TL;DR
This paper introduces a novel LC/MS data processing pipeline to efficiently identify and assess the novelty of secondary metabolites from natural products. By generating "Representative MS Spectra" (RMSs) and developing a "Fresh Compound Index" (FCI), the method helps researchers avoid rediscovering known compounds and accelerate the discovery of new ones.
What Problem Does This Paper Address?
The discovery of novel secondary metabolites from natural products is crucial for drug development and understanding biodiversity. However, a major bottleneck is the frequent rediscovery of already known compounds, a phenomenon known as "dereplication." While mass spectrometry (MS) offers high-throughput analysis, raw MS data is complex and requires extensive processing. Existing databases and computational tools often focus on primary metabolites or lack comprehensive coverage of the vast chemical diversity found in natural products, particularly from diverse plant sources. This gap hinders the efficient identification of truly novel compounds.
Methods
The core of this work is a new LC/MS data processing pipeline designed for data-independent acquisition (DIA) methods. The pipeline involves several key steps:
- Noise Filtering and Deisotoping: Raw MS spectra are cleaned by removing low-intensity signals and resolving isotopic patterns to focus on the most informative ions.
- Clustering and Representative MS Spectra (RMS) Generation: Consecutive MS scans that exhibit high similarity (using a modified dot-product scoring method) are grouped together to form a single RMS. This effectively condenses thousands of raw scans into hundreds of representative spectra, each presumed to correspond to a single metabolite.
- Hierarchical Clustering Analysis (HCA): RMSs from a sample are clustered based on their spectral similarity. This allows for the visualization of structural relationships between metabolites within an extract, grouping similar compounds together.
- Fresh Compound Index (FCI): A novel scoring system is introduced to quantify the novelty of an RMS. The FCI is calculated by comparing an RMS against a large in-house database of reference RMSs. It considers both dissimilarity (DI) and similarity (SI) indices to provide a numerical grade for how "new" a compound is relative to known metabolites.
Datasets
The primary dataset used for method development and validation was derived from methanolic extracts of Agrimonia pilosa roots and aerial parts.
The study also constructed a substantial in-house reference database containing 65,322 reference RMSs. These RMSs were derived from metabolites found in 466 representative Korean medicinal plants. The authors consider this collection to be representative of plants native to East Asia, aiming to capture a broad spectrum of natural product diversity.
Key Insights and Ideas
- RMSs as Metabolite Fingerprints: The concept of generating RMSs from raw MS data is powerful. It simplifies complex spectral data into manageable units that retain essential structural information, making downstream analysis more tractable.
- HCA for Structural Relationships: Hierarchical clustering of RMSs proved more effective than molecular networking (MN) for visualizing structural relationships, especially for compounds with similar scaffolds but differing fragmentation patterns. This allows for the identification of novel compounds that are structurally related to known ones.
- FCI as a Novelty Metric: The Fresh Compound Index (FCI) offers a quantitative way to assess the novelty of a metabolite. By comparing against a large, curated database of "real but unknown" metabolites from diverse plants, it provides a more biologically relevant measure of novelty than simply checking against known compound databases.
- DIA for Comprehensive Profiling: The use of data-independent acquisition (DIA) is highlighted as advantageous for capturing a wider range of metabolites without the need for prior ion selection, leading to more comprehensive spectral data.
Key Results
- The developed data-processing protocols successfully generated RMSs from thousands of raw MS scans, reducing them to hundreds of representative spectra.
- The method was applied to Agrimonia pilosa, leading to the identification of known compounds and the isolation of nine new compounds: three pilosanidin derivatives and four pilosanol-type molecules, along with two HMG-conjugated chromones.
- Hierarchical clustering analysis of RMSs effectively grouped compounds with similar chemical scaffolds, aiding in the identification of novel compounds adjacent to known ones.
- The Fresh Compound Index (FCI) successfully differentiated between novel and common compounds. For instance, pilosanidins, pilosanols, and agrimonolides (which are more specific to Agrimonia) showed higher FCIs compared to triterpenes and flavonoids, which are more widespread.
Limitations and Future Work
- Database Expansion: The authors acknowledge that the in-house database, while substantial, can be further expanded by adding more reference RMSs from diverse plant sources to improve the accuracy and coverage of novelty assessment.
- Computational Intensity: While the RMS generation and clustering are automated, the HCA and FCI calculations can still be computationally intensive, especially with very large datasets.
- "Real but Unknown" Definition: The definition of "real but unknown" relies on the current state of knowledge and the coverage of the reference database. Future discoveries might reclassify some "unknowns" as known.
The authors express optimism that this approach will facilitate the laborious isolation process and accelerate the discovery of novel secondary metabolites by guiding researchers towards compounds with higher FCIs. They plan to continue adding reference RMSs to their database.
Summary updated