Human secretory signal peptide description by hidden Markov model and generation of a strong artificial signal peptide for secreted protein expression

Abstract

A hidden Markov model (HMM) has been used to describe, predict, identify, and generate secretory signal peptide sequences. The relative strengths of artificial secretory signals emitted from the human signal peptide HMM (SP-HMM) correlate with their HMM bit scores as determined by their effectiveness to direct alkaline phosphatase secretion. The nature of the signal strength is in effect the closeness to the consensus. The HMM bit score of 8 is experimentally determined to be the threshold for discriminating signal sequences from non-secretory ones. An artificial SP-HMM generated signal sequence of the maximum model bit score (HMM + 38) was selected as an ideal human signal sequence. This signal peptide (secrecon) directs strong protein secretion and expression. We further ranked the signal strengths of the signal peptides of the known human secretory proteins by SP-HMM bit scores. The applications of high-bit scoring HMM signals in recombinant protein production and protein engineering are discussed.