PERISCOPE-Opt

PERISCOPE-Opt predicts recombinant protein expression yields and optimal fermentation parameters by integrating amino acid sequence and fermentation process features using machine-learning models.

Key Features:

• Integration of Amino Acid Sequence and Fermentation Conditions: Integrates features derived from amino acid sequences and fermentation process conditions to inform predictions of expression yields and fermentation parameters.
• Classification Stage: Utilizes XGBoost classifiers to categorize expression levels into three classes: high (>50 mg/L), medium (0.5–50 mg/L), and low (<0.5 mg/L).
• Regression Stage: Employs regression models, including support vector machines and random forest algorithms, to predict specific expression yields corresponding to each classified level.
• Performance Metrics: Independent testing reported an overall average accuracy of 75% and a Pearson correlation coefficient of 0.91 for correctly classified instances.
• Reduction of Experimental Effort: Provides predictions of optimal fermentation conditions and yields to reduce resource‑intensive trial‑and‑error experimentation in recombinant protein production (RPP).

Scientific Applications:

• Recombinant Protein Production (RPP) in E. coli: Supports optimization of fermentation conditions and expression yields for recombinant proteins produced in E. coli systems.
• Biotechnology and Pharmaceuticals: Assists in maximizing protein yields and process optimization in biotechnology and pharmaceutical development.
• Industrial Microbiology: Enables optimization of protein expression levels in industrial microbiology applications where precise control over expression is crucial.

Methodology:

Integrates features from amino acid sequences and fermentation process conditions, uses XGBoost classifiers to assign expression level classes (high >50 mg/L, medium 0.5–50 mg/L, low <0.5 mg/L), and applies regression models including support vector machines and random forest to predict yields within each class.