Pichia-CLM: A language model-based codon optimization pipeline for Komagataella phaffii

Area of Science:

• Biotechnology
• Computational Biology
• Genomics

Background:

• Synonymous codon usage bias (CUB) influences gene expression and protein production.
• Factors like host organism, gene function, and codon position affect CUB.
• Optimizing codon usage is crucial for efficient heterologous protein expression.

Purpose of the Study:

• To develop a language model-based pipeline for learning and applying codon usage bias.
• To enhance the production of heterologous proteins in the industrial host Komagataella phaffii.
• To compare the performance of the developed pipeline against existing codon optimization tools.

Main Methods:

• Utilized language models to learn codon usage patterns from the host genome.
• Developed the Pichia-Codon language model (Pichia-CLM) pipeline.
• Applied Pichia-CLM to optimize sequences for heterologous protein expression in K. phaffii.
• Evaluated protein production levels and compared results with commercial tools.

Main Results:

• Achieved up to threefold enhancement in heterologous protein production compared to native sequences.
• Pichia-CLM consistently improved productivity for proteins of varying complexity.
• Generated sequences mimicked host cell protein codon usage properties.
• Successfully learned and avoided negative cis-regulatory and repeat elements.

Conclusions:

• Language models can effectively learn codon usage patterns from genomic data.
• Pichia-CLM offers a robust and efficient method for codon optimization.
• The approach holds significant potential for improving heterologous protein production in industrial biotechnology.