How to (not) write a microbiome grant, Part I

I'm involved in something that I shouldn't talk much about, but suffice it to say I'm smack dab in the middle of evaluating multiple different microbiome grants. As such I'm the exact target audience you should be aiming to reach with your microbiome grant, literally, the exact target audience. There are a few grantsmanship things at the forefront of my reviewing mind right now that I'd like to get down in written form with the hopes of helping people out in the future (myself included). I'm especially wary going forward because there is currently a push to make microbiome studies the next BRAINI and we as a community are going to do ourselves a huge disservice if we get this wrong. The last thing I want to see is a huge movement for government funding within an area I care deeply about, only to have it not really come close to living up to the hype. I'll definitely have more to say on this as I digest some more, but here are a few things to keep you going:

1. Think about your hypothesis when designing experiments and keep this hypothesis in the forefront of your thoughts. 

Microbiome constituency is going to change over time at some level, it's going to change in some way with pretty much every manipulation you can think of. It is soooooo tempting to write experiments that focus on measuring this change over time, measuring how individuals differ, or that measure the variation associated with treatment X. Maybe one community changes at different rates than others? Maybe there is something magical and emergent that happens when you put all of this community information together?

When you are writing your grant, you are going to be couching the effect of the microbiome in terms of something. The microbiome is important for human/plant health, the microbiome affects geochemical cycling, the microbiome affects evolution of species, etc.....The problem I'm repeatedly seeing with grants is that the whole project is built with the idea that the microbiome will have an effect on X, and measuring how treatment Y affects the microbiome is important for understanding X, but to me at least it seems a lot of people forget to directly link treatment Y with its effect on X. If the treating the phyllosphere microbiome with jasmonic acid is going to change its constituency or dynamics, and jasmonic acid is therefore predicted to affect "plant health", please try to include measurements of "plant health" within your treatments. Keep the whole hypothesis in mind and don't just measure how treatment Y will change the microbiome and then assume that this change is going to impact X. Directly measure the impact of Y on X in the context of the microbiome.

2. It is very tempting to want to use the latest technology to measure "system level" effects of the microbiome. Proceed at your own risk, and with enough preliminary data to make me believe that you can adequately carry out the experiments. 

A couple of years ago, on a completely different panel, every grant seemingly included RNAseq. Now every grant is including metabolomics and metatranscriptomics. These technologies are awesomely powerful, and they will truly revolutionize some areas of science. If you are writing a grant, however, don't just say you will evaluate the microbiome with "metatranscriptomics". I want to see data for how many reads you might expect in a given environment (pilot experiments work wonders). I want to know that you are proposing to sequence enough depth to actually have a reasonable chance at seeing differences. Every system is different, and just citing papers that it's possible doesn't do the trick. This is especially challenging when studying microbiome communities within a host. Much of the metatranscriptomics work right now is being done in environmental communities, and a lot of the papers/technologies are being developed with these kinds of studies in mind. Any time you include a eukaryotic host in microbiome studies, you are going to get A LOT of eukaryotic RNA in your metatranscriptomes. Sure, pulling down with PolyA might clean up some, but for a lot of plant systems the chloroplast RNA isn't polyadenylated and is present at high frequencies. For 16s based studies you can design PNA blockers to limit the amount of eukaryotic contamination that comes through, but this doesn't work at a metatranscriptomic level yet. Sure, you can just throw everything onto a few HiSeq lanes and only use bacterial RNA reads, but as a reviewer I want to see enough information to convince me that using the "sequence everything and cull what you don't want approach" or maybe "the overkill-ome" is going to work. Tell me the fraction of reads that are host vs. microbiome, even if you only have this information from a small scale pilot study.

3. I don't ever want to see this in your grant:

Figure 1. Bad Hairball Plot. 
mage from: http://www.wired.com/wp-content/uploads/images_blogs/wiredscience/2013/11/bad_hairball.jpg

That is a plot of some random network I found with a Google search. Microbiome studies tend to sample many different taxa under different situations over time. It's tempting to put all of this interaction data together in a plot such as the one above, so that in your grant you can demonstrate that you do "systems biology". THERE IS NOTHING USEFUL VISUALLY IN THESE KIND OF PLOTS. In many cases, the nodes aren't even labelled. The only thing I can assume that this plot is trying to show is that you can do "systems biology". Grant space is so precious and limited, why waste it on a figure that doesn't relay any information at all?

If you do find interesting interactions, feel free to make a small figure including just a few nodes (labelled of course) that shows these interactions and explains what this interaction visually means. A hairball plot like the one above serves absolutely no purpose for the reviewer of a grant, and actually annoys me. Don't annoy the reviewer.

End of rant for now:)