Use indexing and map to replace letters:

df.iloc[:, 2:] = df.apply(lambda x: x[2:].map(x[:2]), axis=1)
print(df)

# Output:
   A  B  i  j  y  z
0  1  2  2  1  1  1
1  2  3  3  2  3  3

Setup:

df = pd.DataFrame({'A': [1, 2], 'B': [2, 3], 'i': ['B', 'B'],
                   'j': ['A', 'A'], 'y': ['A', 'B'], 'z': ['A', 'B']})
print(df)

# Output:
   A  B  i  j  y  z
0  1  2  B  A  A  A
1  2  3  B  A  B  B

Use indexing and map to replace letters:

df.iloc[:, 2:] = df.apply(lambda x: x[2:].map(x[:2]), axis=1)
print(df)

# Output:
   A  B  i  j  y  z
0  1  2  2  1  1  1
1  2  3  3  2  3  3

Setup:

df = pd.DataFrame({'A': [1, 2], 'B': [2, 3], 'i': ['B', 'B'],
                   'j': ['A', 'A'], 'y': ['A', 'B'], 'z': ['A', 'B']})
print(df)

# Output:
   A  B  i  j  y  z
0  1  2  B  A  A  A
1  2  3  B  A  B  B

Details:

For each row, apply the following function over index axis so x contains the whole row at each iteration:

Map the value from the third column (x[2:] <- i, j, y, z) to the index from the two first columns (x[:2] <- A, B) like a dictionary (a Series can act as dictionary, check the map method)

For the first iteration:

A    1  # <- index A
B    2  # <- index B
i    B  # <- value B
j    A  # <- value A
y    A  # <- value A
z    A  # <- value A
Name: 0, dtype: object