I think it's not possible, because mgc coefficients are used to model the MLSA filter as well as the F0 parameter models the source. This filter is responsable to reconstruct the sound in synthesis step. I recomend you read papers to understand better: