do Hierarchical Clustering with any statistical soft or package -> visualize the hierarchy of features (for multidimensional data) -> look for the level of separable groups...

In order to find statistically significant clusters you will also need any statistical soft/package that is capable of estimating statistical significance with p-values... I'm not sure that at the moment scikit-learn estimates statistical significance, but examine its documentation for such opportunities in estimation or find another soft/package that can do this (estimate p-values)... Will be appreciated if you can share it here - at a glance I see one in R-language - package sigclust