Glycan-Centric KEGG Gene Set Enrichment Analysis

Performs glycan-centric KEGG pathway Gene Set Enrichment Analysis (GSEA). It ranks proteins within each glycan trait separately, then compares enriched pathways across traits by running GSEA for each trait.

Usage

enrich_gc_gsea_kegg(
  dea_res,
  rank_by = "signed_log10p",
  aggr = "median",
  organism = "hsa",
  p_adj_method = "BH",
  p_cutoff = 0.05,
  min_gs_size = 10,
  max_gs_size = 500,
  exponent = 1,
  eps = 1e-10,
  seed = FALSE
)

Arguments

dea_res

Differential analysis result. Can be one of:

• Result from glystats::gly_limma() (two groups), glystats::gly_ttest(), or glystats::gly_wilcox(), called on an glyexp::experiment() of "traitproteomics" type.

• A tibble with the following columns:

  • protein: Uniprot ID of proteins

  • trait: A glycosylation trait (e.g. "TFc" for proportion of core-fucosylated glycans)

  • site: The glycosylation site.

  • p_val: p-values, preferably adjusted p-values

  • log2fc: log2 of fold change

rank_by

Criteria for ranking proteins. One of the following:

• "log2fc": log2 fold change with signs

• "abs_log2fc": absolute log2 fold change

• "log10p": negative log10 p-value

• "signed_log10p" (default): log10 p-value with signs of log2 fold change

• "log2fc_log10p": log2 fold change multiplied by negative log10 p-value

aggr

Aggregation method for combining multiple scores across different traits and sites for the same protein. One of "median", "mean", or "max". Defaults to "median".

organism

KEGG organism code. Defaults to "hsa" (Homo sapiens). See clusterProfiler::gseKEGG() for details.

p_adj_method

P-value adjustment method. One of "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr", "none". Passed to pAdjustMethod of downstream enrichment function. Defaults to "BH".

p_cutoff

P-value cutoff to filter significant terms. Passed to pvalueCutoff of downstream enrichment function. Defaults to 0.05.

min_gs_size

Minimal size of each gene set for analyzing. Gene sets with fewer genes than this threshold will be excluded. Passed to minGSSize of downstream enrichment function. Defaults to 10.

max_gs_size

Maximum size of each gene set for analyzing. Gene sets with more genes than this threshold will be excluded. Passed to maxGSSize of downstream enrichment function. Defaults to 500.

exponent

Weight of each step. Passed to exponent of clusterProfiler::gseGO(). Defaults to 1.

eps

Epsilon for calculating p-values. Passed to eps of clusterProfiler::gseGO(). Defaults to 1e-10.

seed

Logical indicating whether to set a random seed for reproducibility. Passed to seed of clusterProfiler::gseGO(). Defaults to FALSE.

Value

A clusterProfiler compareClusterResult object. It can be readily converted to a tibble with tibble::as_tibble(), or visualized with clusterProfiler functions like clusterProfiler::dotplot().

What is glycan-centric enrichment?

In traditional glycoproteomics data analysis, we usually perform differential expression analysis (DEA) on glycoforms, extract proteins that have dysregulated glycosylation, then perform functional enrichment (e.g. GO) on these proteins. This is what enrich_xxx() functions do (e.g. enrich_gsea_go()).

enrich_gc_xxx() functions differ in that they link specific glycan traits with functional annotations. Instead of answering the question "Which functions are enriched in dysregulated glycoproteins?", enrich_gc_xxx() answers questions like "Which functions are enriched in proteins ranked highly for core-fucosylation changes?" Higher specificity, deeper insights. By focusing on distinct glycan motifs, it helps you pinpoint the functional relevance of specific glycosylation changes.

How it ranks proteins

GSEA requires a ranked list of proteins as input. This function first splits the DEA result by glycan trait, then ranks proteins within each trait separately. For each trait, it applies the same ranking and aggregation logic as enrich_gsea_go(), producing one ranked protein list per trait. Those trait-specific ranked lists are then compared with clusterProfiler::compareCluster() using its formula interface.

Common usage pattern

A common pattern of using this function is:

# 1. Use `glydet` to calculate derived traits or motif quantification.
trait_exp <- derive_traits(exp)  # or `quantify_motifs()`

# 2. Perform differential analysis with `glystats`.
dea_res <- gly_ttest(trait_exp)

# 3. Use this function.
go_res <- enrich_gc_gsea_go(dea_res)  # or other `enrich_gc_gsea_xxx()` functions

See also

clusterProfiler::compareCluster(), clusterProfiler::gseKEGG()