Why is a single base-pair deletion mutation within a protein-coding sequence more likely to be harmful than a single base-pair change mutation?