Systematic benchmarking of omics computational tools

Serghei Mangul, Lana S. Martin, Brian Hill, Angela Ka-Mei Lam, Margaret Distler, Alex Zelikovsky, Eleazar Eskin and Jonathan Flint

Go to the profile of Nature Communications
Aug 15, 2018
Like 0 Comment

Received: 23rd July 18

Rapid technological advances, such as the development of next-generation sequencing, are driving the need for comprehensive computational tools to analyze the wealth of generated genomic data. Systematic benchmarking has successfully helped non-computational researchers in many different disciplines evaluate the accuracy and applicability of new computational tools.  Adopting a standardized benchmarking practice and following established principles for the design of new benchmarking studies could help researchers who use omics data to better leverage recent technological innovations. In this Review, we discuss challenges and limitations of benchmarking across various domains of modern biology. A reviewed 26 benchmarking efforts performed in this study, reveals a dominance of non-effective and non-systematics benchmarking practices adopted by the community.  We have summarized systematics principles for guiding the design of new benchmarking studies.  Proposed principles will make computational biology benchmarking studies more sustainable and reproducible, ultimately increasing the transparency of biomedical data and results.

Read in full at OSF PREPRINTS.

This is an abstract of a preprint hosted on an independent third party site. It has not been peer reviewed but is currently under consideration at Nature Communications.

Go to the profile of Nature Communications

Nature Communications

Nature Research, Springer Nature