Remove an item from a list contained in another item in python

[Gribouillis]

>>> l = [[2,3,4,5],[2,3],[7,8],[2],[4,5,6]]
>>> S = sorted((set(x) for x in l), key=len, reverse=True)
>>> S
[{2, 3, 4, 5}, {4, 5, 6}, {2, 3}, {8, 7}, {2}]
>>> roots = []
>>> for s in S:
...     if not any(s.issubset(r) for r in roots):
...         roots.append(s)
... 
>>> roots
[{2, 3, 4, 5}, {4, 5, 6}, {8, 7}]

As supuflounder said, this is a O(n^2) algorithm, which means here that if the initial list has length N, we way call issubset() about N*(N-1)/2 times.