Flag values of microarray data, and how to use them.

Flag values are generated by microarray systems, which indicate how the measured values are reliable or not. Each system or each flag information has unique meaning and characteristics. If you understand flag values and apply filters on them, they greatly help your analyzing data.

Affymetrix GeneChip:

If you have Affymetrix GeneChip data and it's quantified with MAS5 algorithm, ABS_CALL information may be available. Present (P) call indicates the quality of measurement is good. Absent (A) call for bad and Marginal (M) call for intermediate.

Filtering on Affymetrix Detection Values

3' IVT type of Affymetrix GeneChip system generate "Detection" values with intensities. They are flag values indicate the quality of the signals. Present (P) is good, Marginal (M) is intermediate, and Absent (A) is bad. Watch this movie to know the characteristics of Detection values, and learn how to use them.

Please take an Online Training for a full instruction of the data analysis.

Agilent Microarrays:

IsSaturated & IsFeatNonUnifOL

If these flags get 1, it means you can't trust the signal intensity. Still it's ok if 30 or less genes with 1 in them. But if there are hundreds of genes with "1" in these flags, you'd better to doubt the quality of the sample.

IsPosAndSignif & IsWellAboveBG

If these flag values are "0", it indicates the signal intensity is not distinguishable from those of negative controls. In other words, those genes seem not to be expressing.

The difference between the two flags is stringency. IsWellAboveBG assign more genes to "0", and it results in less genes pass quality filter. That's why IsWellAboveBG flag is considered as more conservative. But if you're really interested in genes at very-low-expression-level, using IsPosAndSignif may give you more chance of discovery.

Filtering on Agilent Flag Values

Agilent's microarray system outputs very useful flag information, and we recommend you leave them as they are. We think GeneSpring's way of mimicing Affy's Detection values is ridiculous. I show you why.

Please take an Online Training for a full instruction of the data analysis.