Selection of suitable signal peptides is an important factor for efficient secretion of heterologous proteins. We defined structural fusion degree (SFD) as the compatibility degree of target proteins and signal peptides by a bioinformatics approach. We mathematically analyzed the interaction of fused signal peptides and adjacent residues of proteins, and proposed a mathematical model of extended signal region and the protein. SFD Features was extracted from this model to characterize the secretability of heterologous proteins. Simulation tests showed that SFD features can effectively discriminate high secretory proteins from poor ones in the host Bacillus subtilis. Results from this research will be useful in signal peptide selection and have a better guiding significance for the optimization of heterologous protein secretion.