TL;DR
MultiAlign is a new, free software tool that aligns and clusters features across multiple liquid chromatography-mass spectrometry (LC-MS) datasets. This enables comparative analysis of complex biological samples, particularly in proteomics and metabolomics, even when genomic context or reference databases are limited.
What Problem Does This Paper Address?
Analyzing complex biological samples using LC-MS/MS generates vast amounts of data. A major challenge is comparing these datasets to identify differences and similarities in the proteome or metabolome, especially when dealing with samples from environments with little to no genomic information (e.g., microbial communities). Traditional methods often rely on extensive reference databases, which are frequently incomplete or unavailable for such samples. This leads to a significant number of unidentified features and hinders the ability to capture dynamic changes within biological systems. The paper addresses the need for a robust tool that can align and compare multiple LC-MS datasets, facilitating the discovery of differentially abundant analytes and providing insights into biological processes even in the absence of comprehensive genomic context.
Methods
MultiAlign tackles this challenge through a multi-dataset alignment and clustering approach. The core methodology involves:
- LC-MS Feature Finding: Identifying and deisotoping peaks in LC-MS spectra to extract monoisotopic mass, charge, and elution time features.
- LC-MS Feature Alignment: Using a dynamic time warping algorithm (LCMSWarp) to correct for systematic variations in elution time and mass measurement accuracy across datasets. This allows for meaningful comparison of features.
- LC-MS Feature Clustering: Grouping similar LC-MS features across multiple datasets using average or single linkage clustering. This identifies features that are consistently present or changing across different experimental conditions.
- Peak Matching (with Database): When a reference database (like an AMT tag database for peptides) is available, MultiAlign can match identified clusters to known analytes, enabling peptide or metabolite identification. It uses the STAC algorithm to quantify the confidence of these matches.
- Spectral Clustering (without Database): For situations where databases are absent, MultiAlign can cluster MS/MS spectra associated with LC-MS features. This helps characterize proteome structure and identify similarities between samples, even without prior knowledge of the analytes.
A key innovation is the ability to perform dataset-to-dataset alignment and clustering, which is crucial for comparative studies where a reference database is not the primary focus or is unavailable.
Datasets
The paper evaluates MultiAlign using three types of datasets:
- Environmental Proteomics Datasets: Two large LC-MS/MS datasets from microbial communities were used. One dataset had a known metagenome and an established reference peptide database, allowing for comparison with existing tools like VIPER. The second dataset was from a microbial community without a metagenome, highlighting MultiAlign's utility in data-driven discovery.
- Metabolomics Datasets: Datasets from a study on human hepatocarcinoma cells with altered fatty acid oxidation enzymes were used to demonstrate MultiAlign's applicability to metabolomics.
- Bioreactor Community Proteomics Datasets: Datasets from an anaerobic microbial community degrading cellulose were used to showcase MultiAlign's ability to analyze complex, dynamic systems without a genomic context.
Key Insights and Ideas
- Database-Agnostic Analysis: MultiAlign's strength lies in its ability to perform comparative analysis and identify patterns without relying on a complete reference database. This is revolutionary for environmental and other "omics" studies where such databases are often lacking.
- Traceability of Features: The software allows researchers to trace clusters of features across multiple datasets and link them back to specific MS/MS spectra. This provides a powerful way to investigate unidentified features and guide further experimental validation.
- Unified Platform for Proteomics and Metabolomics: The tool is designed to handle both proteomics and metabolomics data, offering a versatile solution for various omics research questions.
- Revealing Dynamic Biological Processes: By analyzing multiple samples from a dynamic system (like the bioreactor community), MultiAlign can reveal significant differences in proteome or metabolome profiles, suggesting underlying biological changes that might be missed by single-sample analyses.
Key Results
- Comparative Performance: MultiAlign's results for proteomics datasets compared favorably to existing tools like VIPER, with the added benefit of tracing peptide clusters across conditions.
- Unidentified Features: In the environmental proteomics study, MultiAlign identified a significant number of features (approximately 75%) that were unidentifiable through traditional database search techniques, highlighting its capability to discover novel analytes.
- Cluster Analysis: In the bioreactor community study, MultiAlign identified numerous clusters present across multiple time points, indicating dynamic changes in the community's proteome.
- Metabolomics Utility: The application to hepatocarcinoma cell data demonstrated that MultiAlign could effectively align and cluster LC-MS features, revealing patterns in metabolomic profiles.
Limitations and Future Work
- Ambiguity in Identification: While MultiAlign can identify features when a database is available, a significant portion of clusters remained unidentified (e.g., 86% in one proteomics study), indicating the ongoing challenge of comprehensive analyte identification.
- Computational Resources: While efficient, processing large datasets with multiple alignment and clustering steps can still be computationally intensive.
- Future Work: The authors suggest that the software's ability to discover features across conditions without a database is particularly useful for supporting or establishing hypotheses for cause/effect, spatial, or temporal studies. The traceback capability also supports targeted workflows for further experimental validation, such as selected reaction monitoring or additional MS/MS data collection. The tool's applicability to other LC-MS-based omics datasets is also highlighted.
Summary updated