Publication bias, in which positive results are preferentially reported by authors and published by journals, can restrict the visibility of evidence against false claims and allow such claims to be canonized inappropriately as facts.

Although questionable research practices inflate false positive rates, they have little effect on replicability because they also increase power.

Open data provides an opportunity to perform new analyses on preexisting data, but trade-offs are required to limit an increase in false positives.

Funnel plots of the Standardized Mean Difference versus the standard error are prone to distortion, leading to false-positive tests for funnel plot asymmetry, and therefore using the Normalised Mean Difference, or a sample size-based precision estimate, are more reliable alternatives.

A robust library preparation and variant calling strategy (DigiPico/MutLX) eliminates false positive single nucleotide variants from picogram quantities of DNA following whole genome amplification.

The use of Research Resource Identifiers (RRIDs) improves the proper use of cell lines in the biomedical literature.

Analysis of data on drug-gene interactions suggests that decentralized collaboration will increase the robustness of scientific findings in biomedical research.

Great care is needed when interpreting claims about the genetic basis of human variation based on data from genome-wide association studies.

Each stage of the scientific life cycle stands to benefit from the introduction of data science techniques.