Data analysis

Moderator: Bruce KRISTAL (Brigham and Women’s Hospital, Boston)

Background:

In nearly any analysis of large datasets, such as those being generated in nutritional metabolomics, it becomes desirable to extract as much meaning as possible. In addition to a suitable design of the original study, this also relates to learning how to design data analysis experiments optimally for the different types of data and datasets of different structures.

Goal:

Begin to lay down the beginnings of a "best practice" document to help people run and interpret their own experiments, and to serve as more discerning referees to enhance the overall quality of data analysis in the field.

Approach:

Presentation and discussion of the following issues:

• Commonly used algorithms.
• Choice of one or more algorithms that are well suited to solving the problems at hand.
• Commonly used data scaling methods.
• Common fundamental mistakes and misunderstandings and how to avoid these.
• Use of background information such as the past century of biochemical pathway analysis for data analysis.

Result:

First materials to create a “best practice” document to help people run and interpret their own metabolomics experiments.

Data analysis Introduction Kristal.ppt

1233 KB