Statistical Analysis of Metagenomic Data

Metagenomic data possess specific features that call for adapted statistical methods. Learn about the most commonly used statistical tools in this field.

Course Outline

The recommended duration for this course is 5 online session(s).

The most popular version of this course comprises 5 sessions described as follows:Session 1: Statistical Hypothesis Testing

• Accounting for sampling variation in the decision-making process
• Mechanism underlying hypothesis testing

• • Null and alternative hypotheses
  • One- and two-sided tests
  • Test statistic
  • Decision rule: critical value and p-value
  • Statistical vs. practical significance, effect size to detect
• Risk involved in hypothesis testing: confidence level and power
• Controlling risks
• General idea underlying sample size and power determination

Session 2: Understanding the Data Structure, Exploratory Data Analysis and Specific Issues in Metagenomic Data

• Specific features of metagenomics data
• Specific issues
• Differences with classical data
• Summarising data efficiently
• Multiplicity in statistical tests
  • What is multiplicity?
  • Identifying situations leading to multiplicity
  • Dealing with multiplicity - Bonferroni, Tukey, Benjamini-Hochberg, etc.

Session 3: Classical ANOVA and its Alternatives in Genomics

• Refresher on basic principles underlying ANOVA
• Interpretation of an ANOVA table

• • Significance of factors and interactions
  • Interpretation of factor effects and interactions

• Principle underlying MANOVA
• Specifics of metagenomic data and alternative methods
• Principle of ANOSIM and PERMANOVA

• • Comparison of centroids
  • Heterogeneity of dispersion across groups
  • Multiple comparisons
  • Results visualisation

Session 4: Advanced Methods and Data Visualisation

• Various distance measures in metagenomics
  • Principle & selection of an appropriate metric
• An Overview of Visualisation Tools - Ordination Methods
• • Principal Component Analysis (PCA)
  • Correspondence Analysis (CA)
  • MultiDimensional Scaling (MDS)
  • Principal Coordinates Analysis (PCoA)

Session 5: Data Pre-Processing

• Filtering & trimming
• The question of data normalisation
• Clustering

Target Audience

This course is aimed at engineers, researchers, bioinformaticians, biologists and depending on the target audience, the course is adapted. On the one hand by using examples specific to the field of application and on the other hand by presenting specific tools or applications, if applicable.