How Important Is Bioinformatics And How To Make A Case For It?

Question

[Allpowerde]

Bioinformatics is at the heart of modern biological research, especially so in whole genome studies and second generation sequencing (2GS) data analysis.

However, when looking at the methods section of a 2GS paper the space dedicated to describing the bioinformatics is severely disproportional to the amount of work that was actually taken up doing the bioinformatics. It is therefore easy to misjudge the contribution from bioinformatics to a biological papers.

This misconception is also fulled by omnipresent commercial companies claiming that a full bioinformatics analysis of 2GS data can be done by a single easy to use program at the push of a button, which would indeed make the bioinformatics analysis a simple service task.

While bioinformaticians are ready to make circumstantial cases of how bioinformatics is important to produce data of higher quality or rescue noisy data, I could not find a substantial body of evidence showing how bioinformatics are increasingly contributing to biological science as researchers not service personnel. (E.g. growing number of bioinformaticians as first/last authors in nature/science/cell papers).

Can anyone suggest how to convincingly make a case that bioinformatics is not just a service but a proper research area that is significantly contributing to modern biology?

[Chris Evelo]

Well the thing is that the biologists I work with try to solve problems like diabetes, heart failure and cancer. I am happy to contribute to that in whatever helps to solve those problems best. If somebody needs to play the back end of the horse in this play it just needs to be done. This is not to say that I do not think we cannot do very interesting stuff and have a lot of fun doing so. But I absolutely think that bioinformatics now and then should just provide the service needed. (Pfui, am I lucky that comments cannot be downvoted ;-) )

[User 3869]

I am a PhD student of CS. I am working on bioinformatics. I have the same feeling. I were told repeatedly by biologists that "bioinformatics is to serve biologists". Honestly, I totally can not understand the reason of that they thought CS people would be attracted to be servants of biologists. I really appreciate if anyone could give me a good reason.

[Boboppie]

My personal opinion is: the reason we cs people can't be independently doing bioinformatics is that we need biologists make data rational as so called "tell a biological story", proper biology experience and knowledge are what we lack of. I think computational biologists could be the leading force of the future of biology. However, it doesn't mean cs people can't share the cake, one way is to learn more biology and practice the biology way of thinking, thus to contribute to the "story telling", Prof. Lawrence Hunter shows us a good example, and it could be more fun.

[Allpowerde]

I am in a position many post-docs might find themselves in over the next years: many purely biological institutes are setting up a sequencing core facility and because the contribution and difficulty of the bioinformatics involved has been portrait incorrectly by papers and sequencing companies, the resource allocation is insufficient. In my cases: one postdoc to "make sense of the sequencing data", i.e. production informatics, data analysis and hypothesis generation. The question is hence, how to explain that genomics research requires more than just eyeballing the data or running one script?

[Qdjm]

@allPowerde - a core facility is, unfortunately, not a great place to try to do research. Try to find a read-mapping pipeline that satisfies most users' needs and train them on it -- so your basic service is showing people how to analyze their own data. Then spend the rest of your time focusing on projects that you can make a real impact and that are run by people who appreciate your contribution.

Answer 1

[David Quigley]

There are several issues here. It partly hinges on what you mean by "full analysis".

First: the amount of space devoted to describing the result is rarely proportionate to the amount of work that went into the result. Bioinformatics doesn't have a special claim to fame there. Panel 3B of a cell biology paper that has six figures? The one with twelve lanes of western blots? Getting that figure took a post-doc and a tech nine months. To describe the heroic efforts to find antibodies that work, perfect the tissue extraction protocols, figure out the correct time points, etc, you see one sentence. If you want the real story you have to start buying beer for the lab.

That said, I break analytical work into two phases. The first phase is engineering; it requires experience and good judgement. That's the mapping in NGS, the microarray normalization, the genotyping, what have you. The second phase is the biological interpretation. The first part can be done by someone who doesn't know a lot about the purpose of the experiment, though it's better if they do. If a PI comes to you with raw microarrays and asks for normalized data, and that's where it ends, that's a service. It's an expert service, a skilled service. If you contribute to experimental design, if you understand the biological question, if your analysis provides insight beyond a list of differentially expressed genes or a heat map, then you're doing biology. If it's your lab, or you did most of the work, then usually you show up as last or first author. If your work is in developing techniques, but your data are simulated, then that's a very valuable contribution but those papers will usually go into specialty journals for people inherently interested in the techniques. Note however that Nature Genetics now has an "Analysis" paper track.

I think a focus on Nature and Cell and Science as the litmus test of whether research ideas are significantly contributing is too narrow. PLoS Biology, Genome Research, Cancer Cell, PNAS, they don't count? In any case, to publish a story in a biology journal, you need to be telling a biology story. Ultimately that requires that you are doing not just bioinformatics, but genetics, or drug design, or physiology, etc.

Answer 2

[Qdjm]

You need to look harder, there are many examples of bioinformaticians (let's call them computational biologists) who publish high impact papers:

• Manolis Kellis (MIT), Christopher Burge (MIT), and Gil Bejerano (Stanford) are examples of computational biologists who have had "dry lab" papers published in high impact journals.

• Eran Segal (Weizmann) or, more recently, Ziv Bar-Joseph (CMU) are examples of computational biologists who are senior authors on high impact papers that combine wet and dry lab research.

• The EBI is full of "dry lab" researchers publishing high impact papers.

(Some of these researchers now also have wet labs.)

But as David alluded to, there are technicians and there are researchers. Put simply, researchers run the projects, technicians do what they are told to do. If you want first authorship, you have to take responsibility for the whole project. This means that you have to be the one telling the wet lab technicians what to do and play the largest role (with the senior author) in troubleshooting and interpreting the data.

That said, if you feel your contribution isn't being adequately acknowledged -- find new collaborators.

Answer 3

[Egon Willighagen]

There is a bit of a confounding problem here. For bioinformatics to truly succeed, you need all the data you can get. Biologists do not provide that. Just try to look up the raw data for any biology paper. You will only get to see whatever supports to story. That makes it hard for bioinformatics to prove itself: it's success is just rate limited by lack of Open Data in the life sciences.

Now, what you should think of instead, when judging bioinformatics, is how many Nature / Science papers are already using bioinformatics. It's a bit like those cartoons where they depict your living room as if there was no plastic. Now, visualize for yourself what the latest Nature issue would look like without bioinformatics.

Answer 4

[Martin A Hansen]

Is anyone aware of any open notebook science projects with a major bioinformatic part? In such a project one might get a true view on where the efforts were - and this could be used to "make the case".

http://en.wikipedia.org/wiki/Open_Notebook_Science

Answer 5

[Paul]

I can only illustrate my point using the example of signal sequences.

Signal sequences are N-terminal extensions that direct proteins destined for protein translocation.

The pioneric work of Gunnar von Heijne (started in the eighties of the last century) has attributed significantly to the understanding of signal sequences (general structure, differences of signal sequences in different translocation systems etc.).

His work was initially purely theoretical, still his work (and of co-workers) has helped the field of protein translocation significantly further.

So yes bioinformatics is a proper research area and can attibute to our understanding in life sciences.