Equation (2.4) of "Berry Phase Effects on Electronic Properties" by Xiao, Chang and Niu uses the identity
$$\langle\frac{\partial n}{\partial q}| k\rangle (E_n-E_k)= \langle n| \frac{\partial H_q}{\partial q} |k\rangle \quad ,$$ where $E_i$ is the energy of the eigenstate $|i\rangle$ of the crystal Hamiltonian $H_{q}=e^{-ikr}He^{ikr}$ corresponding the crystal momentum vector $\mathbf{k}$. I have tried to reverse engineer this equation starting from the left hand side and have gotten as far as proving that
$$\frac{\partial H}{\partial q} = H^* \frac{\partial}{\partial q^*} -\frac{\partial}{\partial q*} H\quad ,$$
where I have used $*$ to denote the Hermitian conjugate. Then I made the assumption that $H$ and $q$ should be self-adjoint such that their eigenvalues are real and found that this required the commutator of $H$ and the derivative of $q$ to be
$$[H, \frac{\partial}{\partial q}] = \frac{1}{2} \quad ,$$
but I'm struggling to justify this claim. Is there another way to prove this identity / a reason the commutator takes this value?