Lessons Learned: Recommendations for Establishing Critical Periodic Scientific Benchmarking

Abstract

AbstractThe dependence of life scientists on software has steadily grown in recent years. For many tasks, researchers have to decide which of the available bioinformatics software are more suitable for their specific needs. Additionally researchers should be able to objectively select the software that provides the highest accuracy, the best efficiency and the highest level of reproducibility when integrated in their research projects.Critical benchmarking of bioinformatics methods, tools and web services is therefore an essential community service, as well as a critical component of reproducibility efforts. Unbiased and objective evaluations are challenging to set up and can only be effective when built and implemented around community driven efforts, as demonstrated by the many ongoing community challenges in bioinformatics that followed the success of CASP. Community challenges bring the combined benefits of intense collaboration, transparency and standard harmonization. Only open systems for the continuous evaluation of methods offer a perfect complement to community challenges, offering to larger communities of users that could extend far beyond the community of developers, a window to the developments status that they can use for their specific projects. We understand by continuous evaluation systems as those services which are always available and periodically update their data and/or metrics according to a predefined schedule keeping in mind that the performance has to be always seen in terms of each research domain.We argue here that technology is now mature to bring community driven benchmarking efforts to a higher level that should allow effective interoperability of benchmarks across related methods. New technological developments allow overcoming the limitations of the first experiences on online benchmarking e.g. EVA. We therefore describe OpenEBench, a novel infra-structure designed to establish a continuous automated benchmarking system for bioinformatics methods, tools and web services.OpenEBench is being developed so as to cater for the needs of the bioinformatics community, especially software developers who need an objective and quantitative way to inform their decisions as well as the larger community of end-users, in their search for unbiased and up-to-date evaluation of bioinformatics methods. As such OpenEBench should soon become a central place for bioinformatics software developers, community-driven benchmarking initiatives, researchers using bioinformatics methods, and funders interested in the result of methods evaluation.