normalization_algorithms: returns all normalization algorithms integrated with MS-DAP
In ftwkoopmans/msdap: Mass Spectrometry Downstream Analysis Pipeline

normalization_algorithms

R Documentation

returns all normalization algorithms integrated with MS-DAP

Description

median: scale each sample such that median abundance values are the same for all samples in the dataset.

loess: Loess normalization as implemented in the limma R package (PMID:25605792) https://bioconductor.org/packages/release/bioc/html/limma.html. code: limma::normalizeCyclicLoess(log2_data, iterations = 10, method = "fast"). Normalize the columns of a matrix, cyclicly applying loess normalization to normalize each columns against the average over all columns.

vsn: Variance Stabilizing Normalization (VSN) as implemented in the vsn R package (PMID:12169536) https://bioconductor.org/packages/release/bioc/html/vsn.html. code: vsn::justvsn(). From bioconductor: "The model incorporates data calibration step (a.k.a. normalization), a model for the dependence of the variance on the mean intensity and a variance stabilizing data transformation. Differences between transformed intensities are analogous to 'normalized log-ratios'".

rlr: Robust Linear Regression normalization, as implemented in the MSqRob package (PMID:26566788) https://github.com/statOmics/msqrob. For each sample s, perform a robust linear regression of all values (peptide intensities) against overall median values (e.g. median value of each peptide over all samples) to obtain the normalization factor for sample s.

msempire: log-foldchange mode normalization, as implemented in the msEmpiRe package (PMID:31235637) https://github.com/zimmerlab/MS-EmpiRe. Instead of computing all pairwise sample scaling factors (i.e. foldchange distributions between all pairs of samples within a sample group), MS-EmpiRe uses single linkage clustering to normalize to subsets of 'most similar' samples and iteratively expands until all within-group samples are covered.

vwmb: Variation Within, Mode Between (VWMB) normalization. In brief, this minimizes the median peptide variation within each sample group, then scales between all pairwise sample groups such that the log-foldchange mode is zero. The normalization algorithm consists of two consecutive steps:

samples are scaled within each group such that the median of variation estimates for all rows is minimized
summarize all samples per group by respective row mean values (from row*sample to a row*group matrix). Then rescale at the sample-group-level to minimize the mode log-foldchange between all groups See further MS-DAP function normalize_vwmb.

mwmb: Mode Within, Mode Between (MWMB) normalization. A variant of VWMB. Normalize (/scale) samples within each sample group such that their pairwise log-foldchange modes are zero, then scales between groups such that the log-foldchange mode is zero (i.e. the between-group part is the same as VWMB). If the dataset has (unknown) covariates and a sufficient number of replicates, this might be beneficial because covariate-specific effects are not averaged out as they might be with VWMB. See further MS-DAP function normalize_vwmb.

modebetween: only the "Mode Between" part of VWMB described earlier, does not affect scaling between (replicate) samples within the same sample group. Note that this is mode-between normalization at peptide level, in most cases you'll want to use "modebetween_protein" instead.

modebetween_protein (also referred to as "MBprot", e.g. in the MS-DAP manuscript and some documentation): only the "Mode Between" part of VWMB described earlier, but the scaling factors are computed at protein-level !! When this normalization function is used, the normalize_modebetween_protein function will first rollup the peptide data matrix to protein-level, then compute between-sample-group scaling factors and finally apply those to the input peptide-level data matrix to compute the normalized peptide data.

vw: only perform peptide-level variation-within normalization. This rescales samples such that the peptide standard deviations (intensity values across samples in the same sample group) are minimized. i.e. this is only the first part of the VWMB algorithm. Note that in most cases / experimental designs, you want to end your normalizations with modebetween_protein (only normalization variation within group is often not the most suitable normalization solution)

mw: analogous to the "vw" option, but here performing peptide-level normalization that uses mode normalization within groups (first part of MWMB). No between group rescaling. Same recommendations/notes as with "vm" apply