[munificent]

An interesting follow-up would be to run PageRank on the citation graph. That should lower the importance of the scientific methods papers since they are likely to be cited by a very large number of random papers of limited importance, which bumping up papers that have led to further work that is itself important.

[tokai]

Citation count and co-authorship apparently does a better job at pointing out good papers. Even though PageRank was in part inspired by citations studies. http://link.springer.com/article/10.1007%2Fs11192-007-1908-4 (Paywalled)