This final “ility” is also critical as scientists rely on the ability to compare results from different organizations. Issues surrounding reproducibility have reached unprecedented heights, as a significant rise of peer reviewed scientific paper retractions takes hold (New York Times, April 16, 2012). In addition, a number of industry studies and reports have cited an unacceptably low reproducibility rate in scientific publications (Nature Chemical Biology 9, 345, 2013). It’s no wonder that www.reproducibilityinitiative.org now exists to provide authenticity to scientific publications.
At a more granular level, it is easy to see why results are difficult to reproduce. Results used during lead optimization programs typical in the industry are compared in many cases using a single, or maybe a few, numeric results from individual biological assays. For example, an IC50 (the concentration of a test compound that exhibits a 50% inhibition of the biological process being measured) is calculated using a four-parameter logistic fit model. Whilst the IC50 is the key measure, it should be taken in context with the other three parameters calculated (slope, max asymptote, min asymptote). Then add the choice of more than one fit model and allow for human intervention. The scientist may indeed remove a number or two from the original dataset in order to improve the quality of the result. With even these seemingly minor variables in data analysis, differences in results will be seen and may or may not affect subsequent decisions.
Publishing (internally or externally) these four parameters on their own, without important details such as the binding Michaelis constant, or the concentration of substrate or ligand used whilst performing the assay, negates the value and comparability of the results. The specifics of every calculation should be encoded within a portable and incorporable entity that can be passed between collaborators or organizations. Better still, the full scientific method should be encapsulated with this package of information.