10 years on, same old same old

September 2, 2002

So what new skills will postdocs need to ensure that they don’t become science relics? The answer is math, statistics, and knowledge of a scripting language for computers.

— ­The Scientist, “Bioinformatics Knowledge Vital to Careers.” 16(17): 53.

February 8 2012

But two other skills are increasingly necessary: expertise in computer-programming languages designed to aid manipulation of large data sets, such as R, Perl or Python, and the ability to use these languages to analyse large amounts of data quickly.

— Nature, “Biostatistics: Revealing analysis.” 482: 263–265.

9 thoughts on “10 years on, same old same old

  1. I think the key is Statistics. Knowledge of statistical analysis will enhance every postgrad professional.

  2. ummm … I think if they think Perl and Python are sufficient for manipulating large data sets by themselves, people are in for a rude shock (depending on your definition of large)

  3. I think one thing that’s different from 10 years ago is that back then, nearly all bench types were frightened of the UNIX command line (sometimes even of PAUP on a Macintosh or running BLAST from the NCBI website!). These days, while there are still technophobes, there is a new breed of bench scientist, one that isn’t really a bioinformatican, but one who can at least write simple scripts to reformat data and navigate around the command line to run existing pipelines.

  4. I think these days there are a lot more scientists that are using some scripting, stats and re-using existing data to enhance their own bio experimental results. I know a couple of great example of this. There is also such a huge competition for jobs that they are getting competitive advantages so its going to keep trending up. I have been teaching a PhD course for biomedical students and I try to pass on this message that you need these skills to be competitive but not all of them (want to) believe me :).

  5. Thanks for the comments – keep them coming! I posted this one because I had a strong reaction to it, but I can’t figure out exactly why it “set me off.” So I find the reactions of others very interesting.

  6. I’ve come into closer contact with bench-oriented folks again over the last year or so and I’m rather saddened at the lack of quantitative insight or instinct. People who measure things – complex things, in big machines that go “whoomph” a lot and have kinda slick interfaces that spew numbers – rarely seem to care about little things like reproducibility, measurement error, power or experimental design.

    Make an observation on a sample? Good, now do a second one and it’s a wrap. Paper writes itself!

    Ok, ok, I’m going back to my nursing home. But in my day, biochemists knew an integral from a differential. They did kinetics for pity’s sake!

  7. Belatedly, but having sat in on a few of these conversations, the pool of “essential” skills tends to blossom out ferociously – UI design, ontologies, functional testing, chemistry! – perhaps in line with the widediversity of things that pass for bioinformatics. I’ve decided there are only 3 essential core skills – scripting / programming, statistics, biology (of the domain you are investigating). Any less, and you’re “just” a programmer or following instructions. Everything else is optional, depending on your field.

    • I’d rank them differently: 40% statistics, 40% biology, 20% scripting. My justification is that knowing what to measure and how to do it is crucial. Tool use is then /relatively/ easy to acquire. All the scripting + biology in the world don’t give you any insight into the robustness of results (or teasing out signal from noise); conversely, all the statistics + scripting in the world don’t guarantee you’re asking a sensible question.

Comments are closed.