Improved Classification in NMR Metabolomics Data
Author Information
Author(s): Helen M Parsons, Christian Ludwig, Ulrich L Günther, Mark R Viant
Primary Institution: The University of Birmingham
Hypothesis
The glog transformation can improve classification accuracy in NMR metabolomics data compared to other scaling methods.
Conclusion
The glog and extended glog transformations significantly improve classification accuracy in NMR metabolomics datasets.
Supporting Evidence
- The glog transformation achieved 100% accuracy in classifying 1D NMR spectra of hypoxic and normoxic invertebrate muscle.
- The glog transformation also achieved 100% accuracy for 2D JRES spectra of fish livers from different rivers.
- For urine samples from dogs, the glog and autoscaling methods achieved equal highest accuracies.
Takeaway
This study shows that using a special math trick called the glog transformation helps scientists better tell different types of samples apart in their experiments.
Methodology
The study compared the glog transformation with autoscaling and Pareto scaling using PCA and LDA on three different datasets.
Potential Biases
Potential bias from technical variance in sample preparation and analysis.
Limitations
The study primarily focused on three datasets, which may limit the generalizability of the findings.
Participant Demographics
Urine samples from two dog breeds, muscle tissue from marine mussels, and liver tissue from fish.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website