Analysis of T-RFLP data using analysis of variance and ordination methods: a comparative study

Area of Science:

• Microbial ecology
• Bioinformatics
• Statistical modeling

Background:

• T-RFLP (Terminal Restriction Fragment Length Polymorphism) analysis is widely used for microbial community profiling.
• A lack of consensus exists regarding the optimal statistical methods for analyzing T-RFLP data and identifying ecological trends.
• Multivariate ordination techniques are commonly employed but their performance varies with dataset characteristics.

Purpose of the Study:

• To empirically and theoretically compare ten common ordination methods for T-RFLP data analysis.
• To assess the distribution of variation within T-RFLP datasets using ANOVA.
• To identify the most robust and informative multivariate methods for soil microbial community analysis.

Main Methods:

• Analysis of ten diverse T-RFLP datasets from soil microbial communities.
• Application of principal component analysis (PCA), nonmetric multidimensional scaling (NMS), correspondence analysis (CA), detrended correspondence analysis (DCA), and the Additive Main Effects and Multiplicative Interaction (AMMI) model.
• Utilized analysis of variance (ANOVA) to examine variation sources (Environment, T-RFs, TxE interactions).

Main Results:

• ANOVA indicated small Environment effects, large T-RF effects, and intermediate T-RFxEnvironment (TxE) interactions, with higher TxE signifying greater community dissimilarity.
• AMMI, T-RF-centered PCA, and DCA demonstrated the highest robustness, yielding consistent results across datasets.
• NMS with Sørensen and Jaccard distances showed superior sensitivity in detecting complex gradients within heterogeneous datasets.

Conclusions:

• Method selection for T-RFLP data analysis should consider dataset complexity and theoretical criteria.
• AMMI, T-RF-centered PCA, and DCA are recommended for robust ordination of soil microbial community T-RFLP data.
• Binary or relativized peak height data are suggested for exploratory analyses of soil T-RFLP data.