SANDPUMA

SANDPUMA predicts substrate specificities of nonribosomal peptide synthetase adenylation (A) domains from DNA sequences to support discovery and characterization of nonribosomally synthesized peptides (NRPs).

Key Features:

• Ensemble algorithm: Integrates multiple high-performing algorithms into an ensemble to improve accuracy of A-domain substrate predictions.
• prediCAT phylogenetics-inspired estimation: Incorporates prediCAT to quantitatively estimate the predictability of individual A-domains using phylogenetics-inspired principles.
• Benchmarking on independent test set: Evaluated against an independent test set of 434 A-domain sequences, demonstrating that active-site-motif-focused algorithms outperform whole-domain methods.
• antiSMASH integration: Integrates with the antiSMASH biosynthetic gene cluster analysis pipeline to combine genomic cluster annotation with A-domain substrate prediction.

Scientific Applications:

• NRP chemical diversity exploration: Applied to 7635 Actinobacteria genomes to reveal greater nonribosomal peptide (NRP) diversity than previously estimated.
• Natural product discovery and biotechnology: Guides identification of candidate A-domains and predicted substrates to inform natural product discovery and biotechnological exploitation.

Methodology:

High-performing algorithms were retrained on a comprehensive dataset and integrated into an ensemble; prediCAT provides per-domain predictability estimates; models were benchmarked against an independent test set of 434 A-domain sequences with comparisons between active-site-motif-focused and whole-domain methods.