Package 'glyclean'

Description

This function adds a new "site_sequence" column to the variable information table. It is the protein sequence around the glycosylation site. If the head and tail amino acids of the peptide sequence are insufficient, fill with "X".

This function requires the following columns in the variable information tibble:

• "protein": The protein uniprot accession.

• "protein_site": The site on the protein sequence.

Usage

add_site_seq(exp, fasta = NULL, n_aa = 7, taxid = 9606)

## S3 method for class 'glyexp_experiment'
add_site_seq(exp, fasta = NULL, n_aa = 7, taxid = 9606)

## Default S3 method:
add_site_seq(exp, fasta = NULL, n_aa = 7, taxid = 9606)

Arguments

Value

A glyexp::experiment() object with the new "site_sequence" column.

Description

This function adjusts the glycopeptide expression by the protein expression. In another word, it "strips out" the protein expression from the glycopeptide expression, so that the expression reflects the glycosylation status only.

Usage

adjust_protein(exp, pro_expr_mat, method = c("ratio", "reg"))

## S3 method for class 'glyexp_experiment'
adjust_protein(exp, pro_expr_mat, method = c("ratio", "reg"))

## Default S3 method:
adjust_protein(exp, pro_expr_mat, method = c("ratio", "reg"))

Arguments

Details

For simplicity, glycopeptide expression is denoted as GP, and protein expression is denoted as PRO.

"ratio" method

GP-adj = (GP / PRO) / (median(GP) / median(PRO))

The first part is to adjust glycopeptide expression by protein expression. The second part is to rescale the expression to the original scale.

"reg" method

Use linear regression to remove the protein expression from the glycopeptide expression. Both glycopeptide and protein expression values are log2-transformed (with +1 to avoid log(0)) before fitting the linear model: log2(GP+1) ~ log2(PRO+1). The residuals represent the glycosylation-specific signal and are converted back to the original scale using 2^residuals, ensuring all adjusted values are positive.

In both methods, only glycoproteins identified in both exp and pro_expr_mat will be retained.

Value

A glyexp::experiment() object with adjusted protein expression.

Description

Aggregate glycoproteomics data to different levels (glycoforms, glycopeptides, etc.). This function sums up quantitative values for each unique combination of specified variables. It is recommended to call this function after missing value imputation.

The following levels are available:

• "gf": Aggregate to glycoforms, which is the unique combination of proteins, protein sites, and glycan compositions. This is the default level.

• "gp": Aggregate to glycopeptides, which is the unique combination of peptides, proteins, protein sites, and glycan compositions.

• "gfs": Like "gf", but differentiates structures with the same composition.

• "gps": Like "gp", but differentiates structures with the same composition.

Different levels of aggregation require different columns in the variable information.

• "gf": "protein", "glycan_composition", "protein_site"

• "gp": "peptide", "protein", "glycan_composition", "peptide_site", "protein_site"

• "gfs": "protein", "glycan_composition", "glycan_structure", "protein_site"

• "gps": "peptide", "protein", "glycan_composition", "glycan_structure", "peptide_site", "protein_site"

Other columns in the variable information tibble with "many-to-one" relationship with the unique combination of columns listed above will be kept. A common example is the "gene" column. For each "glycoform" (unique combination of "protein", "protein_site", and "glycan_composition"), there should be only one "gene" value, therefore it is kept for "gf" level. On the other hand, the "peptide" column is removed for "gf" level, as one "glycoform" can contain multiple "peptides".

Usage

aggregate(
  exp,
  to_level = c("gf", "gp", "gfs", "gps"),
  standardize_variable = TRUE
)

## S3 method for class 'glyexp_experiment'
glyclean_aggregate(
  exp,
  to_level = c("gf", "gp", "gfs", "gps"),
  standardize_variable = TRUE
)

## Default S3 method:
glyclean_aggregate(
  exp,
  to_level = c("gf", "gp", "gfs", "gps"),
  standardize_variable = TRUE
)

Arguments

Value

A modified glyexp::experiment() object with aggregated expression matrix and updated variable information.

Description

Aggregates glycoproteomics data to "gfs" (glycoforms with structures) level if the glycan structure column exists, otherwise to "gf" (glycoforms with compositions) level.

Usage

auto_aggregate(exp, standardize_variable = TRUE)

Arguments

Value

A modified glyexp::experiment() object with aggregated expression matrix and updated variable information.

Examples

library(glyexp)
exp <- real_experiment
auto_aggregate(exp)

Description

Perform automatic data preprocessing on glycoproteomics or glycomics data. This function applies an intelligent preprocessing pipeline that includes normalization, missing value handling, imputation, aggregation (for glycoproteomics data), and batch effect correction.

For glycomics data, this function calls these functions in sequence:

• auto_remove()

• auto_normalize()

• normalize_total_area()

• auto_impute()

• auto_correct_batch_effect()

For glycoproteomics data, this function calls these functions in sequence:

• auto_normalize()

• auto_remove()

• auto_impute()

• auto_aggregate()

• auto_normalize()

• auto_correct_batch_effect()

Usage

auto_clean(
  exp,
  group_col = "group",
  batch_col = "batch",
  qc_name = "QC",
  normalize_to_try = NULL,
  impute_to_try = NULL,
  remove_preset = "discovery",
  batch_prop_threshold = 0.3,
  check_batch_confounding = TRUE,
  batch_confounding_threshold = 0.4,
  standardize_variable = TRUE
)

Arguments

Value

A modified glyexp::experiment() object.

See Also

auto_normalize(), auto_remove(), auto_impute(), auto_aggregate(), auto_correct_batch_effect()

Examples

library(glyexp)
exp <- real_experiment
auto_clean(exp)

Description

If the data has more than 50 variables, call transform_alr(). Otherwise, call transform_clr(), following the same strategy in glycowork.

Usage

auto_coda(x, by = NULL, gamma = 0.1, group_scales = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with a CoDA-transformed expression matrix (ALR if >50 variables, CLR otherwise).

Algorithmic details

The stochastic branch follows glycowork: when gamma > 0, CLR noise is sampled per feature-sample entry rather than once per sample. Without informed scales, this corresponds to jittering around the sample-specific log2 geometric mean of the non-zero components and then returning to ratio space.

The group_scales argument is interpreted in the same way as glycowork's custom_scale. For exactly two groups, provide the total-signal ratio of the second group relative to the first, either directly as a scalar or implicitly through two per-group scales. For multi-group data, provide one positive scale per group; these scales are used only in the stochastic branch.

Motif quantification

If you intend to use ALR/CLR transformation with motif quantification, you should apply the transformation after motif quantification rather than before. In another words, the recommended workflow is:

1. Perform data preprocessing.

clean_exp <- auto_clean(exp)

1. Perform motif quantification on the cleaned experiment.

motif_exp <- glydet::quantify_motifs(clean_exp, motifs)

1. Perform ALR/CLR transformation on the motif-quantified experiment.

coda_motif_exp <- auto_coda(motif_exp, by = "group", gamma = 0.1)

1. Proceed with downstream analyses on the transformed motif experiment.

dea_res <- glystats::gly_ttest(coda_motif_exp)

Description

Detects and corrects batch effects in the experiment. If batch information is available, this function performs ANOVA to detect batch effects. If more than 30% (controlled by prop_threshold) of variables show significant batch effects (p < 0.05), batch correction is performed using ComBat. If group information exists, it will be used as a covariate in both detection and correction to preserve biological variation. If no batch information is available, the function will return the original experiment.

Usage

auto_correct_batch_effect(
  exp,
  group_col = "group",
  batch_col = "batch",
  prop_threshold = 0.3,
  check_confounding = TRUE,
  confounding_threshold = 0.4
)

Arguments

Value

A glyexp::experiment() object with batch effects corrected.

Examples

exp <- glyexp::real_experiment
exp <- auto_correct_batch_effect(exp)

Description

This function automatically selects and applies a deterministic default imputation method by sample count and experiment type. Quality Control (QC) samples are inspected for workflow consistency, but they are not used to benchmark or select the imputation method.

Usage

auto_impute(exp, group_col = "group", qc_name = "QC", to_try = NULL)

Arguments

Details

The automatic strategy uses these defaults:

• n_samples < 30: impute_min_prob() for glycomics and glycoproteomics.

• ⁠30 <= n_samples <= 100⁠: impute_bpca() for glycomics and impute_min_prob() for glycoproteomics.

• n_samples > 100: impute_miss_forest() for glycomics and impute_bpca() for glycoproteomics.

Other experiment types use the glycoproteomics defaults as a conservative fallback. impute_sample_min() and impute_half_sample_min() remain available for manual use, but they are not selected automatically.

Value

The imputed experiment.

Examples

library(glyexp)
exp_imputed <- auto_impute(real_experiment)

Description

This function automatically applies a deterministic default normalization method based on experiment type. Quality Control (QC) samples are not used to benchmark, select, or report normalization behavior.

Usage

auto_normalize(exp, group_col = "group", qc_name = "QC", to_try = NULL)

Arguments

Details

The automatic strategy uses these defaults:

• glycomics: normalize_total_area().

• glycoproteomics: normalize_median().

• missing or other experiment types: normalize_median().

Value

The normalized experiment.

Examples

library(glyexp)
exp_normed <- auto_normalize(real_experiment)

Description

This function uses preset rules to remove variables with low quality. Available presets:

• "simple": remove variables with more than 50% missing values.

• "discovery": more lenient, remove variables with more than 80% missing values, but ensure less than 50% of missing values in at least one group.

• "biomarker": more strict, remove variables with more than 40% missing values, and ensure less than 60% of missing values in all groups.

Usage

auto_remove(exp, preset = "discovery", group_col = "group", qc_name = "QC")

Arguments

Value

A modified glyexp::experiment() object.

Examples

library(glyexp)
exp <- real_experiment
auto_remove(exp)

Description

Correct batch effects in glycoproteomics/glycomics data using ComBat algorithm from the sva package or removeBatchEffect from the limma package.

Usage

correct_batch_effect(
  x,
  batch = "batch",
  group = NULL,
  check_confounding = TRUE,
  confounding_threshold = 0.4,
  method = c("combat", "limma")
)

## S3 method for class 'glyexp_experiment'
correct_batch_effect(
  x,
  batch = "batch",
  group = NULL,
  check_confounding = TRUE,
  confounding_threshold = 0.4,
  method = c("combat", "limma")
)

## S3 method for class 'matrix'
correct_batch_effect(
  x,
  batch = "batch",
  group = NULL,
  check_confounding = TRUE,
  confounding_threshold = 0.4,
  method = c("combat", "limma")
)

## Default S3 method:
correct_batch_effect(
  x,
  batch = "batch",
  group = NULL,
  check_confounding = TRUE,
  confounding_threshold = 0.4,
  method = c("combat", "limma")
)

Arguments

Details

This function performs batch effect correction using either the ComBat algorithm or the limma removeBatchEffect function. It requires batch information provided via the batch parameter. If no batch information is available, the function will return the original data unchanged.

If group information is provided via group, the function will check for confounding between batch and group variables. If batch and group are highly confounded (|Cramer's V| > threshold), the function will issue a warning and return the original data unchanged to avoid over-correction.

When both batch and group information are available and not highly confounded, the group information will be included in the model to preserve biological variation while correcting for batch effects.

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a batch-corrected matrix.

Examples

# With glyexp_experiment and column names
exp <- glyexp::toy_experiment
exp$sample_info$batch <- c("A", "A", "A", "B", "B", "B")
exp$sample_info$group <- c("Ctrl", "Ctrl", "Treat", "Ctrl", "Treat", "Treat")
corrected_exp <- correct_batch_effect(exp, batch = "batch", group = "group")

# With matrix and factor vectors
mat <- matrix(abs(rnorm(200)), nrow = 20, ncol = 10)
batch_factor <- factor(rep(c("A", "B"), each = 5))
group_factor <- factor(rep(c("Ctrl", "Treat"), times = 5))
corrected_mat <- correct_batch_effect(mat, batch = batch_factor, group = group_factor)

# Using limma method
corrected_mat <- correct_batch_effect(
  mat, batch = batch_factor, group = group_factor, method = "limma"
)

Description

Use ANOVA to detect if batch effect is present in the data. If group is provided, it will be used as a covariate in the ANOVA model.

Usage

detect_batch_effect(x, batch = "batch", group = NULL)

## S3 method for class 'glyexp_experiment'
detect_batch_effect(x, batch = "batch", group = NULL)

## S3 method for class 'matrix'
detect_batch_effect(x, batch = "batch", group = NULL)

## Default S3 method:
detect_batch_effect(x, batch = "batch", group = NULL)

Arguments

Value

A double vector of p-values for each variable, i.e., the same length as nrow(x) (for matrix) or nrow(get_expr_mat(x)) (for experiment)

Examples

# With glyexp_experiment and column names
exp <- glyexp::toy_experiment
exp$sample_info$batch <- c("A", "A", "A", "B", "B", "B")
exp$sample_info$group <- c("Ctrl", "Ctrl", "Treat", "Ctrl", "Treat", "Treat")
p_values <- detect_batch_effect(exp, batch = "batch", group = "group")

# With matrix and factor vectors
mat <- matrix(rnorm(200), nrow = 20, ncol = 10)
batch_factor <- factor(rep(c("A", "B"), each = 5))
group_factor <- factor(rep(c("Ctrl", "Treat"), times = 5))
p_values <- detect_batch_effect(mat, batch = batch_factor, group = group_factor)

Description

A wrapper around the pcaMethods::pca(). Impute missing values using Bayesian principal component analysis (BPCA). BPCA combines an EM approach for PCA with a Bayesian model. In standard PCA data far from the training set but close to the principal subspace may have the same reconstruction error. BPCA defines a likelihood function such that the likelihood for data far from the training set is much lower, even if they are close to the principal subspace.

Usage

impute_bpca(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
impute_bpca(x, by = NULL, ...)

## S3 method for class 'matrix'
impute_bpca(x, by = NULL, ...)

## Default S3 method:
impute_bpca(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

A wrapper around the impute::impute.knn(). Impute missing values with values from the k-nearest neighbors of the corresponding feature.

Usage

impute_fw_knn(x, k = 5, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
impute_fw_knn(x, k = 5, by = NULL, ...)

## S3 method for class 'matrix'
impute_fw_knn(x, k = 5, by = NULL, ...)

## Default S3 method:
impute_fw_knn(x, k = 5, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

Impute missing values with half of the minimum value of the corresponding sample. This method assumes that missing values are MCAR, i.e. missing values are induced by an ion below the detection limit. Compared to impute_sample_min(), this method is more conservative.

Usage

impute_half_sample_min(x)

## S3 method for class 'glyexp_experiment'
impute_half_sample_min(x)

## S3 method for class 'matrix'
impute_half_sample_min(x)

## Default S3 method:
impute_half_sample_min(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

Impute missing values using random draws from the left-censored gaussian distribution. Missing values are imputed on the log2 intensity scale and then transformed back to the original scale.

Usage

impute_min_prob(x, by = NULL, q = 0.01, tune.sigma = 1, ...)

## S3 method for class 'glyexp_experiment'
impute_min_prob(x, by = NULL, q = 0.01, tune.sigma = 1, ...)

## S3 method for class 'matrix'
impute_min_prob(x, by = NULL, q = 0.01, tune.sigma = 1, ...)

## Default S3 method:
impute_min_prob(x, by = NULL, q = 0.01, tune.sigma = 1, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

A wrapper around the missForest::missForest(). Impute missing values using recursive running of random forests until convergence. This is a non-parametric method and works for both MAR and MNAR missing data.

Usage

impute_miss_forest(x, by = NULL, seed = 123, ...)

## S3 method for class 'glyexp_experiment'
impute_miss_forest(x, by = NULL, seed = 123, ...)

## S3 method for class 'matrix'
impute_miss_forest(x, by = NULL, seed = 123, ...)

## Default S3 method:
impute_miss_forest(x, by = NULL, seed = 123, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

A wrapper around the pcaMethods::pca(). Impute missing values using probabilistic principal component analysis (PPCA). PPCA allows to perform PCA on incomplete data and may be used for missing value estimation.

Usage

impute_ppca(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
impute_ppca(x, by = NULL, ...)

## S3 method for class 'matrix'
impute_ppca(x, by = NULL, ...)

## Default S3 method:
impute_ppca(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

Impute missing values with the minimum value of the corresponding sample. This method assumes that missing values are MCAR, i.e. missing values are induced by an ion below the detection limit. See also impute_half_sample_min().

Usage

impute_sample_min(x)

## S3 method for class 'glyexp_experiment'
impute_sample_min(x)

## S3 method for class 'matrix'
impute_sample_min(x)

## Default S3 method:
impute_sample_min(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

A wrapper around the pcaMethods::pca(). Impute missing values using singular value decomposition (SVD) imputation. SVD is a matrix factorization technique that factors a matrix into three matrices: U, Σ, and V. SVD is used to find the best lower rank approximation of the original matrix.

Usage

impute_svd(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
impute_svd(x, by = NULL, ...)

## S3 method for class 'matrix'
impute_svd(x, by = NULL, ...)

## Default S3 method:
impute_svd(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

A wrapper around the impute::impute.knn(). Impute missing values with values from the k-nearest neighbors of the corresponding sample. If there are strong patterns among the samples (such as group clustering relationships or experimental conditions), this method can better utilize the overall relationships among samples.

Usage

impute_sw_knn(x, k = 5, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
impute_sw_knn(x, k = 5, by = NULL, ...)

## S3 method for class 'matrix'
impute_sw_knn(x, k = 5, by = NULL, ...)

## Default S3 method:
impute_sw_knn(x, k = 5, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

Impute missing values in an expression matrix by replacing them with zeros.

Usage

impute_zero(x)

## S3 method for class 'glyexp_experiment'
impute_zero(x)

## S3 method for class 'matrix'
impute_zero(x)

## Default S3 method:
impute_zero(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with missing values imputed. If x is a matrix, returns a matrix with missing values imputed.

Description

This function is a wrapper around limma::normalizeCyclicLoess() with method = "pairs". Each pair of columns is normalized mutually to each other. See this paper for more information. Also see limma::normalizeCyclicLoess().

Usage

normalize_loesscyc(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
normalize_loesscyc(x, by = NULL, ...)

## S3 method for class 'matrix'
normalize_loesscyc(x, by = NULL, ...)

## Default S3 method:
normalize_loesscyc(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This function is a wrapper around limma::normalizeCyclicLoess() with method = "fast". Each column is simply normalized to a reference array, the reference array being the average of all the arrays. See this paper for more information. Also see limma::normalizeCyclicLoess().

Usage

normalize_loessf(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
normalize_loessf(x, by = NULL, ...)

## S3 method for class 'matrix'
normalize_loessf(x, by = NULL, ...)

## Default S3 method:
normalize_loessf(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

Normalize the expression matrix by dividing each column (sample) by the median of that column (NA ignored), so that the median of each column is 1. This is the most common normalization method for proteomics data. It effectively and robustly removes the bias introduced by total protein abundance, and removes batch effects in part.

Usage

normalize_median(x)

## S3 method for class 'glyexp_experiment'
normalize_median(x)

## S3 method for class 'matrix'
normalize_median(x)

## Default S3 method:
normalize_median(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

Normalize the expression matrix by dividing each column (sample) by the absolute median of that column (NA ignored), so that the absolute median of each column is 1.

Usage

normalize_median_abs(x)

## S3 method for class 'glyexp_experiment'
normalize_median_abs(x)

## S3 method for class 'matrix'
normalize_median_abs(x)

## Default S3 method:
normalize_median_abs(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This approach is based on the calculation of the dilution factor of each sample with respect to a reference sample. Here, the reference sample was calculated as the median value of each glycan's abundance across all measured samples. For each sample, a vector of quotients was then obtained by dividing each glycan measure by the corresponding value in the reference sample. The median of these quotients was then used as the sample's dilution factor, and the original sample values were subsequently divided by that value. The underlying assumption is that the different intensities observed across individuals are imputable to different amounts of the biological material in the collected samples. See this paper for more information.

Usage

normalize_median_quotient(x, by = NULL)

## S3 method for class 'glyexp_experiment'
normalize_median_quotient(x, by = NULL)

## S3 method for class 'matrix'
normalize_median_quotient(x, by = NULL)

## Default S3 method:
normalize_median_quotient(x, by = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This function is a wrapper around limma::normalizeQuantiles(). It normalizes the expression matrix by quantile normalization. This method is used to remove technical variation between samples. Proteomics data rarely uses this method, but it is common in microarray data. See wikipedia for more information.

Usage

normalize_quantile(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
normalize_quantile(x, by = NULL, ...)

## S3 method for class 'matrix'
normalize_quantile(x, by = NULL, ...)

## Default S3 method:
normalize_quantile(x, by = NULL, ...)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This method is based on robust linear regression. The reference sample is calculated as the median value of each variable's abundance across all measured samples. For each sample, a robust linear regression model is fitted to the sample's abundance values against the reference sample's abundance values. The fitted model is then used to normalize the sample's abundance values. The underlying assumption is that the different intensities observed across individuals are imputable to different amounts of the biological material in the collected samples.

Usage

normalize_rlr(x, by = NULL)

## S3 method for class 'glyexp_experiment'
normalize_rlr(x, by = NULL)

## S3 method for class 'matrix'
normalize_rlr(x, by = NULL)

## Default S3 method:
normalize_rlr(x, by = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This method is based on robust linear regression with median adjustment. First, the median of each variable's abundance across all measured samples is subtracted from the sample's abundance values. Then, it's like the normalize_rlr() method.

Usage

normalize_rlrma(x, by = NULL)

## S3 method for class 'glyexp_experiment'
normalize_rlrma(x, by = NULL)

## S3 method for class 'matrix'
normalize_rlrma(x, by = NULL)

## Default S3 method:
normalize_rlrma(x, by = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This method is based on robust linear regression with median adjustment and cyclic normalization. The method is applied iteratively to each pair of samples. For each pair of samples, the median of the differences between the two samples is calculated. A robust linear regression model is fitted to the differences against the averages of the two samples. The fitted model is then used to normalize the two samples. The process is repeated for a number of iterations.

Usage

normalize_rlrmacyc(x, n_iter = 3, by = NULL)

## S3 method for class 'glyexp_experiment'
normalize_rlrmacyc(x, n_iter = 3, by = NULL)

## S3 method for class 'matrix'
normalize_rlrmacyc(x, n_iter = 3, by = NULL)

## Default S3 method:
normalize_rlrmacyc(x, n_iter = 3, by = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

Normalize the expression matrix by dividing each column (sample) by the sum of that column, so that the sum of each column is 1. This is the most common normalization method for glycomics data. It removes the bias introduced by total glycan abundance. However, it results in compositional data, which may result in unrealistic downstream analysis results.

Usage

normalize_total_area(x)

## S3 method for class 'glyexp_experiment'
normalize_total_area(x)

## S3 method for class 'matrix'
normalize_total_area(x)

## Default S3 method:
normalize_total_area(x)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

This function is a wrapper around limma::normalizeVSN(). Evidence proved that this method performs well in reducing noise and boosting differential expression detection. Log-transformation is not needed for downstream statistical analysis, for this normalization method performs a log-like transformation internally. Due to the same reason, fold change estimates will be severely distorted. Please use this method with caution. See this paper for more information.

Usage

normalize_vsn(x, by = NULL, ...)

## S3 method for class 'glyexp_experiment'
normalize_vsn(x, by = NULL, ...)

## S3 method for class 'matrix'
normalize_vsn(x, by = NULL, ...)

## Default S3 method:
normalize_vsn(x, by = NULL, ...)

Arguments

Details

At least 42 variables should be provided for this method. This follows the convention of the vsn package.

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with normalized expression matrix. If x is a matrix, returns a normalized matrix.

Description

Draw a PCA score plot for samples and color points by batch. PCA is computed on log2-transformed intensities after removing variables with missing values.

Usage

plot_batch_pca(exp, batch_col = "batch")

Arguments

Value

A ggplot object of PCA scores.

Examples

exp <- glyexp::toy_experiment
exp$sample_info$batch <- rep(c("A", "B"), each = 3)
plot_batch_pca(exp, batch_col = "batch")

Description

Compute coefficient of variation (CV) for each variable and plot its density. When by is provided, CVs are computed within each group and densities are shown with different fills.

Usage

plot_cv_dent(exp, by = NULL)

Arguments

Value

A ggplot object of CV density.

Examples

plot_cv_dent(glyexp::toy_experiment)
exp <- glyexp::toy_experiment
exp$sample_info$group <- rep(c("A", "B"), length.out = ncol(exp$expr_mat))
plot_cv_dent(exp, by = "group")

Description

Draw boxplots of log2-transformed intensities for each sample. Optionally color and group samples by a metadata variable.

Usage

plot_int_boxplot(exp, by = NULL)

Arguments

Value

A ggplot object of log-intensity boxplots.

Examples

plot_int_boxplot(glyexp::toy_experiment)
plot_int_boxplot(glyexp::toy_experiment, by = "group")

Description

Draw a bar plot of missing value proportions for each sample or variable. Items are ordered from low to high missing proportion.

Usage

plot_missing_bar(exp, on = "sample")

Arguments

Value

A ggplot object of missing value proportions by item.

Examples

plot_missing_bar(glyexp::toy_experiment)

Description

Draw a binary heatmap showing the missing value pattern in an experiment. Values are binarized: 1 (present) or 0 (missing). Rows (variables) are sorted by missing value proportion from low to high. Columns (samples) are clustered using hierarchical clustering.

Usage

plot_missing_heatmap(exp, ...)

Arguments

Value

A ggplot object of the missing value heatmap.

Examples

plot_missing_heatmap(glyexp::toy_experiment)

Description

Draw a scatter plot of proteins ranked by mean log2 intensity. Proteins are ordered from high to low mean intensity along the x-axis.

Usage

plot_rank_abundance(exp)

Arguments

Value

A ggplot object of protein rank abundance.

Examples

plot_rank_abundance(glyexp::toy_experiment)

Description

Randomly draw replicate sample pairs and plot log2 intensity scatter plots. The plot title shows sample names, and the subtitle reports the R2 value.

Usage

plot_rep_scatter(exp, rep_col, n_pairs = 9)

Arguments

Value

A patchwork object containing replicate scatter plots.

Examples

exp <- glyexp::toy_experiment
exp$sample_info$replicate <- rep(c("A", "B"), length.out = ncol(exp$expr_mat))
plot_rep_scatter(exp, rep_col = "replicate", n_pairs = 4)

Description

Draw boxplots of relative log expression (log2 intensity minus row median) for each sample. Optionally color and group samples by a metadata variable.

Usage

plot_rle(exp, by = NULL)

Arguments

Value

A ggplot object of RLE boxplots.

Examples

plot_rle(glyexp::toy_experiment)
plot_rle(glyexp::toy_experiment, by = "group")

Description

Draw a bar plot showing total intensity (TIC) for each sample. Samples are ordered from high to low TIC from left to right.

Usage

plot_tic_bar(exp)

Arguments

Value

A ggplot object of total intensity by sample.

Examples

plot_tic_bar(glyexp::toy_experiment)

Description

Constant variables are variables with the same value in all samples. This function is equivalent to remove_low_var(x, var_cutoff = 0, by = by, strict = strict).

Usage

remove_constant(x, by = NULL, strict = FALSE)

Arguments

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a filtered matrix.

See Also

remove_low_var()

Description

Filters variables whose coefficient of variation falls below a threshold. Default behavior is to remove variables with zero coefficient of variation.

Usage

remove_low_cv(x, cv_cutoff = 0, by = NULL, strict = FALSE)

## S3 method for class 'glyexp_experiment'
remove_low_cv(x, cv_cutoff = 0, by = NULL, strict = FALSE)

## S3 method for class 'matrix'
remove_low_cv(x, cv_cutoff = 0, by = NULL, strict = FALSE)

## Default S3 method:
remove_low_cv(x, cv_cutoff = 0, by = NULL, strict = FALSE)

Arguments

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a filtered matrix.

See Also

remove_low_var()

Description

Filters variables based on median expression values. Variables with median expression values below certain percentile will be removed.

Usage

remove_low_expr(x, percentile = 0.05, by = NULL, strict = FALSE)

## S3 method for class 'glyexp_experiment'
remove_low_expr(x, percentile = 0.05, by = NULL, strict = FALSE)

## S3 method for class 'matrix'
remove_low_expr(x, percentile = 0.05, by = NULL, strict = FALSE)

## Default S3 method:
remove_low_expr(x, percentile = 0.05, by = NULL, strict = FALSE)

Arguments

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a filtered matrix.

Description

Filters variables whose variance falls below a threshold. Default behavior is to remove variables with zero variance.

Usage

remove_low_var(x, var_cutoff = 0, by = NULL, strict = FALSE)

## S3 method for class 'glyexp_experiment'
remove_low_var(x, var_cutoff = 0, by = NULL, strict = FALSE)

## S3 method for class 'matrix'
remove_low_var(x, var_cutoff = 0, by = NULL, strict = FALSE)

## Default S3 method:
remove_low_var(x, var_cutoff = 0, by = NULL, strict = FALSE)

Arguments

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a filtered matrix.

See Also

remove_low_cv(), remove_constant()

Description

Remove Rare Variables with Too Many Missing Values

Usage

remove_rare(x, prop = NULL, n = NULL, by = NULL, strict = FALSE, min_n = NULL)

## S3 method for class 'glyexp_experiment'
remove_rare(x, prop = NULL, n = NULL, by = NULL, strict = FALSE, min_n = NULL)

## S3 method for class 'matrix'
remove_rare(x, prop = NULL, n = NULL, by = NULL, strict = FALSE, min_n = NULL)

## Default S3 method:
remove_rare(x, prop = NULL, n = NULL, by = NULL, strict = FALSE, min_n = NULL)

Arguments

Value

For glyexp_experiment input, returns a modified glyexp_experiment object. For matrix input, returns a filtered matrix.

Examples

# With glyexp_experiment
exp <- glyexp::toy_experiment
exp$expr_mat[1, 1] <- NA    # V1: 1/6 missing
exp$expr_mat[2, 1:3] <- NA  # V2: 3/6 missing
exp$expr_mat[3, 1:5] <- NA  # V3: 5/6 missing
exp$expr_mat[4, 1:6] <- NA  # V4: 6/6 missing
exp$expr_mat

# Remove variables with more than 50% missing values.
remove_rare(exp, prop = 0.5)$expr_mat

# Remove variables with more than 2 missing values.
remove_rare(exp, n = 2)$expr_mat

# Remove variables if they have more than 1 missing value in all groups.
# In another word, keep variables as long as they have 1 or 0 missing value
# in any group.
remove_rare(exp, by = "group", strict = FALSE)$expr_mat

# Keep only variables with no missing values.
remove_rare(exp, prop = 0)$expr_mat

# Use custom min_n to require at least 4 non-missing values
remove_rare(exp, min_n = 4)$expr_mat

# With matrix
mat <- matrix(c(1, 2, NA, 4, 5, NA, 7, 8, 9), nrow = 3)
mat_filtered <- remove_rare(mat, prop = 0.5)

Description

This function implements a glycoWork-compatible ALR preprocessing strategy. Internally, the data are transformed relative to an automatically selected reference glycan using glycoWork-style ALR logic, then back-transformed to the original ratio space before returning the result.

Usage

transform_alr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## S3 method for class 'glyexp_experiment'
transform_alr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## S3 method for class 'matrix'
transform_alr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## Default S3 method:
transform_alr(x, by = NULL, gamma = 0.1, group_scales = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with an ALR-transformed expression matrix. If x is a matrix, returns an ALR-transformed matrix. When ALR succeeds, the reference glycan is excluded from the result and the output therefore has one fewer row than the input. When ALR falls back to CLR, the returned object keeps the original dimensions. The returned values are back-transformed to the original ratio space, corresponding to x / x_ref.

Algorithmic details

The stochastic branch follows glycowork: when gamma > 0, CLR noise is sampled per feature-sample entry rather than once per sample. Without informed scales, this corresponds to jittering around the sample-specific log2 geometric mean of the non-zero components and then returning to ratio space.

The group_scales argument is interpreted in the same way as glycowork's custom_scale. For exactly two groups, provide the total-signal ratio of the second group relative to the first, either directly as a scalar or implicitly through two per-group scales. For multi-group data, provide one positive scale per group; these scales are used only in the stochastic branch.

Motif quantification

If you intend to use ALR/CLR transformation with motif quantification, you should apply the transformation after motif quantification rather than before. In another words, the recommended workflow is:

1. Perform data preprocessing.

clean_exp <- auto_clean(exp)

1. Perform motif quantification on the cleaned experiment.

motif_exp <- glydet::quantify_motifs(clean_exp, motifs)

1. Perform ALR/CLR transformation on the motif-quantified experiment.

coda_motif_exp <- auto_coda(motif_exp, by = "group", gamma = 0.1)

1. Proceed with downstream analyses on the transformed motif experiment.

dea_res <- glystats::gly_ttest(coda_motif_exp)

Description

This function implements a glycoWork-compatible CLR preprocessing strategy. Internally, the data are transformed on the log2 scale following glycowork::clr_transformation(), then back-transformed to the original ratio space before returning the result.

Usage

transform_clr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## S3 method for class 'glyexp_experiment'
transform_clr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## S3 method for class 'matrix'
transform_clr(x, by = NULL, gamma = 0.1, group_scales = NULL)

## Default S3 method:
transform_clr(x, by = NULL, gamma = 0.1, group_scales = NULL)

Arguments

Value

Returns the same type as the input. If x is a glyexp_experiment, returns a glyexp_experiment with transformed expression matrix. If x is a matrix, returns a transformed matrix. The returned values are back-transformed to the original ratio space. Zeros in the input therefore remain zeros in the output.

Algorithmic details

The stochastic branch follows glycowork: when gamma > 0, CLR noise is sampled per feature-sample entry rather than once per sample. Without informed scales, this corresponds to jittering around the sample-specific log2 geometric mean of the non-zero components and then returning to ratio space.

The group_scales argument is interpreted in the same way as glycowork's custom_scale. For exactly two groups, provide the total-signal ratio of the second group relative to the first, either directly as a scalar or implicitly through two per-group scales. For multi-group data, provide one positive scale per group; these scales are used only in the stochastic branch.

Motif quantification

If you intend to use ALR/CLR transformation with motif quantification, you should apply the transformation after motif quantification rather than before. In another words, the recommended workflow is:

1. Perform data preprocessing.

clean_exp <- auto_clean(exp)

1. Perform motif quantification on the cleaned experiment.

motif_exp <- glydet::quantify_motifs(clean_exp, motifs)

1. Perform ALR/CLR transformation on the motif-quantified experiment.

coda_motif_exp <- auto_coda(motif_exp, by = "group", gamma = 0.1)

1. Proceed with downstream analyses on the transformed motif experiment.

dea_res <- glystats::gly_ttest(coda_motif_exp)