Today, I am trying to do something different. I will offer advice on something I have never achieved myself: publishing bioinformatics-centric papers in high profile journals. The journals I have in mind are not Bioinformatics, BMC Bioinformatics, or PLoS Computational Biology; those do have a decent impact factor (ISI) but they are supposed to publish bioinformatics papers. I am talking about Nature, Science and Cell – journals that claim to publish the most exciting stuff in the biosciences. While this does not formally exclude bioinformatics papers, you rarely see them published there. A small number of examples shows that it is possible, though, and I am trying to analyze what it takes to publish your next bioinformatics paper in the Premier League of bioscience journals (depending on your country and interest, insert your favorite equivalent here).
As I said, I never managed this feat myself. This is not the fault of the journals, though, as I hardly ever considered submitting my (bioinformatics) manuscripts there. The main reason is that my bioinformatical interest (and possibly my capabilities) are not compatible with the guidelines that I will outline below. Whenever I see a bioinformatics paper in one of the real high-impact journals, I experience one of the two reactions: i) Wow, what a great paper, I wish I had done this myself! or ii) Whoa, what did they do to get this junk published by that journal??
After analyzing many of the type i and ii cases (more of the latter), I collected the following guidelines:
1) Your work should come to a conclusion that cannot be verified experimentally.
This sounds paradoxical enough. Shouldn’t bioinformatics research strive to make predictions, or to give explanations for observed phenomena? And isn’t a prediction worthless if it cannot be verified? Why would anybody prefer one explanation over another, if they are not testable? I would be the first one to subscribe to this point of view. However, if you want to publish your work in high places, and you conclude something amenable to experimental verification, editors and reviewers will jump at you demanding that you also provide the verification experiments.
As you work in bioinformatics, you typically cannot do the experiments on your own. You will have to collaborate with experimentalists, which immediately leads to a scenario that I have covered in a recent post. Unless you are extremely lucky (or famous), the best you can expect from this collaboration is a “type 2b paper“, which is neither a bioinformatics paper, nor is it yours anymore. If your goal was to publish a bioinformatics paper in Nature or Science, you have lost. Better stick to topics that cannot be tested – some examples can be found at the bottom.
2) Present large-scale work (one ome is not enough)
Nowadays, working on a single protein, gene, modification or other phenomenon is clearly not sufficient. Whatever you are doing, do it on the entire genome, proteome, or whatever ome is relevant. Even better: integrate several omes.
3) Provide bioinformatics support for the latest hype
This might not be everyone’s cup of tea, but is always a good recipe for getting a Nature or Science paper. Go find a new hype (in statu nascendi) and show by bioinformatics that this phenomenon is even more pervasive/important than everybody thinks. If someone just discovered alternative splicing, show that every gene is alternatively spliced. If they discover miRNAs, show that the entire intergenic region consists of miRNAs. Alternatively, show that each miRNA can regulate every gene in the genome. Something along these lines, anyway.
4) Results trump methods
At the target journals, the section editors and probably also the reviewers will probably be biologists and they will judge the submission on the basis of biological criteria. It seems like the hurdles for method-centric papers are be extremely high. In fact, I can only remember a single bioinformatics method published in Science: a paper on a Gibbs-sampling strategy for multiple alignment by Chip Lawrence, published in 1993. On the other hand, there are quite a few bioinformatics papers focusing on interesting results.
5) Standard recommendations apply
The previous guidelines specifically address bioinformatics papers. Obviously, all the other rules for publishing in Nature or Science also apply. It is clearly not sufficient to submit any large-scale analysis that produces non-verifiable results, and enrich it with the latest buzzwords. Your work either has to be extremely good, or you have to claim something quite extraordinary. If you choose the latter route, you don’t expect problems with the editor; the trick is to get past the reviewers. Like with all other submissions to high-profile journals, it helps to i) work at the right place, ii) know the right people, iii) have a history of previous high-profile papers.
Having formulated these five golden rules, how about some suggestions for work that stands a chance of being published in high-profile journals.
- As said previously, methods are problematic. If you want to try anyway, maybe you should create an algorithm that successfully calculates 3D structure from the sequence – based on first principles.
- You could also find a method to reliably predict which alternative splice form is made in which particular cell type.
- You could find out what the conserved regions within introns and between genes are actually doing. Not one region, all of them!
- Maybe you could find the gene for the elusive reverse translatase – but wait, this would require some experimental verification.
- A more conventional approach would be to create an enormous network encompassing genomics, proteomics, transcriptomics, metabolomics, interactomics, and at least two more omics datasets, all compiled from at least 10 model organisms. (This network graph should be aesthetically pleasing – they will need it for the title page). As a result, you should find at least five properties that follow a power-law.
- When looking for other non-verifiable areas, it is always a good idea to turn to evolution. As a suggestion, you could make a strong case that the last common ancestor of all extant organisms had exactly 42 miRNAs. Or protein folds. Or whatever.
- Undoubtedly, the most exciting papers of all deal with the metaphysical aspects of sequence analysis. You might have seen that I am trying to fine-map the locus of the soul. Alternatively, you could look out for hidden messages from our Intelligent Designer. Considering the overall bizarreness of nature (the real one, not the journal), it would be more than reasonable to expect a number of ‘easter eggs’ hidden in the genome. What better place than the alleged ‘junk DNA’ to hide a secret message to the creation? Let’s go and jump-start the area of comparative religiomics!