Sustained software development, not number of citations or journal choice, is indicative of accurate bioinformatic software

• DOI: 10.1186/s13059-022-02625-x
• Publication Date: 2022-02-16T15:02:36Z
• Authors: Paul P. Gardner; James M. Paterson; Stephanie McGimpsey; Fatemeh Ashari-Ghomi; Sinan U. Umu; Aleksandra Pawlik; Alex Gavryushkin; Michael A. Black
• Keywords: Human genetics; Genome Biology

Computational biology provides software tools for testing and making inferences about biological data. In the face of increasing volumes data, heuristic methods that trade speed accuracy may be employed. We have studied these trade-offs using results a large number independent benchmarks, evaluated whether external factors, including speed, author reputation, journal impact, recency developer efforts, are indicative accurate software.We find citations age unreliable predictors accuracy. This is unfortunate because frequently cited reasons selecting tools. However, GitHub-derived statistics high version numbers show bioinformatic generally product many improvements over time. also an excess slow inaccurate tools, this consistent across sub-disciplines. There few middle-of-road in terms trade-offs.Our findings indicate primarily long-term commitments to development. addition, we hypothesise bioinformatics suffers from publication bias. Software intermediate both difficult publish-possibly due author, editor reviewer practises. leaves hole literature, as ideal fall into gap. High not always useful if they slow, while inaccurate.