The future of science, gradical change, and tools for the people

Maybe you've felt it - the buzz in a room, the tension in the air, the accelerating pace at which people are connecting and the realization that we're all in this together, even if we don't quite know what "this" is. At least in my small pocket of the world (wide web), something is brewing.

That something is The Future of Science. Michael Nielsen has written about this at length in preparation for his forthcoming book of the same name, with a lively discussion in the comments following. At BioBarCamp this past weekend (many thanks to John Cumbers and Attila Csordas for organizing!), the future of science became a recurring theme, with an impromptu discussion on open science the first day and spirited sessions on open science, web 2.0, the data commons, change in science, science "worship", and redefining "impact" and "failure" the second. Each of these topics could be their own blog series, and, in fact, many of them are. Even if people didn't always agree on the details, it was clear that everyone there (a biased group, inarguably) agreed that change is necessary, and inevitable. The question is, what will that change look like, and how will we get there?

The creators of Labmeeting.com put forth the following thesis:

Science relies on trust. Trust only remains intact when change occurs through consensus. Change through consensus is inherently gradual. (Therefore change in science must be gradual to succeed.)

Though you could agree or disagree with each statement, there are two things I'd like to discuss in particular. One is the issue of trust. Science relies on trust, right? I would say instead that science could be built on trust, if people weren't so worried about it! The most popular argument made against radical openness in science is based on the fear that other people will not act in good faith, i.e. if you make your lab notebook public, you could get scooped. And yet it is exactly this current climate of secrecy and cutthroat competition that encourages scooping and offers little recourse when it happens. If all research were open, digital, and timestamped, there would be an indisputable record of work and ideas that could be used to argue precedence.

Of course, this all starts to sound a little chicken and egg after a while. How do we assuage the fear of scooping enough for things to get sufficiently open so that scooping really isn't a problem? This brings us to the next point - that change must be gradual. Let me add the session leaders' conclusion to this: "the first step is to create incentives for scientists to voluntarily start doing the same everyday things on the same web platform." I think this is a valuable statement to keep in mind as more and more web 2.0 tools and platforms keep cropping up - that in some sense, the best way to enact satisfied change is to make it beneficial to the individual researcher, and allow them to discover this on their own terms. Scientists are a skeptical lot by training; the fact that they are also generally time-strapped and resource-starved makes them, ironically, reluctant to experiment, at least with the way they do their work. They neither need, nor want, another social networking tool.

The key that some groups have discovered (Labmeeting, Epernicus, and OpenWetWare among them) is to discover what people need, and then build something they will want. For Labmeeting, it is online paper management, for Epernicus it is effective question answering and resource finding (no more wild goose chases looking for someone who can help you with a specific problem), and for OWW it is tools for managing group websites and sharing protocols. Although Epernicus does rely on there being a social/professional network in place, the other two provide services that are useful even if you're the only one using it; the online community therefore can build itself without pressure. And Epernicus along with the others recognizes that in order to be successful among scientists, you need to provide them with something useful. In other words, you need to make tools for the people, rather than tools that need people.

So what about change? How will it happen and when? Well, I'm hoping Michael's book will tell us. ;) But I have a feeling it will be "gradical" - gradual at first, and then...

Mac hacks for research

As much as I sometimes want to think that Apple is the new Microsoft, I can't deny that they've got something that the evil empire never had - fanatical users who are loyal not because they have to be, but because they truly love Macs. In fact, they love Macs so much that they often devote their free time to developing stunning software applications that range from the quirky and fun (think Delicious Library) to the "how did I ever live without it?" (think Papers). The enormous array of applications available for Macs, unrivaled in their aesthetics, ease of use, and depth of features, serves to reinforce the Mac's reputation as the platform of choice for trendsetting computer users.

It turns out that this is true in the scientific domain as well. Joel Dudley, founder of MacResearch, gave a guest talk for my lab today on a dizzying array of Mac tips, tricks, and software meant to optimize the Mac experience, especially in a scientific research environment. Some of the applications he mentioned looked truly extraordinary, and I thought I'd describe some of the more notable ones here for those interested in getting more out of their Macs.



Macnification
For the cell and molecular biologists out there, here's a solution for your image processing needs. Macnification is like an extended iPhoto for microscopy. The full feature set looks impressive - you can track experiments, manage metadata, make measurements, create movies, and generate virtual z-slices through multiple images, all in one sleek application. I don't work with microscopy images, but now I wish I did!







NodeBox
For Python programmers wanting to flex their artistic side, NodeBox allows you to create amazingly complex graphics and animations with just a few lines of Python code. NodeBox is free and open source, with plenty of example scripts to get you started. Just looking through their online gallery is enough to get the "what-if" juices flowing.




Graph Sketcher and DataGraph
If you hate pretty much everything about Excel graphs, you might like everything about these two graph programs. Graph Sketcher is for quick, brainstorming type graph drawing - use their simple tools to draw pretty much any abstract relationship in 2D, with or without data. DataGraph is more powerful and meant to plot large volumes of data. The defaults start out fairly aesthetically pleasing, but there are many many ways to tweak the look of graphs, add or switch data, add additional axes, and plot multiple dimensions simultaneously. Both applications export to PDF for high-resolution figures, with DataGraph allowing export to vector-based formats as well for use in publications.

In addition to these, the Omni group has a suite of applications for boosting productivity, managing information, and drawing high-quality graphics (much more easily than the impossibly hard to use Adobe Illustrator); Journler is a great Mail-like program for organizing notes such as your lab notebook; and, of course, Papers is a must for anyone who reads scientific papers on a regular basis.

Be sure to check out MacResearch for more innovative applications geared especially towards science and research.

Online collaborative manuscript annotation

While at the inaugural AMIA Summit on Translational Bioinformatics the first half of this week (stay tuned for another post summarizing that), I started thinking about some ideas for tools that could help make discussion of papers easier and more productive.

Currently, it seems that there are a few avenues for discussing a paper: 1) have an informal conversation in person, 2) hold a journal club where one person presents the paper and discussion ensues, or 3) blog about it and hope others comment. (You could argue that another avenue exists through some journals - especially open access ones - allowing comments on published articles, but this hasn't caught on as far as I can tell.) There are several disadvantages of the current systems. In-person conversations or journal clubs can be stimulating as they happen, but are transient and usually go unrecorded, resulting in little tangible benefit to others (or often even the participants); they also usually preclude remote participation without some sort of audio-visual setup. Going the blog route allows anyone to participate, but it's difficult to connect the comments back to the paper and the discussion may be less productive than hoped.

A group of students in my human-computer interaction class a few years ago developed an idea called Collaboread for their final project. In essence, it allowed multiple online users to markup a document, enabling collaborative annotation. I'm sure there are several products out there that allow either online markup of documents (Adobe, for one) or collaborative editing (Google Docs), but I haven't seen anything that resembles exactly what I envision.

Suppose you are viewing a document on the screen - maybe a full-text articles at BioMed Central, or a PDF. Clearly, things like web URLs and references should be hyperlinked already. But suppose you could create additional hyperlinks, such as to wikipedia pages, other papers that were not referenced but are relevant, blog posts, etc. You can also start individual discussion threads attached to a particular results, claims, or points made in the paper, or to tables or figures. Mousing over or clicking on the icon indicating such a thread would bring up a summarized view of the thread overlaid on the screen which you could browse more deeply or hide if you decide you're not interested. The idea is to make a richly annotated document that is easy to read but at the same time make it easy to see what other people thought or were confused by and respond if so inclined without too much disruption. When I envision this, I see a Google Maps-like navigation and manipulation style with lots of linked text, little colored balloons at the POIs - the discussion threads, and liberal use of tags to help with filtering and searching of the document and annotations.

A tool like this would be useful not just for journal club-style discussion of papers, but also as a teaching and editing tool. Authors could collaboratively comment on a paper, or learn from others' comments after it is published and made the focus of such a discussion. Readers and students would benefit from the additional linked resources and learn from the discussions how to critique a paper. And the annotated document would be available to everyone long after discussion has tapered off.

Of course, there are potentially many technical, legal, social, *al issues surrounding this, but I think some kind of tool along this vein would be useful and interesting. Does anyone know of any tools that do these things already? If not, I am already looking into what it would take to develop it, and would appreciate tips, suggestions, warnings...

Tools for analyzing "lists" in biology

My latest research is focused on cluster/list annotation in biology. Given a cluster or list of genes or proteins that were grouped together using some metric (expression profile, sequence or structure similarity, interactions, etc), how can you discover descriptive terms or labels for that cluster? This seems to be a common question, and yet I've had trouble finding tools that help you do what I am specifically trying to do (investigation of a list of biological entities). I've found many that can give you tons of information for single genes or proteins, which I don't consider that helpful, and a few that can give you information for a group, but these are either organism specific or limited to one or two types of data (e.g. GO terms).

Since I am developing a method to do this based on text, I'd like to be able to compare my method to existing ones that solve the same problem. What I am looking for is two or three available methods that give you information relevant to a list of biological entities from multiple species, at least one of which uses literature or text-mining. Does anyone know of such methods, or have ideas of where to look? Various PubMed and Google searches have failed me!

Unrelated, but also done today: Submitted the PSB proposal to Nature Precedings as per several of your requests. Will update once word is back from their review process.

Science journal feedmixes

The topic of literature review came up at a recent group meeting. Our advisor receives a number of print subscriptions to journals, but these often languish in some forgotten corner. Even when they are brought out of the depths, it seems a daunting task to leaf through them to find articles of interest to each student. Since everyone is on the interwebs, it is much easier (and complete) to get updates on relevant articles through a website or email, peruse the titles and blurbs online, and then decide what to actually sit down and read. There are a few problems with this, however.

1. Getting alerts from journals, search engines, or aggregators like Faculty of 1000 still usually produces too many articles to sift through.
2. To limit the amount of junk you get, you provide keywords - but, if you're like me, you will browse through unlikely articles in Science or Nature or PLoS ONE on a regular basis because they look interesting, so keywords will filter these out.
   

I've set up a feedmixer for science journals and related information using Feed Digest. It's little more than an aggregator right now so it doesn't really address those problems. If anyone knows of any cool tricks to help sift through the ridiculous amounts of information we're supposed to keep up with, without losing the unexpected gems, I'd love to hear it!

Update: A new tool called Persai claims to learn your preferences through what you accept and what you reject (review on Slate), and filters your feeds accordingly. I'm not sure it will help with issue #2, but it's probably just an irreconciliable trade-off between #1 and #2. Perhaps the solution is to have a couple different pages set up with Persai - narrow ones for specific fields or interests, and broader ones for the science "pleasure reading"!

Review on open source CMS for bioinformatics

An alum of my lab recently published a review on open source content management systems and their uses in bioinformatics.