April 6, 2006 | BOSTON -- Unless science adopts a “bottom-up” rather than a “top-down” approach to Web-based knowledge sharing, it risks being bypassed by developments on the Web. That was the main thrust of a “Web 2.0” workshop at Bio-IT World’s Life Sciences Conference + Expo in Boston this week.

Tim O’Reilly, founder and CEO of O’Reilly Media, pointed out that during the dot-com bust, those dot-coms that survived did so because they harnessed the collective intelligence of their users to make their application better. Google, Amazon, Yahoo, and other survivors have three to four invitations per page for users to contribute, O’Reilly pointed out.

The idea behind Web 2.0, said O’Reilly, is to let users add value to the content you provide. Another example of the power of this model is craigslist.com, which, despite being the seventh most popular site on the Web, has just 19 employees – most of its content is generated by users. O’Reilly would like to see the scientific world apply these principles to their applications.

However, O’Reilly’s characterization of the Semantic Web as a “top-down” approach was rebutted by Teranode product strategist Eric Neumann, who argued that designating vocabulary for the Semantic Web is simply a “starting point,” and that user input is also taken into account.

Scientists as Copper Cable
Declan Butler, Europe-based reporter for Nature magazine, said Web 2.0 can bring us back to original idea of the Web as a collaborative workspace. Just as developing nations bypassed copper-wire telephones and leapfrogged directly to cell phones, “the biggest revolution on the Web is passing science by,” he warned.

Science has a collective intelligence that can be a drawback, Butler said, if scientists stay wedded only to traditional methods such as paper-based peer-reviewed journals.

Very few scientists blog, he pointed out, for fear it could damage their careers. But many younger scientists see the current system as over-competitive and prefer to collaborate. And he believes scientists are enriched by blogging. In fact, a peer-reviewed blog such as scienceblogs.com, postgenomic.com, and realclimate.org, can serve as a journal by acting as a collaborative filtering and sorting service.

Beyond the NCBI and other highly organized databases, Butler said, there is very little data sharing among scientists. He quoted UC Santa Cruz bioinformatician David Haussler, as saying: “Scientists are still in the Dark Ages when it comes to sharing data.” For example, Butler said, the World Health Organization has no usable data on avian flu. And a Swedish site on avian flu presents the data in PDF files.

But there are signs of change. Mashups – user-generated applications that integrate new data into existing applications – are beginning to gain attention in the scientific world. For example, AntBase and AntWeb bring together data on 12,000 species, using Web services. And a mashup that Butler created single-handedly, mapping poultry density per nation and outbreaks of avian flu over Google Earth, resulted in six nations contacting him about a collaboration.

PubMed and Productotypes
Jim Ostell, chief of information engineering at the National Center for Biotechnology Information (NCBI), explained the design principle behind the NCBI site that has resulted in ever-increasing amounts of traffic. The NCBI site, which includes PubMed, PubMed Central, GenBank, and other public databases, has become a content aggregator that allows users to make connections across various fields, said Ostell. When linking information, the “central dogma” is key. This means selecting elements that are things and then finding “computable values” between them.

Despite what appears to be a “top-down” approach, user input is definitely part of the process of building the NCBI site, Ostell said. “We don’t put out prototypes, but rather ‘productotypes,’ ” quipped Ostell. “Productotypes” are prototypes that are “good enough” to be put into the public space. When they get calls from users that a certain “productotype” is not working as it should, NCBI programmers use that input to refine the model.

Ostell also pointed out that the NCBI is working on the Discovery Initiative to present users with deep links for their topic.

Timo Hannay, director of Web publishing for Nature Publishing Group, presented Web 2.0 projects under development, including the Open Text Mining Interface (OTMI), and Connotea.org, which helps researchers organize and share their references.