The phrase "as a default normalization" is correct and usable in written English.
It can be used in contexts related to data processing, statistics, or machine learning where a standard or baseline normalization method is being referenced.
Example: "In our analysis, we will use min-max scaling as a default normalization to ensure all features are on the same scale."
Alternatives: "as a standard normalization" or "as a baseline normalization".
We used this special case to illustrate that it may not be wise to conduct differential expression blindly before making sure that a default normalization method is appropriate.
Besides using RMA as our default normalization method, we also assessed the impact of normalization on data analysis by comparing RMA results with results derived from MAS 5.0 and dChip [ 28].
As demonstrated in the current simulation results comparing two groups (for example, samples A and B), the use of default normalization methods implemented in the existing R packages performed poorly in simulations where almost all the DEGs are highly expressed in Sample A (i.e., the case of PA > > 50% when the range is defined as 50% ≤ PA ≤ 100%).
To assess the effect of normalization on the outcome of the different packages, we considered both the use of the default normalization of each package, as well as the use of similarly preprocessed data normalized with the TMM normalization [ 18].
The data were normalized using the default normalization method.
Therefore, for the rest of the results, we focus on the default normalization methods that an average user is likely to use.
As can be seen, TbT performed better than the individual default normalization methods implemented in the other three packages (DESeq [ 20], baySeq [ 18], and NBPSeq [ 21]).
Thus, to our surprise, changing the default normalization methods to TMM normalization did not have a significant impact on the results in the present comparisons.
This normalization of the Ti/Tv ratio is proposed as a default option of PHYLORPH.
We compared a total of eight package combinations: four R packages (edgeR, DESeq, baySeq, and NBPSeq) with their default normalization settings and with our normalization strategy.
