Recently, a research group reported that they can fit the F2 data well by using the Altarelli-Parisi evolutionequation with a valence-like initial gluon distribution, which means that we need not consider the high twisteffects in the proton. In this letter, we point out that their valence-like initial gluon distribution is unreasonable for the proton structure function. With the reasonable initial distributions, we find that the high twist corrections to the Altarelli-Parisi evolution equation is necessary.