Question: Tversky' alpha and beta should not be int, should they?
0
2.4 years ago by
stephan.struckmann0 wrote:

Hello,

thank you for your great ChemmineR toolkit. Alas, I have to report a potential bug. I am using version ChemmineR_2.24.2, but I compared with http://bioconductor.org/packages/devel/bioc/src/contrib/ChemmineR_2.27.1.tar.gz and the relevant parts seem to be the same still. If I call fpSim with method="Tversky", alpha=0.5, beta=0.5), which should according to https://en.wikipedia.org/wiki/Tversky_index be the same as the Dice-Method, I receive a matrix out of zeros. Reading the code, I found, that fpSim from sim.R calls .Call("similarity",x,y,method,addone,alpha,beta). I think this is implemented in fingerprints.cc, and there I find two lines:

int alpha= t2(0);

int beta = t3(0);

These should be double values, shouldn't they? Sorry again for bringing bad news, but maybe, you can check that in just in case I am right, maybe you may fix that?

Thank you very much

Stephan

chemminer • 348 views