(assumed stable/definable character of data)
|LOWESS||Relationship between ratio and signal intensity
→ not universal, but only unique to the pair of data
|Alteration of the ratio with arbitrary functions
→ will make distortions, lead to the loss of objectivity
|SuperNORM||Statistical data distribution
→ universally found
|Three parameter lognormal model based on the universal character
→ No distortion, ensuring objectivity
LOWESS sometimes gives stable logratios that are not signal intensity dependent. However, the reproducibility in ratio determination is poor.
SuperNORM not only gives stable logratios, but finds out and cancels noise-affected data ranges. With noise-affected ranges determined, false-positive data could be identified. Data in the model-consistent range (dots in black) show high reproducibility in ratio determination.
SuperNORM can be applied to dChip data which could be generated with PM-MM or PM only. In terms of reproducibility in ratio determination, SuperNORM will give better results than dChip.