Efforts to Improve Data Transparency at the JBC Roger J. Colbran - PowerPoint PPT Presentation

Efforts to Improve Data Transparency at the JBC Roger J. Colbran Associate Editor Credit: Amanda Fosang, Associate Editor 2

Why Improve Transparency? • Irreproducibility of pre-clinical studies • Negative impact on political and public will to fund research…... 3

Reason for Irreproducibility • Poor methodological descriptions • Reagents not described adequately and/or validated – antibodies, siRNAs, small molecules • Inappropriate data interpretation 4

NIH Principles and Guidelines Rigorous statistical analysis • Transparency in reporting • Establish guidelines for: • • Image based data • Western blots Descriptions of biological materials • Antibodies: source, dilutions, validation • Cell lines: source, authentication, mycoplasma • • Animals: species, source, strain, sex, age http://www.nih.gov/research-training/rigor-reproducibility 5

Efforts Toward a Solution…... Improve transparency of JBC papers: 1. Improve description of methods, including reagents 2. Clearly define reproducibility (technical and biological replicates) 6

JBC Policies and Guidelines 1. Experimental uncertainty, statistics 2. Graphical presentation of data 3. Western blots – presentation/quantitation 7

1. Experimental uncertainty, statistics NIH: • Require full reporting of statistics, including tests used, exact value of N, dispersion/precision (e.g., mean, median, SD, SEM, confidence intervals). • Inclusion and exclusion criteria: State criteria used for exclusion of any data or subjects. • Policies for statistical analysis should be included in Information for Authors. Journals should have mechanisms to check for statistical • accuracy. 8

1. Experimental uncertainty, statistics "There are three kinds of lies: lies, damned lies, and statistics." Attr. (by Mark Twain) Benjamin Disraeli, British Prime Minister Transparency is the key….... 9

1. Experimental uncertainty, statistics JBC Instructions to Authors: • Authors must include information on uncertainty and reproducibility of data in figure legends. • Authors must state numbers of independent samples (biological replicates) and replicate samples (technical replicates) analyzed and report how many times each experiment was repeated. • Experimental variability/precision should be reported by standard deviation (SD) (strongly preferable), confidence intervals (CI). Standard error of the mean (SEM) may be acceptable when combined with visual scatter plots. 10

1. Experimental uncertainty, statistics Data can be plotted to be misleading…… Motulsky, 2014. J. Pharmacol. Exp. Ther. 351:200 11

1. Experimental uncertainty, statistics Instructions to Authors: Animal studies • State whether or not animals and/or samples analyzed were randomized, and indicate how. • State whether studies were blinded or not. If so, indicate the method of blinding. • State whether any data were excluded. If so, indicate the reason and/or criteria for exclusion. 12

1. Experimental uncertainty, statistics http://xkcd.com/882 CC BY-NC 2.5 license. 13

2. Graphical presentation of data Mean ± SEM bar graphs can be misleading…….. Weissgerber et al., 2015. PLOS Biology 13: e1002128 14

2. Graphical presentation of data Mean ± SEM bar graphs can be misleading Weissgerber et al., 2015. PLOS Biology 13: e1002128 15

2. Graphical presentation of data Instructions to authors: Scatter plots strongly recommended for small data • sets; use box and whisker plots for large data sets, if necessary. 16

3. Western blots - antibodies Instructions to authors: • Define species of origin and source of all antibodies used, including catalog/lot numbers. • Describe how novel antibodies were generated, including preparation/purification of epitope/antigen. • Describe data supporting antibody specificity, including post- translational modifications or neoepitopes. • If possible, demonstrate loss of immunoreactivity following genetic or other molecular modification to the antigen. 17

3. Western blots - presentation Instructions to authors: • Immunoblots should be cropped in a way that retains information about antigen size, antibody specificity. • Cropped images should ideally include positions of molecular weight markers above AND below the band(s) of interest. • Avoid assembling figures of blots by splicing lanes from different sections of a gel. If blots must be spliced, borders must be clearly marked and explained in the figure legend. 18

3. Western blots - presentation Keep original data and unprocessed scans! 19

3. Western blots – presentation 20

3. Western blots - quantitation Instructions to authors: • Explain how data were obtained, linearity of signal intensity with antigen loading, and how protein loading was normalized. Some detection methods (e.g., ECL) have a very limited linear range. • Strongly prefer normalizing signal intensity to total protein loading (assessed by staining membranes for total protein). “House-keeping” proteins should not be used for normalization without evidence that manipulations do not affect expression. • Phospho-specific antibody signals should be normalized to total levels of the target protein. Supported by a large literature making the same point. 21

3. Western blots Instructions to authors: During the editorial review process, authors may be asked to provide high resolution images of original immunoblots, quantification details, and antibody validation data. Keep your original data and the unprocessed scans! 22

Implementation 1. Clearly define expectations/policies 2. Author education 3. Consistency in reviews (by EBMs!) 23

Implementation – an author’s response “The revision process provided us an opportunity to learn more about using total protein as a loading control, and we are glad to learn that the total protein staining of the Western blots is a superior and simpler method. We will use this method in our ongoing and future experiments/projects.” 24