Deep learning driven biosynthetic pathways navigation for natural products with BioNavi-NP

Area of Science:

• Biochemistry
• Computational Chemistry
• Bioinformatics

Background:

• Most natural product (NP) biosynthetic pathways remain unelucidated.
• Accurate prediction of these pathways is crucial for NP discovery and engineering.

Purpose of the Study:

• To develop a user-friendly toolkit, BioNavi-NP, for predicting biosynthetic pathways of NPs and NP-like compounds.
• To enable efficient and accurate bio-retrosynthesis predictions.

Main Methods:

• A single-step bio-retrosynthesis prediction model was developed using end-to-end transformer neural networks.
• An AND-OR tree-based planning algorithm was employed for iterative multi-step pathway sampling.

Main Results:

• BioNavi-NP successfully identified biosynthetic pathways for 90.2% of test compounds.
• It recovered reported building blocks for 72.8% of compounds, significantly outperforming existing methods.
• The toolkit demonstrated accuracy in predicting pathways for complex NPs from recent literature.

Conclusions:

• BioNavi-NP provides a valuable computational tool for predicting NP biosynthetic pathways.
• The freely available toolkit facilitates NP elucidation and reconstruction, advancing natural product research.