cgreene
commented
5 years ago
My basic thought is that it might be helpful for me to summarize here
the older work Paul and I did . For example, we propose an edge prior
in the paper you mentioned earlier
(https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0017258).
The real idea of making predictions using the "global prior" of node
degrees or multifunctionality is to approximate the concept of a rank
1 edge prior (which we explicitly use within that first paper and call
the IPN for Individual Property Network). Then much of our subsequent
work is about understanding exceptions from this principle. The GBA
is the exception paper
(https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1002444)
is about showing that non-prior performance is often not
generalizable. And then our CAFA and several other papers
(https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-14-S3-S15,
https://academic.oup.com/bioinformatics/article/29/4/476/199838,
https://www.sciencedirect.com/science/article/pii/S1874391914000384)
are about nailing down that it is often literature bias or
circularity, particularly in PPI data. This is part of why I am such
a fan of expression data. Between Paul and I, we have a lot of other
papers on the topic. While that's certainly not the last word, I
think that work lays out a lot of ideas that would be useful for you
to discuss in your introduction, because you are operating in much the
same territory. Other related work is not quite so closely related,
I'd guess (Marcotte, Friedberg, Dessimoz, Moreau, Khatri), but our
intros cover some of the literature, at least as of a few years ago.
We should probably write a new review at some point.☝️ comment by @gillislab via email - posted with permission
zietzm
PDF for reading convenience manuscript.pdf