Within the NPARC package documentation (Childs et al., 2019), the global normalization procedure of (Savitski et al., 2014) is mentioned as being applied to transform data prior to using the functions of the package. This normalization process involves fitting two curves - one to experimental data and one to control - and selecting the normalization curve that has the 'best' R2 value to calculate correction factors and normalize both datasets. The NPARC package then relies on relAbundance rather than a normalized or corrected fold change value. How is the user to apply the global normalization procedure within the context of input data to the NPARC package? When plotting for a first observation of the data - the corrected fold change against Temperature matches the curve quite well, while just relative abundance does not. The plot of the F-stat distribution for 'empirical' df type appears to closely match theoretical, while there are no results generated for topHits.
Code should be placed in three backticks as shown below
fits <- NPARCfit(x = df$temperature,
y = df$relAbundance,
id = df$uniqueID,
groupsNull = NULL,
groupsAlt = df$compoundConcentration,
BPPARAM = BPPARAM,
returnModels = FALSE)
sessionInfo( )
```
Hi Aidan,
you can use the TPP (https://bioconductor.org/packages/release/bioc/html/TPP.html) Bioconductor package for import and normalisation and then use the NPARC package for analysis. Please see also TPP and NPARC: A problem about the conversion of data frame. for help with converting the normalised ExpressionSet obtained from TPP to the data frame needed for NPARC analysis. Concerning your second point: the empirical degree of freedom estimation is definitely recommended. It has the limitation however, that it expects are large proportion of the proteins in the dataset to be not affected by the treatment, do consider whether this is true for your dataset. If not, the classical analysis approach available via the TPP package may be more appropriate.