Protein sequence map (pLM)
763 unique proteins across
303 natural product records (4541 protein–record mentions),
embedded with facebook/esm2_t33_650M_UR50D (revision 08e4846e537177426273712802403f7ba8261b6c,
1280-d mean-pooled residue representations) and projected to 2D with
PaCMAP (n_neighbors=10, mean-centred cosine prep).
Positions depend only on amino-acid sequence via the protein language model
facebook/esm2_t33_650M_UR50D. Biosynthetic filing pathways, enzymatic vs.
target roles, and producer taxonomy are layered on as colors and filters
without influencing the coordinates.