Fig. 1: The motivation and overview of BioNavi-NP.
From: Deep learning driven biosynthetic pathways navigation for natural products with BioNavi-NP
a The vast natural products and rare biosynthetic pathways reported to date. Natural products were collected from DNP1 and visualized by TMAP73 (left). Biosynthetic reactions were collected from MetaCyc5, KEGG6 and MetaNetX7, and the network was visualized by Cytoscape74 (right). The structures were represented by the nodes and similar structures converged. The edges and arrows in the biosynthetic network represent the structural transformation. Fatty acids and others from the AA/MA pathway were colored yellow. Terpenoids and steroids from the MVA/MEP pathway were colored blue. Flavonoids and others from the CA/SA were colored red. Alkaloids and others from the AAs pathway were colored green. Others, such as nucleic acids and some hybrid-origin compounds, were colored black. b The protocol of BioNavi-NP to explore biosynthetic pathways of target natural product. We trained the transformer neural networks by combining biosynthetic and organic reactions, and four models trained with different hyperparameters form the ensemble model, which was finally used to make the single-step prediction (see details in Methods, Supplementary Figs. 1 and 2).
