On the origin of the boson peak in globular proteins