There are often discussions about whether the bioinformaticians should spend more time thinking about better algorithms or building more user-friendly implementations. The answer is neither, as a cancer benchmarking study found out. Quoting from ‘A Comprehensive Assessment of Somatic Mutation Calling in Cancer Genomes’ -
We also detected distinct clustering of different analysts. Even though analysts used similar combinations of software in their pipelines, very few similarities were detected in their calls. The combination of software is not as critical as how each piece of software is applied and what settings are applied. A slight correlation of true positive SSM calls of pipelines using MuTect and Strelka can be seen (Figure 4). Data analysis pipelines are constructed by integrating software from diverse sources. In many instances the approaches used in the different software and the assumptions made therein are not evident to the user and many parts are black boxes. In order to account for unknowns, pipeline developers resort to calibrating their pipelines against known results, for example from a genotyping experiment done on similar samples, and by using combinations of tools for the same process step, assuming that a result shared by two different approaches has a higher likelihood to be correct. Of note is that many of the pipelines apply multiple tools for the same process step in their pipeline and then use intersects to weight calls. This practice, together with good use of blacklists, results in best outputs. No best tools emerged from our benchmark and it is also clear that no strict best threshold settings exist. However, it is clear that how analysts use their pipeline is critical for the quality of their output.