explore_path GUI shows cos(theta) but ccmap uses pearson correlation

[alexvpickering]

From email with Nasir (why I switches from cosine to pearson):

Pearson correlation actually performed a little better than cosine similarity (see attached plots). As you can see from the plot, the improvement is mainly for queries where the expected drug ranked relatively poorly with the queried signature (below top ~20%). That being said, Pearson correlation is about twice as fast as cosine similarity so I will move towards that soon.