tl;dr: A universal solution is very nearly impossible, biblatex offers the bookpagination (and pagination field) and a well-coded style will make use of those, so bookpagination = {none} seems a good way to go for biblatex.
The Long Answer
A solution that works for both natbib and biblatex is almost impossible (one might say all but impossible) for the simple reason that it is not natbib or biblatex that decides what is printed and how the bibliography looks like, but the bibliography style (\bibliographystyle in "BibTeX packages"for lack of a better name [the .bst styles]; style, bibstyle with biblatex [the .bbx styles]); so a universal solution would have to work with all the styles (.bst as well as .bbx) available today - that is quite a lot of styles.
While most of the styles adhere to a general semantics and best practice, we cannot be sure that all styles will display, say, the pages field in a certain way (well, we can be quite sure with the pages field as it is so essential and standard, but the eid field is a different matter altogether).
Most of the "exotic" fields such as eid, eprint etc. are bound to be unsupported by a number of styles - even the more basic (and arguably quite un-exotic - by modern standards at least) URL field is lacking in some styles, simply because they were completed before the advent of the URI, which was standardised in 1994 (RFC1738), case in point: ieeetr.bst seems to have been the same since 1988 (save for a clarification of the licence).
Having said all that, the pages field is probably your best bet for a widely applicable solution, plus it does not seem to be too far off semantically.
But even the treatment of the pages field is not standardised or normalised across bibliography styles: Some .bst styles check whether the pages field contains multiple pages in order to decided whether to print "p." or "pp." (plain.bst and ieeetr.bst), but there is no need to do that in a German style since in both cases the abbreviation will be "S.", consequently natdin.bst does not care to check for "multiple pages", it behaves differently if a URL is present, though.
biblatex offers a lot of these functionalities out of the box, so almost all styles make use of them (that is also thanks to the fact that a lot of the biblatex styles out there are coded quite well), I have yet to come across a style that tries to implement these checks itself.
biblatex will check if the pages field actually contains a "page range", page ranges consist of characters recognised as numbers (see \DeclareNumChars) and those recognised as range separators/indicators (see \DeclareRangeChars) as well as certain other commands (see \DeclareRangeCommands and \DeclarePageCommands). If the pages field contains only characters and commands declared there, it passes the \ifpages test. If a pages field passes the \ifpages test and it is printed with \mkpageprefix, the page prefix "p."/"pp." will be added if it does not, the pages field is printed as is without a prefix. In fact, we can control the prefix added to the pages field as described in §2.3.10 Pagination, p. 35, of the biblatex documentation, we can even define our own prefix (as shown in Citing specific slides of a presentation).
The "pagination scheme" is even settable on a per-entry basis with pagination (controls how in-text citations have the postnote formatted if it passes \ifpages) and bookpagination (controls the pages and pagetotal fields).
So what you could do for these special pages fields is to add bookpagination = {none} to the corresponding entry (or hope for a character that throws \ifpages to false, the e in the PLoS entries does exactly that).
An example with bookpagination = {none}:
@article{Clune22032013,
author = {Clune, Jeff and Mouret, Jean-Baptiste and Lipson, Hod},
title = {The Evolutionary Origins of Modularity},
volume = {280},
number = {1755},
pages = {20122863},
bookpagination = {none},
date = {2013},
doi = {10.1098/rspb.2012.2863},
eprint = {http://rspb.royalsocietypublishing.org/content/280/1755/20122863.full.pdf+html},
journal = {Proceedings of the Royal Society B: Biological Sciences}
}
The MWE
\documentclass{article}
\usepackage[style=authoryear, backend=biber]{biblatex}
\usepackage{filecontents}
\usepackage{hyperref}
\begin{filecontents*}{\jobname.bib}
@article{Clune22032013,
author = {Clune, Jeff and Mouret, Jean-Baptiste and Lipson, Hod},
title = {The Evolutionary Origins of Modularity},
volume = {280},
number = {1755},
pages = {20122863},
bookpagination = {none},
date = {2013},
doi = {10.1098/rspb.2012.2863},
eprint = {http://rspb.royalsocietypublishing.org/content/280/1755/20122863.full.pdf+html},
journal = {Proceedings of the Royal Society B: Biological Sciences}
}
\end{filecontents*}
\addbibresource{\jobname.bib}
\begin{document}
\cite[3]{Clune22032013}
\printbibliography
\end{document}
yields, note that for in-text references the pages prefix is enabled (and the default page one), while the prefix for the journal pages is disabled.
biblatexit would probably be no problem to use thepagesfield and usepagination={none}in those cases. To say how a solution might work withnatbib, we would need to know what style you use, as the.bstfile decides what gets printed, and how it is going to look..bibfiles such thatbiblatexandnatbibcan access and show the field correctly. Also, I want to avoid abiblatex-only solution as switch betweenbiblatexandnatbibfor different projects. Iseidnot the right field to use? What happens whenpages={e1000112}that prevents pagination from appearing (I assume it's thee)? It seems that best option would be to use something in thepagesfield that prevents pagination.eid. But as Ulrike Fischer says (and I mentioned above), it is the.bstwho decides the output, and if the.bstfile always prints "p." in front of the pages field and allows for no mechanism to disable this behaviour, you will have to live with that. A universal solution that works with everything seems quite unlikely.biblatexdoes, but I imagine not allnatbib.bstfiles do). As such a work-around like this might work with some styles (not a lot actually, I suppose), but is useless as the universal solution you want it to be.