Gunnar von Heijne presents a new method for predicting the cleavage sites of signal sequences in proteins, which is based on a weight-matrix approach. The method aims to improve the accuracy of predicting the cleavage sites in both prokaryotic and eukaryotic proteins, which are typically only known from their DNA sequences. The predictive accuracy of the new method is estimated to be around 75-80%, significantly higher than the 65% and 45% accuracy of the previous method for prokaryotic and eukaryotic proteins, respectively. The method involves constructing weight-matrices from observed amino acid counts in known signal sequences, with a focus on the N-terminal region (-3, -1) and the C-terminal region. The most probable cleavage site is identified by scanning the sequence with these weight-matrices and summing the weights for each position. The study uses a large dataset of 161 eukaryotic and 36 prokaryotic non-homologous signal sequences with known cleavage sites. The results show that the new method correctly identifies 78% of eukaryotic and 89% of prokaryotic cleavage sites, with a standard deviation of about ±10%. The method also allows for efficient discrimination between putative signal sequences and the N-terminal regions of cytosolic proteins. The study concludes that the new method represents a significant improvement over previous methods in terms of accuracy and practicality.Gunnar von Heijne presents a new method for predicting the cleavage sites of signal sequences in proteins, which is based on a weight-matrix approach. The method aims to improve the accuracy of predicting the cleavage sites in both prokaryotic and eukaryotic proteins, which are typically only known from their DNA sequences. The predictive accuracy of the new method is estimated to be around 75-80%, significantly higher than the 65% and 45% accuracy of the previous method for prokaryotic and eukaryotic proteins, respectively. The method involves constructing weight-matrices from observed amino acid counts in known signal sequences, with a focus on the N-terminal region (-3, -1) and the C-terminal region. The most probable cleavage site is identified by scanning the sequence with these weight-matrices and summing the weights for each position. The study uses a large dataset of 161 eukaryotic and 36 prokaryotic non-homologous signal sequences with known cleavage sites. The results show that the new method correctly identifies 78% of eukaryotic and 89% of prokaryotic cleavage sites, with a standard deviation of about ±10%. The method also allows for efficient discrimination between putative signal sequences and the N-terminal regions of cytosolic proteins. The study concludes that the new method represents a significant improvement over previous methods in terms of accuracy and practicality.