An Integrated Data-Mining Strategy via Transcriptomics and DIA Quantitative Proteomics to Unveil Chemical Attribution Signature Peptides for Tracing Origins of Castor Beans

AI Summary

Integrated RNA-seq transcriptomics and DIA quantitative proteomics across 17 origins to derive chemical attribution signatures for castor bean provenance.
A feature-screening pipeline prioritised 59 signature peptides; machine learning on these yielded 93.7% classification accuracy and revealed altitude and latitude effects.
Parallel reaction monitoring validation confirmed a minimal panel of 24 peptides as robust molecular CAS markers for forensic origin attribution.

Anal Chem. 2026 May 18. doi: 10.1021/acs.analchem.6c00346. Online ahead of print.

ABSTRACT

The increasing frequency of castor bean or ricin-induced intoxication or terror events threatens public safety and national security, making the tracing of castor bean origins critical for law enforcement and counterterrorism efforts. Chemical attribution signatures (CAS) could address this issue by providing inherent and forensic links between ricin-containing samples and their geographical origins; however, practical implementations of this approach remain scarce. Omics data sets offer substantial potential to generate comprehensive biological insights and high-dimensional data for provenance attribution, where transcriptomics and proteomics profiling are better suited to castor beans than traditional genomics and metabolomics, regarding that castor beans exhibit low genetic diversity but high phenotypic polymorphism. In this work, toward castor bean samples from 14 provinces or autonomous regions in China and three international locations, including Ethiopia, Pakistan, and South Sudan, we proposed an integrated local-global data-mining strategy by systematically integrating RNA-seq transcriptomics and data-independent acquisition quantitative proteomics, and developed a straightforward feature-screening pipeline to prioritize 59 signature peptides as provenance-related CAS with distinct interregional and international expression patterns. A subsequent machine learning model trained on these CAS achieved 93.7% classification accuracy, identifying robust discriminative patterns among samples from different global regions, as well as fine-scale differences across altitude gradients and north-south divisions in China, in which the attribution index of altitude and latitude is reported for the first time. Finally, after validation by parallel reaction monitoring in nanoLC-HRMS/MS, we confirmed a minimum signature panel of 24 peptides as molecular CAS markers for the origin attribution of castor beans.

PMID:42148881 | DOI:10.1021/acs.analchem.6c00346

Document this CPD