greenelab / annorxiver_manuscript Goto Github PK

There is work from @jpolka that just came out : https://www.biorxiv.org/content/10.1101/2021.02.20.432090v1

In the versions of our paper that were available before this weekend, the statements around priority of analysis based on text content have been true. At this point, we should revise things to note this recent work.

The axis labels (and maybe other text - harder to tell) here appear to be all subtly different fonts or sizes

A: Key labels should be formatted for presentation (not underscore separated). You might want to present them as bioRxiv-PMC, bioRxiv-NYTAC, etc. It also looks like you may have room to spell out Kullback–Leibler on the axis. If you don't do this, many journals will make you explain the acronym in the figure legend so it's better to spell it out if possible.

B: I think you may wish to present the log_2(OR)s. This will make them symmetric around the axes and make it less important which one is ordered first. You might also consider formatting the X axis as with Odds Ratio in the middle (at 0 of the log odds) and then an arrow to the left saying "More in PMC" and one to the right towards "More in bioRxiv". This will make it easier for someone to understand your results at a glance.

C: PMC should be capitalized for consistency.

D: See B.

E: use consistent capitalization strategies throughout.

The URL column makes these somewhat harder to look through now. We should consider removing that column.

In "Discovering Unannotated Preprint-Publication Relationships" the methods section says:

First, we used CrossRef [@doi:10.1629/uksg.233] to identify bioRxiv preprints that were linked to a corresponding published article.

This actually gets used earlier as well when you compare preprint-published versions. I think you should break this out into its own short methods subsection (saying when you obtained the links, etc). It seems like it needs to come before "Comparing Corpora" and examining the subset of linked articles should be mentioned there as well.

It would be helpful to include the precise numbers of articles with a linked publication, etc, in the new subsection.

This is a ticket to remind myself to add description for the confidence intervals and tweak Figure B axis labels.