Improving protein secretion in yeasts: From secretion leader engineering and AI-driven prediction to bioprocess performance.

Yeast is widely used for recombinant protein production because it supports eukaryotic protein expression and high-cell-density cultivation. Efficient secretion depends on secretion leaders, short N-terminal sequences that direct nascent polypeptides into the secretory pathway. Multiple engineering strategies have been developed to enhance secretion efficiency. These include rational modification of physicochemical properties of secretion leaders, hybrid leader design through the combination of distinct signal elements, codon optimization of the leader sequence, and fusion with translational fusion partners. In parallel, artificial intelligence-based prediction tools enable high-throughput identification of signal peptides, cleavage sites, and domain organization, facilitating systematic screening and data-driven optimization. However, a leader that performs well in small-scale screening does not necessarily retain its advantage at the fermenter scale, where host secretory capacity, metabolic burden, scale-up dynamics, and downstream processing economics jointly determine the final outcome. This review summarizes recent advances in secretion leader engineering and artificial intelligence-driven prediction tools and connects these molecular strategies to bioprocess performance under industrial conditions. Current limitations and future perspectives for developing robust and broadly applicable secretion leaders in yeast expression systems are also discussed.