gene analysis and protein suspension arrays using commercially available kits to nail down growth factors that could be responsible. Experiments involving neutralization of some of those growth factors are on-going but this is my dissertation project on a nutshell. I am curious to know if anybody has any suggestions as to how I can incorporate some computational approaches into this project. 

I appreciate your comments.

ADD COMMENTlink written 4.5 years ago by yuggnurug70

If you've already carried out the analysis of various arrays, it sounds like you might have a better start than you think.

I think the best way to see what else you can do is identify specific problems to approach with bioinformatics/computational biology. There's plenty of places to go, and many of them might not suit your goals.

Maybe use computational approaches to find inhibitors of your targets, or use other approaches to identify new targets that may be more drugable, say factors that are up/down stream of your current targets that might not have the drawbacks of targeting a growth factor. You could use computational approaches to see if the human equivalents of your mouse targets might respond the same to your drugs or what differences may exist in those pathways.

I'd be really to see if you have any insight into the differences between mouse and human VEGF and MMP (especially MMP9) signalling.

ADD REPLYlink written 4.5 years ago by pld4.8k
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: 1600 users visited in the last hour