AI-Prioritized Enzyme Variants for Sitagliptin Synthesis

An AI-driven 6D-grid protein-engineering framework integrated a dataset of 1.39 million structural fragments to predict productive enzyme mutations. Five AI-prioritized variants, each containing nine mutations, were validated at 7 L fermentation scale, achieving up to 89% conversion for sitagliptin synthesis. The dataset, shared by Pravin Kumar R on figshare in May 2026, likely contains the computational and experimental results underpinning this biocatalyst development.

Use Cases

• Training machine learning models for enzyme property prediction based on structural fragment data
• Benchmarking computational protein engineering frameworks against experimental validation results
• Analyzing the relationship between mutation combinations and catalytic stability under scale-up conditions

Strengths

• Dataset underpins a published methodology that reduced screening to five prioritized variants from a potential search space of thousands
• Experimental validation includes specific performance metrics: 51-89% conversion rates and >99% enantiomeric excess
• Workflow integrates multiple data types: interaction energy descriptors, solvent effects, and 1.39 million structural fragments

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment for large-scale ML training

Provenance

Source

figshare

Collection Method

Likely generated through computational protein design and experimental validation as described in the associated research.

Freshness

Last updated 2026-05-14 12:34:37