External (bioinformatics) specialists: best on the grant from the onset

By Grant Jacobs 15/01/2010

Specialists should be involved in planning grant applications, not brought in later. (I’m writing mainly about data analysis, as I see this as less of a problem for data generation.)

Researchers should involve those who will analyse their data as they write their grant applications to assist the experimental design and to make more pragmatic employing any external researchers or consultants (or contractors) they might need.

Most (better) researchers recognise the need for good data analysis and exploration, but they seem to less often consider to have these people involved as the project plans are being laid down, i.e. while writing the grant application; these people seem to more often be brought in after the fact.

When large amounts of data are involved, researchers seem to readily look to external assistance, presumably frightened by the scale of the data, but the need for specialist advice is in many ways even more true for smaller projects that have specific questions their data is to address. The nature of the question is important than the amount of the data in many ways. (Well-established large-scale analyses often have relatively “standard” analyses compared to focused projects.)

I’m mainly writing with molecular biologists or geneticists in mind, as these are the scientists I interact with, but what I’m writing about applies to other forms of data analysis and also to data generation. (My own interest comes being an independent computational biologist or consultant, contracting to research groups and commercial companies. Some of what I write may be biased by being computational biologist, as opposed to a bioinformaticist, i.e. someone who reads extensively on the experimental biology side of topics as well as the analysis/development side.)

In the case of data generation, this seems less of an issue. Data generation frequently requires expensive equipment (think: DNA sequencing, proteomics, spectroscopy, etc.) and odd as this might be to say, experimental researchers seem to have a better appreciation of “upstream” skills in creating data, than they do of “downstream” skills in data analysis.

For data analysis, I suspect the ready availability of software (in the case of bioinformatics or statistics for example) creates an optimism amongst researchers that “we’ll figure it out.” The time and infrastructural issues can be bigger than people realise. While it’s likely that many (if not most) will eventually figure it out–if they can commit the time involved–it’s often better use of time and money to use people that are already up the learning curve.

(This is also a concern for my own work, too: if the effort to come up to speed on a job is too great, I simply have to turn it down as the overheads won’t be covered. Operating “commercially” forces me to be hard-nosed about this: it’s my income gone if the overheads prove higher than anticipated.)

In my experience it’s certainly an issue in cases of bioinformatics taking place within biological research groups.

There is a tendency for some research groups to try do everything in-house. I would suggest for things that involve background in a field that they’re not familiar, research groups are likely to be better served by working with those that already have the necessary background (or, if those with the general expertise are new to the particular analysis, are better placed to get up to speed than someone from outside the field).

While this must seem obvious to the point of being trite, it runs against some scientists’ natural instinct to want to learn to analyse one’s own data themselves and to “control” their data or project. Ordinary old human emotions, in other words.

Approaching an external worker “after the fact”, when researchers have discovered it wasn’t the best move to go it alone, leaves the external worker in the position of doing a “rescue effort.” This can result in a delay. It’s much better that everyone be in place in advance, so that preparations can be made and any issues sorted in a timely manner.

There are other reasons it’s wiser to bring those who analyse their data (or work as external researchers providing other services/skills) onto the grants as they are written, at least under the current regime in NZ.

I don’t get to review grant applications (as I’m not part of the university/CRI system), but I’d like to think that they’re more likely to succeed if you can show you have run your plans run past someone familiar with data analysis. The impression I get is that it’s more common to write that the project will “hire someone with appropriate expertise when the time comes”, which assumes that the data analysis portion of the plan is sound.

Another issue is that it’s easy to miss opportunities. While a few of the more widely-used bioinformatics methods are reasonably well-known to experimental workers, they’re unlikely to know of more niche approaches that may better suit their specific project and may yield additional information. (They’re even less likely to see opportunities to develop new analytical methods.)

I favour a computational biology approach. I’m sometimes asked to apply a particular analysis to a dataset, but this has the researcher having already made the decision that I should be assisting them making. (The same general issue applies to external work in any other area: take note of who is best placed to make what decisions.)

My response is almost invariably to try coax the researcher to back up and explain the biological problem that they’re trying to address and why they have chosen the experimental approaches that they have. In this way I can address the problem as a biological question and also spot any opportunities that might benefit their project. (I am a biologist, after all, just one who specialises in theoretical approaches.)

You’re unlikely to get this perspective from junior staff hired to “fill” a role, it’s one place where experienced external workers have more to offer. If nothing else, the external worker can give an independent view on the project from a quite different angle which can be helpful in writing the grant application.

A more pragmatic reason to get people involved while writing the grant application, which might be limited to New Zealand, is if people try acquire the use an external worker after the grant is awarded it can be fiddly to sort out full-cost payments as the money is already committed. It’s easier to do this at the time the grant is being written.

There is at least two main aspects to this, one about understanding how much research work actually costs and the other about the politics of paying full-cost rates once the university administrations are controlling the grant funds.

What an external worker is seeking is not a salary, but a contract or consulting rate that includes overheads.

Overheads are higher than people not familiar with accounting their own time and funds often appreciate. In particular, people have a tendency to under-estimate time overheads. This is also true of those starting a business for the first time, or who don’t think their business time through thoroughly; this one isn’t peculiar to academics!

University overheads in NZ are high enough that I (for example) can compete with even a post-doc, yet academics commonly compare my rates with salaries, not with the full-cost of their own staff. Needless to say, the false comparison draws the wrong conclusion! (I’m not going to try break the overheads down but it’s sufficient to say it’s more than just a simple multiplier of the staff’s income or comparing with their nominal university “full cost”.)

Long story short, to avoid the “bother” of dealing with the university administration to get the overheads accounted for from a current grant, it is simpler to have the external worker on the grant at the time the application is made.

(I sometimes think that we’d get a “spend” of the tax payers’ dollars if the issues of paying full rates to external workers were resolved as it’d enable external workers to compete more effectively against university staff , but I imagine the effort to get this in place is large…)

Other bioinformatics posts on Code for life:

Developing bioinformatics methods: by who and how

More on “What is a computational biologist?” (and related disciplines)

Retrospective: The mythology of bioinformatics

Bioinformatics — computing with biotechnology and molecular biology data

Computational biology: Natural history v. explanatory models

0 Responses to “External (bioinformatics) specialists: best on the grant from the onset”

  • Hard to disagree with much of what you write! I too am a research consultant and am often brought in to analyze data after the fact. It is all to common for researchers to not have a clear idea of how the data are to be analyzed at the time of experimental design (i.e., during grant applications), or even after they’ve finished data collection (and thus not necessarily allocating enough funding to cover that part of the research). All the grant applications I’ve been involved with though, do cover this aspect, and those projects tend to go much more smoothly. However, I’ve found across many projects that the time sink is in screening large datasets and preparing them for analysis. Too often I get given an excel spreadsheet that has data hand-entered with no thought as to how those data might be cranked. Text is mixed with numbers, no limts are placed on what constitutes valid data, formatting and comments are used to indicate key information, no sanity-checking done on manual data entry, no unique code to cross-link subjects to their data, etc., etc. I’ve spent entire contracts simply getting the data into a form that it can be analyzed. Much of this can be mitigated by good upfront planning by someone who understands how to collect data in a form that makes it easy to analyze, with minimum chance for human data-entry error. But I never see that part of the effort mentioned in grant apps; even those that discuss how the data are to be analyzed.

  • Great comment. I agree with all you write! (It’s is rare for me to entirely agree with a comment… that worries me a little…)

    Bioinformatics datasets usually are better than hand-collected data (!) but there are sanity checking and validation issues, which are some of my pet peeves. (I have long had concern about the lack of decent quality control on some “data pipeline”-styled efforts in bioinformatics, for example.)

  • […] My first suggestion would be to talk to experienced bioinformatics scientists or computational biologists. It’s a lot quicker, and it’ll save you making mistakes that the research literature assumes you know better not to. You’ll also avoid the trap of thinking that something is simple that in fact isn’t. In particular, you should talk to your local analytical experts well before collecting the data, ideally before drafting the grant application. […]