Hi all,

I want use arrayQualityMetrics for outlier detection and have two questions:

1- I will select only some of GSM of a GSE for my analysis.  Should I load all GSM for QC or only selected ones?

2- The arrayQualityMetrics report result of 6 different method for outlier:

- Distances between arrays
- Boxplots
- Relative Log Expression (RLE)
- Normalized Unscaled Standard Error (NUSE)
- MA plots
- Spatial distribution of M

what is best criteria for mark a sample as outlier? (any method, three of them, ...)

Thanks. 

Dear Horcsct

1. It shouldn't really make a difference. If it does, you'll need to dig in, find out why, and proceed from there.

2. One person's outlier is the other person's Nobel prize. The reason why we provided those different criteria, and not one single summary, is that outlier detection is subjective and context-dependent. So you'll need to make that choice based on your specific question.

Wolfgang