The document discusses normalization methods for reducing analytical variance and addressing batch effects in replicated measurements. It covers topics such as sample normalization, variable transformation, and visualizations of batch effects. The analysis highlights that while normalization can reduce analytical variance and batch effects, it is crucial to determine the impact of different normalization techniques on the data.

1. Biology
Chemistry
Informatics
Normalization methods for
analytical variance reduction
Statistics
Goals: Evaluate batch effects in replicated
measurements and overview normalization methods
Topics:
1. Batch effects
2. Sample normalization
3. Variable transformation
4. Variable normalization

2. Biology
Chemistry
Informatics
Identify the effects of sample
drying
Statistics
Use DATA: Normalization Data.csv
Visualize:
1. Batch effects in replicated measurements
2. The effect of normalization on samples and variables
Questions:
1. How can batch and outlier effects be mitigated?

3. Biology
Chemistry
Informatics
Question:
Statistics
Are there any batch effects in this data?

4. Biology
Chemistry
Informatics
Answer:
Statistics
Are there any batch effects in this data? (Yes!)

5. Biology
Chemistry
Informatics
Answer:
Statistics
Not all metabolites are similarly affected

6. Biology
Chemistry
Informatics
Question:
Statistics
Can sample and variable normalizations reduce
analytical variance and batch effects?

7. Biology
Chemistry
Informatics
Answer:
Statistics
Sample and variable normalizations can reduce
or increase analytical variance and batch
effects.
Sum
raw normalization

8. Biology
Chemistry
Informatics
Answer:
Statistics
Sample and variable normalizations can reduce
analytical variance and batch effects.
Sum
normalization
raw

9. Biology
Chemistry
Informatics
QC sample based normalizations
Statistics
Can be useful for estimating and removing
analytical variance

10. Biology
Chemistry
Informatics
Example: qcLOESS workflow
Statistics

11. Biology
Chemistry
Informatics
Overview of Normalizations (PCA)
Raw (%RSD = 13)
qcISTD (9)
LOESS (12)
qcISTD +
LOESS (8)
Only LOESS included
normalizations effectively
remove analytical batch
effects