Generative AI designs functional thiolation domains for reprogramming non-ribosomal peptide synthetases

This record contains the supplementary datasets underlying the study "Generative AI designs functional thiolation domains for reprogramming non-ribosomal peptide synthetases." Pretrained generative models (ESM3, ProteinMPNN and EvoDiff) were combined with iterative design–build–test–learn cycles to generate and experimentally evaluate de novo thiolation (T) domains for non-ribosomal peptide synthetases (NRPSs), testing 578 recombinant NRPS variants in vivo across minimal, full-length and hybrid assembly lines.

The record includes two files: Supplementary_data_1_0303.xlsx — sequences and identifiers of all AI-designed T-domains; processed per-variant production data across design rounds; RapidFire LC–MS screening results (including minor products); residue-level molecular dynamics contact and hydrogen-bond interaction statistics for each catalytic state and replicate; and the T-domain sequences used for the phylogenetic analysis.