In this thesis a new method based on the Tight-Binding Linear Muffin Tin Orbital (TB-LMTO) formalism and the Quasiparticle Self-consistent GW (QSGW) approximation is proposed. The method is capable of generating accurate electronic bands structure of large supercells necessary to model alloys structures. The strategy consist in building simple and small hamiltonian from linear Muffin-tin-orbitals (LMTO). Parameters in this hamiltonian are then used to fit the difference in QSGW self-energies and LDA exchange-correlation potentials. The parameter are assumed to transfer to new environments --- a procedure we check carefully by comparing our predicted band to QSGW bands for small supercells. The method possess both the accuracy of the QSGW approximation, (which is the most reliable way to determine energy bands accurately, and yet too expensive for the large supercells required here), and the efficiency of the TB-LMTO method. The accurate and highly efficient hamiltonian is used to predict the electronic and optical transitions of Si1-xGex alloys and SnxSiyGe1-x-y alloys. The goal is to engineer direct band gap material compatible with the silicon technology. The results obtained are compared to available experimental data.