If for significance you intend 'statistical significance' you need a population of reference, otherwise your question cannot be solved. Thus you can devise a sample of molecules (e.g. a set of drugs taken at random or with similar biological activities between your drug pair, the choice of the two strategies depends upon the scale of analysis you are interested in) of reference and compute the pairwise Tanimoto coefficientsbetween each drug of the refeence sample and the drug you are interested into. If 0.65 is outside 3 standard deviations of the tanimoto coefficients distribution of the reference population your value is significant.

This is of utmost importance: when you want to use statistics you must have a population of reference...

Thank you , a good tip