Fig. 5: A compositional data meta-analysis of oligo- and pauci-mannose N-glycans in cancer.

Both full structures (a, b) and motifs (c, d) are consistently dysregulated across cancer types in a fixed-effects meta-analysis of CLR-transformed relative abundances (using the get_meta_analysis function within glycowork, version 1.3), using a single two-tailed t-test of the combined effect size. Adding an informed-scale model based on the ion intensities (b, d) improves analytical sensitivity and reveals an absolute increase of most considered structures and motifs in cancer. Total n = 194 and combined effect sizes are shown as Cohen’s d. All results can be found in Supplementary Data 11–14. A pairwise analysis of CLR-transformed data from healthy samples and cancer tissues reveals structures of moderate predictiveness (e; AUC = 0.69), which is substantially improved by the informed-scale model (f; AUC = 0.81). ROC curves were analyzed and generated with the get_roc function (glycowork, version 1.3). g In a pan-cancer analysis of oligo- and pauci-mannose N-glycans, cancer samples exhibited a significantly greater alpha diversity of these structures (p = 0.002), measured as the number of expressed unique structures per sample, calculated with the get_biodiversity function (glycowork, version 1.3), using a single two-tailed Welch’s t-test. h Correlating CLR-transformed N-glycomics and O-glycomics data from the same prostate cancer patients in different stages reveals substantial cross-class regulation. Only significant correlations, derived from the get_SparCC function (glycowork, version 1.3), are shown as Spearman’s rho, with everything else set to zero. **p < 0.01. Source data are provided as a Source Data file.