add description of databases to manuscript

[ahwagner]

In order to understand the relevance of the difference arising from the comparisons, it is essential to have upfront much more description about the different databases and their remit/intention. e.g PMKB has a much lower number of variants – is this because it only includes example variants (or variant types) that led to approval of a drug? Likewise, what are the obvious reasons for differences in gene and publication content between the different databases.

[ahwagner]

Initial comments: me:

My first thought is that describing the nuances of each database in detail is probably best left to references–to try to characterize and compare each of the databases against one another could easily be another paper by itself. However, some discussion/analysis about the differences in gene/publication content could be explored and added to the text.

Adam Margolin:

We should include at least a paragraph as he suggests with, e.g., just giving one sentence for each database. This is an easy comment to address, to good to give the review what he/she wants and say we’ve added his good suggestion to the text. Specifics aren’t correct, but maybe something like “In addition to differences in gene/drug associations, each database differs in the kind of information contained. The Jaxson Labs knowledge base contains the largest number of associations, but contains many genes prognostically related to cancer without associated therapies. PMKB contains the fewest associations, but contains detailed information on the specific variants linked to each drug and excludes associations with lower tiers of evidence. OncoKB includes many associations linked to gene, rather than, variant level alterations.” Also can describe differences with respect to germline vs. somatic variants between databases, or whatever, per next comment.

Then have references for each as suggested. Would be ideal to summarize in a small table or figure if the differences are easy to lay out and we think this helps the paper and would show we took reviewer’s comment seriously, but not necessary if this doesn’t show the information so well.

Mark Lawler:

Agree with Adams point - a paragraph in the text about each database with references - think our overall response to reviewers is to (mostly) accede to their requests as the majority are reasonable and will improve the paper

[ahwagner]

Added a paragraph in results describing these resources and how differences in their curation strategy and data modeling may account for differences in observed variants.