Written by: C. Titus Brown
Primary Source: Living in an Ivory Basement
We all know that biology (along with other sciences) is becoming ever more data intensive. Biologists (among other scientists) are not terribly well prepared for this, because of a lack of computational culture, lack of computational training, and a lack of tools. What do they need to know?
This question drove us in organizing a July 8 & 9, 2013 meeting at SESYNC, in Annapolis, MD. Greg Wilson (of Software Carpentry) and I (MSU/BEACON) co-organized it: I provided the money (via an NSF grant), Greg provided the advice on how to spend it :).
A truly fantastic group of people deigned to attend — I discovered that one real value of having blogged, taught, and proselytized open science, education, and training for years, is that even when you invite people to a hot, muggy locale in the middle of summer, they come! We ended up a pretty diverse group of people, almost all actively and energetically involved in some form of teaching, with many focused on teaching computing to biologists.
We spent two days discussing, brainstorming, pitching, and outlining ideas to and at each other. We generated a 500 line+ Etherpad document in the process, and Greg has said that he will try to summarize things in some upcoming blog posts. However, since I’m still riding the enthusiasm and energy of the meeting, I thought I’d blog a few of my morning-after thoughts — nothing systematic, just some of what stuck.
Mixing up activities is really important!
The social and topic mixing is even more important than I thought. I arrived early Monday morning due to travel delays, and was fried for much of the day, so Greg ran things on Monday; we had a few presentations, a breakout session to write a “driver’s license” for biologists who wanted to use computing, a use-case writing stint, and (carried over to Tuesday) an opportunity to pitch an idea to the group as a whole. In contrast, I organized… a few whole-classroom discussions, which, while useful, were kind of boring and un-energetic. If I had to take one thing away from this organizing experience, it’s that smart, creative, and dedicated people don’t mind non-linear conference flow, and in fact can profit from a certain amount of creative chaos to mix things up.
Everyone who actively teaches has something to contribute
I asked people, what do you do to keep your online material up to date? A perennial problem for trainers is that the pace of change in technology makes most material go stale relatively quickly, and it’s hard to detect and takes time to fix.
I was hoping for a few nuggets of wisdom — things that I, with my vast experience and ginormous brain (<- sarcasm), had not prioritized or emphasized. I got much, much more, in terms of practical experiences from other people! People are actively using a mix of continuous integration, high trainer turnover with active remixing, doctest-style output generation, interspersing flipped material with lots of in-person courses (which makes sure the flipped material stays up to date), and “report an error” buttons. No magic bullets, but, collectively, a set of attitudes and low- and high-tech approaches that provide a wide range of options.
More importantly, I realized that stale material that no one looks at is rather unimportant; and that there really was no magic bullet. (See also “Now I know what no one knows,” below.)
Everybody agreed that figshare was what we should be teaching for basic data sharing. I was surprised at the unanimity on this point — the only negative bit was that we worried that figshare would go the way of Mendeley, and be bought out by an untrustworthy actor.
A bunch of reasons to use the command line (or not)
Vicky’s 5 minute pitch was to have all of us — in the remaining 4 minutes and 30 seconds — pair up and write three reasons why the command line is (or is not) important for biologists to learn.
Here are some of the results:
- showing them that they can automate stuff that currently takes them ages
- showing them how it helps the, to reproduce/repeat things
- being able to do things in a less error-prone manner-doing things manually may introduce human errors.
- saves time
- makes it easier to automate, link stuff together
- identifying a package necessary to perform derived analysis then show how to download and install through command line
- give tasks like renaming 100 files manually win 10min: see how slow many errors then show very fast and accurate on command line
- show simple common data + problem: solve it: data transformation problem
- Make commands available (like R): side-by -side cheat sheets for common commands_cut-paste sampler
- Lots of useful tools require command line input
- you can automate some of your processes + save time for beer-drinking
- you can modify packages that ALMOST do what you want
- not for everyone, specify group need
- working with large amounts of data and compute is easier on the command line (but may change)
- command lines do not lie
- if you want to be involved in active research in five years you need it.
- direction of where science is going is going to depend more and more on comp. literacy
- fetching, access to data speeds up research
- it’s fun
- more empowering
- always get latest versions of programs
- provides record of what you actually did
- run on bigger computer-provide speed efficiency
- provides ability to repeat tasks easily
- example command line versus excel for tasks of interest tobiologists (data crunching), lead/proof by example/hand-on
- repeat (lead by example) for workflow (R or Python script vs screen shots)
- buy them or provide them with a copy of “Data Crunching” (Greg’s :-)
I didn’t really sift through them yet, but there are some gems in there.
Now I know what no-one knows
One important outcome for me was this: there are many approaches to w2tbac that have not been tried, or not tried thoroughly. Once you know what people have tried, and understand what they are doing currently, you have two valuable pieces of information: a list of people to go talk to about how they do what they do & what hasn’t worked; and, equally, some idea of what hasn’t been tried and might be worth trying.
More succinctly, I now have a much better idea (through the process of elimination :) of what has not been tried!
We came out with some ideas that we really want to try and that I think will actually happen. Syndication, above; focusing more on building lessons around narrative paths; collaboration on protocols & questionnaires for assessment; trying to share more of the process of recovering from errors; and thinking more about blended learning, where we regard online training as one component of training, and don’t force a false dichotomy between online and in-person training.
My personal TODOs
My TODO list now includes:
- Implementing a “data biologist” driver’s license quiz;
- Write up an Excel -> R/Python day-long lesson;
- Go out to iPlant to grok their cloud-for-scientists infrastructure;
- Run a focused workshop at NEON;
- Develop a simple, short, common pre- and post- questionnaire for workshop assessment, for those workshops that are bio focused;
- Integrate IPython Blocks and IPython Browsercast into my term-long course on Computing for Evolutionary Biologists;
- Start making cheat-sheets one of the specific products of my various training activities;
- Get permission from instructors to relicense all of the NGS course material under CC0;
- Arrange a week-long hackathon somewhere nice;
and hey, look, I have funding and/or motivation for all of that. Yay!
Latest posts by C. Titus Brown (see all)
- On gaining tenure as an open scientist - December 26, 2014
- My review of “Determining the quality and complexity of NGS…” - December 18, 2014
- The post-apocalyptic world of binary containers - December 17, 2014