Fig. 2: Identification of taxane biosynthetic gene modules.

a, Schematic of Taxol biosynthesis and previously hypothesized gene order. Blue, previously identified Taxol biosynthesis enzymes; red, hypothesized enzymes. b, PCC between known Taxol-related genes using mpXsn data. To identify substructures, genes were hierarchically clustered (SciPy fcluster, Euclidean distance) on both axes. c, Schematic for matrix factorization. mpXsn data were factorized using cNMF²⁸. d, Heat map showing the rank of known Taxol biosynthetic genes in each of the modules produced by matrix factorization. e, As in d, but showing only the three modules enriched in Taxol genes (modules 1, 2 and 3). f, Heat map of Taxol modules, showing module rankings for the two isoprenoid pathways in the primary metabolism potentially upstream of the Taxol biosynthesis. Only the MEP pathway is co-expressed with the first Taxol module, supporting its role in synthesizing Taxol precursors. g, All gene modules ranked by the total number of 2-ODD, P450 and acetyltransferase (AcylT) genes in the top 100 genes of each module. h, Module usage of each cell, which is analogous to gene expression, plotted onto the single-nucleus transcriptomic UMAP. Taxol modules 1–3 are expressed in non-overlapping cell states, and were mainly identified in different experiments. i, Unfiltered lists of the top genes in each module, plotted as module rank and score. Blue, previously identified genes associated with Taxol biosynthesis; red, new biosynthetic genes identified in this study.