Calculation of FC Dose Levels in Anova Output

[empiricalbayes]

I wonder if you could clarify how FCs in ANOVA's output for each level of doses is calculated. Are they calculated for each individual gene/probe separately? For example, using this simplified data set

NormalizedCounts_Test.xlsx

and after running ANOVA, only one probe passes the pre-defined filters
ANOVA_Test.xlsx

I wonder if you could clarify how those FC Dose Levels are calculated for this probe.

Thanks.

[auerbachs]

It appears that you left your data unlogged. I would suggest you log2 transform the data, reload it and when loading indicate that it is log2 transformed. Once you do this and then run the ANOVA along with FC you should see the result that you are expecting. In the software FC is calculated by comparing the zero or lowest dose group to the treated groups.

[empiricalbayes]

Thanks for the explanation. I am a bit confused. Why should one necessarily log2 transform input data while when uploading data into BMDExpress you can choose whether your data are log transformed or not?

However, I log2 transformed my (updated) data
Log2_NormalizedCounts_Test.xlsx

Here is ANOVA's output using the default parameters
ANOVA_Test.xlsx

I still cannot understand how those FC Dose Levels are calculated. Do you mind writing down your definition in terms of R code or formula? Thanks.

[auerbachs]

The reason it is more appropriate to log the data is that the ANOVA assumes normality.

Fold change when data is non-transformed (i.e, designated upon loading as NONE)
Ratio = Avg(Dose group values loaded by user)/Avg(Control dose group values loaded by user)
If Ratio > 1, then Ratio= FC
If Ratio <1 then FC = -1/Ratio

Fold change when data is logarithmically transformed (i.e, designated upon loading as log2, log10, ln)
Log Ratio = Avg(Log dose group values loaded by user) – Avg(Log control dose group values loaded by user)
Depending on the form of the log, Ratio = 2^Log Ratio OR 10^Log Ratio OR e^Log Ratio
If Ratio > 1, then Ratio= FC
If Ratio <1 then FC = -1/Ratio

This will mean that the FC values will differ slightly when the data are non-transformed vs log transformed as one is based on an arithmetic mean of the groups (non-transformed) compared to the logarithm of the Geometric Mean of the groups.

[jichaojoyce]

Hi I am also confused on the FC Dose level. What does Avg(Log dose group values loaded by user) and Avg(Log control dose group values loaded by user) mean?