Access by using the .loc indexer. : represents a slice that returns all rows.

In [2]:
df.loc[:, ["a", "c"]]
Out[2]:
a c
1 4.0 10.0
2 5.0 11.0
3 NaN NaN

Search columns based on items, like or regex.

In [3]:
df.filter(items=['a', 'c'])
Out[3]:
a c
1 4.0 10.0
2 5.0 11.0
3 NaN NaN

Select columns by data type.

In [4]:
df.select_dtypes(include=['float'])
Out[4]:
a c
1 4.0 10.0
2 5.0 11.0
3 NaN NaN