Utility To Find The Shortest Distance (Publication-Wise) Between Two Bioinformaticians?

Question

[Egon Willighagen]

Not bioinformatics in itself, but I am running increasingly often to find the shortest path between to bioinformaticians, in terms of publications. If they are co-author on the same paper, their distance is zero. If they they are not found as co-author on a single paper, but both co-author a publication with a third author, then the index becomes one, etc. This idea is know as the Erdős index. Basically, I would like to calculate such index for any author, such as Lindenbaum (or should I exemplify Saunders now? :).

Despite the fun character of this index, my use case is very serious. I am editing a BMC Thematic Series, and need to know how 'close' a potential referee is to the authors of a paper.

[Chris Evelo]

I was wondering how high you think the index for a good reviewer should be. You probably think 0 is too low (i.e. they should not have co-authored a previous papers. But I would guess that if it is larger than 3 it is probably not really an expert on the topic. Also you might want to consider to exclude so called conceptual papers (where ideas or organizations are announced) or papers about standards from your analysis, since for these larger groups of authors are often sought that do not necessarily really collaborate on anything else.

[Egon Willighagen]

Good points. Note that I do in fact review papers of others with index = 0. We're all grown up, and can objectively review work my befriended researchers, as long it is not work I collaborated on or advised for already.

Answer 1

[Simon Cockell]

Without something like ORCID, this is never going to be totally possible.

Retrieving all papers by author A and author B, then walking through their co-authors is not conceptually difficult, but how can I be sure that the Cockell, SJ who publishes (occasionally) about bioinformatics is the same person as the Cockell, SJ who writes about eating disorders. (Definitely not the same person by the way). It isn't as though my name is exactly common either.

[Egon Willighagen]

True, but ORCID is not here yet, and cannot help out right now. Moreover, the results returned by the software are just a starting point, not facts carved in stone. My use case is so specific, and the number of papers I edit right now, that such manual verification is no problem.

[Rm]

Biomedexperts (BME) shows graphic scientific network(s) of co-authors based on their publications. http://www.biomedexperts.com/Portal.aspx

Answer 2

[Pierre Lindenbaum]

Hi Egon,

I played with this kind of graph on my blog using the neo4j API. But here, I only used a small set of articles....

[Egon Willighagen]

So, we only need to hook this up to Mendeley then?

Answer 3

[Casey Bergman]

If your ultimate goal is to address whether two bioinformaticians are or have been collaborators, irrespective of whether they have published together (yet), you may want to try the distance function in the computational biology academic tree.