Related Experiment Video
Updated: Mar 26, 2026
09:38
Single-throughput Complementary High-resolution Analytical Techniques for Characterizing Complex Natural Organic Matter Mixtures
Published on: January 7, 2019
9.3K
Random sample consensus combined with partial least squares regression (RANSAC-PLS) for microbial metabolomics data
Shao Thing Teoh1, Miki Kitamura2, Yasumune Nakayama3
1Department of Biotechnology, Graduate School of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan.
Journal of Bioscience and Bioengineering
|February 11, 2016
Summary
This study introduces a novel RANSAC-PLS method for strain engineering, improving metabolite identification for enhanced phenotypes. This approach effectively uncovers unique metabolic correlations missed by traditional methods.
Area of Science:
- Metabolic Engineering
- Systems Biology
- Data Science
Background:
- High-throughput omics technologies enable strain engineering by identifying gene targets from omic data.
- Metabolomics is crucial for linking cellular metabolism to phenotypes.
- Traditional methods like Partial Least Squares (PLS) regression struggle with complex datasets containing outliers or conflicting trends.
Purpose of the Study:
- To develop a robust data-mining strategy for identifying metabolite-phenotype correlations in complex biological systems.
- To improve the accuracy of regression models in metabolomics by addressing data inconsistencies.
- To discover novel metabolic targets for enhancing strain phenotypes, such as stress tolerance.
Main Methods:
- A novel data-mining strategy combining Random Sample Consensus (RANSAC) with Partial Least Squares (PLS) regression was developed (RANSAC-PLS).
- RANSAC was used to identify subsets of samples with consistent metabolite-phenotype trends.
- The RANSAC-PLS model was applied to gas chromatography/mass spectrometry metabolomics data from yeast strains to predict 1-butanol tolerance.
Main Results:
- The RANSAC-PLS approach identified previously unrecognized metabolites correlated with 1-butanol tolerance in specific sample subsets.
- These findings were validated using single-deletion strains of the corresponding metabolic genes.
- The RANSAC-PLS method demonstrated superior performance over traditional PLS modeling in detecting unique metabolic insights.
Conclusions:
- RANSAC-PLS is a powerful and promising strategy for robustly identifying unique metabolites linked to specific phenotypes.
- This method enhances phenotype improvement strategies by revealing metabolic correlations not detectable with conventional techniques.
- The findings provide valuable insights for optimizing yeast strains for industrial applications, such as improved tolerance to chemicals like 1-butanol.
