Applies modular normalization including 80% filtering rule, total-ion count normalization, missing value imputation, and outlier detection using Hotelling's T2 test.
Usage
processing(
data,
metadata_sample = NULL,
metadata_feature = NULL,
metadata_info = c(Conditions = "Conditions"),
featurefilt = "Modified",
cutoff_featurefilt = 0.8,
tic = TRUE,
mvi = TRUE,
mvi_percentage = 50,
hotellins_confidence = 0.99,
core = FALSE,
save_plot = "svg",
save_table = "csv",
print_plot = TRUE,
path = NULL
)Arguments
- data
SummarizedExperiment or data frame. If SummarizedExperiment, metadata_sample is extracted from colData. If data frame, provide unique sample identifiers as row names and metabolite numerical values in columns with metabolite identifiers as column names. Use NA for undetected metabolites.
- metadata_sample
Data frame (optional). Only required if data is not a SummarizedExperiment. Contains information about samples, combined with input data based on unique sample identifiers used as row names. Must contain Conditions column. If experiment has no multiple conditions, assign all samples to same condition. Default: NULL.
- metadata_feature
Data frame (optional). Only used if data is not a SummarizedExperiment (if it is, this is instead extracted from rowData). Contains feature-level metadata, with feature identifiers matching the column names of data as row names. Only required if featurefilt = "Manual", in which case it must contain a logical TRUE/FALSE column (TRUE = remove this feature) named "FilteredFeatures", or as configured via metadata_info. Default: NULL.
- metadata_info
Named character vector (optional). Contains names of experimental parameters: c(Conditions="ColumnName", Biological_Replicates="ColumnName"). Column "Conditions" (mandatory) contains sample conditions (e.g., "N"/"T" or "Normal"/"Tumor"), used for feature filtering and PCA color coding. Column "BiologicalReplicates" (optional) contains numerical values. For core=TRUE, must also add core_norm_factor="ColumnName" and core_media="ColumnName". Column core_norm_factor is used for normalization; core_media specifies media controls in Conditions. Optionally, FilteredFeatures="ColumnName" names the column in metadata_feature holding the manual TRUE/FALSE flags for featurefilt = "Manual"; if not given, a column literally named "FilteredFeatures" is expected. Default: c(Conditions="Conditions").
- featurefilt
Character (optional). If NULL, no feature filtering is performed. If "Standard", applies 80% filtering rule (Bijlsma et al., 2006) on metabolite features across whole dataset. If "Modified", filtering is done per condition and Conditions column must be provided (Yang et al., 2015). If "Manual", the features to remove are taken directly from metadata_feature instead of being computed from a cutoff. If the expected column is missing from metadata_feature, a warning is raised and "Modified" is used instead; if "Modified" itself cannot be used because Conditions has fewer than 2 levels, a further warning is raised and "Standard" is used. Default: "Standard".
- cutoff_featurefilt
Numeric (optional). Percentage threshold for feature filtering. Ignored if featurefilt = "Manual". Default: 0.8.
- tic
Logical (optional). Whether total ion count normalization is performed. Default: TRUE.
- mvi
Logical (optional). Whether missing value imputation (MVI) based on half minimum is performed. Default: TRUE.
- mvi_percentage
Numeric (optional). Percentage (0-100) of imputed value based on minimum value. Default: 50.
- hotellins_confidence
Numeric (optional). Confidence level for outlier identification in Hotelling's T2 test. Default: 0.99.
- core
Logical (optional). Whether consumption-release experiment was performed and core value should be calculated. If TRUE, provide normalization factor column "core_norm_factor" in metadata_sample where Conditions column matches. The normalization factor must be numerical value from growth rate (growth curve) or growth factor (ratio of cell count/protein quantification at start vs. end point). Additionally, control media samples must be available in data and defined as "core_media" in Conditions column of metadata_sample. Default: FALSE.
- save_plot
Character (optional). File type of output plots: "svg", "png", "pdf". If NULL, plots are not saved. Default: "svg".
- save_table
Character (optional). File type of output table: "csv", "xlsx", "txt". If NULL, tables are not saved. Default: "csv".
- print_plot
Logical (optional). Whether to print overview of resulting plots. Default: TRUE.
- path
Character (optional). Path to folder where results should be saved. Default: NULL.
Value
List with two elements: DF (all output tables generated) and Plot (all plots generated). If featurefilt is not NULL, DF includes a Feature_matrix table listing every input feature (not only the removed ones) with a FilteredFeatures column (TRUE/FALSE). If data is a SummarizedExperiment, the same FilteredFeatures column is added to the rowData (feature matrix) of the returned data_Rawdata SummarizedExperiment.
Examples
data(intracell_raw)
Intra <- intracell_raw %>% tibble::column_to_rownames("Code")
ResI <- processing(
data = Intra[1:30, -c(1:3)],
metadata_sample = Intra[1:30, c(1:3)],
metadata_info = c(
Conditions = "Conditions",
Biological_Replicates = "Biological_Replicates"
),
save_plot = NULL,
save_table = NULL,
print_plot = FALSE
)
#> feature_filtering: Here we apply the modified 80%-filtering rule that takes the class information (Column `Conditions`) into account, which additionally reduces the effect of missing values (REF: Yang et. al., (2015), doi: 10.3389/fmolb.2015.00004). Filtering value selected: 0.8
#> 3 metabolites where removed: FAICAR, SAICAR, AICAR
#> Missing Value Imputation: Missing value imputation is performed, as a complementary approach to address the missing value problem, where the missing values are imputing using the `half minimum value`. REF: Wei et. al., (2018), Reports, 8, 663, doi:https://doi.org/10.1038/s41598-017-19120-0
#> total Ion Count (tic) normalization: total Ion Count (tic) normalization is used to reduce the variation from non-biological sources, while maintaining the biological variation. REF: Wulff et. al., (2018), Advances in Bioscience and Biotechnology, 9, 339-351, doi:https://doi.org/10.4236/abb.2018.98022
#> Outlier detection: Identification of outlier samples is performed using Hotellin's T2 test to define sample outliers in a mathematical way (Confidence = 0.99 ~ p.val < 0.01) (REF: Hotelling, H. (1931), Annals of Mathematical Statistics. 2 (3), 360-378, doi:https://doi.org/10.1214/aoms/1177732979). hotellins_confidence value selected: 0.99
#> There are possible outlier samples in the data
#> Filtering round 1 Outlier Samples: MS55_29
#> Filtering round 2 Outlier Samples: MS55_30