Given the amazing advances in software and computational infrastructure in the past decades, one would be tempted to conclude that this is a one-way path, with researchers attempting to offload as much as possible of the complex data analysis and prediction to computers. But instead, researchers are finding that for tasks that have a strong visual element to them, humans are still supreme. Harnessing the visual acuity, smarts and even intuition of many people, a crowd can outperform even the most sophisticated algorithms.

Crowds of people are still superior to computers for many visually based data-analysis tasks. Credit: PhotoDisk

Crowdsourcing of scientific data analysis partly grew out of distributed computing efforts that harnessed personal computers to create networks whose power rivaled or exceeded that of the most powerful supercomputers. Creation of the Foldit crowdsourcing project, for example, was spurred by users of the distributed computing program Rosetta@home, who wanted to show the software how to solve protein structure problems that appeared easy to a user looking at the structure. The power of crowdsourcingprotein structure analysis with Foldit was demonstrated with the solution of the crystal structure of a retroviral protease in 10 days (Nat. Struct. Mol. Biol. 18, 1175–1177, 2011) and the redesign of an enzyme to increase its activity >18-fold (Nat. Biotechnol. 30, 190–192, 2012).

Although there is a strong desire among the public to participate in the analysis of scientific data, it can be challenging to make the tasks stimulating enough. Tracing neurites through neuronal tissue is a challenge for computers. Humans are more reliable, but neurite tracing is both tedious and exacting. When someone's attention wanes, mistakes are made. A potential solution to this quandary is to incorporate the task into a game. A company called scalable minds is working with the Max Planck Institute of Neurobiology on a crowd-sourcing effort called Brainflight to create games to help map the wiring of the brain.

It is also possible for researchers who lack the ability to distribute crowdsourcing software tools to get into the act. The Amazon Mechanical Turk allows 'requesters' to post tasks to be completed by 'workers' for a nominal fee. The structure of the service limits the types of analyses that can be performed, but the ability to restrict workers to those who have passed specified qualification requirements can be valuable.

There is no doubt that computational analysis will continue to expand and dominate most areas of biological research. For some tasks, however, a crowd of people and the right tool to connect them to the data could be superior for years to come.