Probably, the only drawback that one could find in Python, which is not due to personal preferences, is its execution speed. Typically, Python is slower than its compiled brothers. The standard implementation of Python produces, when you run an application, a compiled version of the source code called byte code (with the extension .pyc), which is then run by the Python interpreter. The advantage of this approach is portability, which we pay for with a slowdown due to the fact that Python is not compiled down to machine level as are other languages.

However, Python speed is rarely a problem today, hence its wide use regardless of this suboptimal feature. What happens is that, in real life, hardware cost is no longer a problem, and usually it's easy enough to gain speed by parallelizing tasks. Moreover, many programs spend a great proportion of the time waiting for IO operations to complete; therefore, the raw execution speed is often a secondary factor to the overall performance. When it comes to number crunching though, one can switch to faster Python implementations, such as PyPy, which provides an average five-fold speedup by implementing advanced compilation techniques (check http://pypy.org/ for reference).

When doing data science, you'll most likely find that the libraries that you use with Python, such as Pandas and NumPy, achieve native speed due to the way they are implemented.

If that wasn't a good-enough argument, you can always consider that Python has been used to drive the backend of services such as Spotify and Instagram, where performance is a concern. Nonetheless, Python has done its job perfectly adequately.