engleski

Harmonisation of glycan data obtained by different quantification platforms for use in genome-wide association meta-analysis

Goal: LC-MS (liquid chromatography-coupled mass spectrometry) and UPLC (Ultra performance liquid chromatography) are two of many analytical methods used to measure levels of IgG released N-glycans. To maximize the number of samples for genome-wide association meta analysis of IgG N-glycome, we aimed to find an appropriate pre-processing and harmonisation method to enable the joint analysis of LC-MS and UPLC datasets. Materials and methods: We used 661 samples from CROATIA-Vis cohort for which both LC-MS and UPLC glycan measurements were available. We defined glycan traits as overall percentage of presence of sugar moiety in total IgG glycome. Prior to calculation of glycan traits we applied different normalizations followed by batch correction using empirical Bayes method. In case of LC-MS measurements, we tested weighting the abundance of individual glycoprotein by response factor and/or concentration of the given IgG subclass. The performance of harmonization was assessed using Pearson correlation coefficient for UPLC and LC-MS glycan trait values in the same sample. Results: Defined glycan traits represented the percentage of following glycosylation traits: fucosylation, monogalactosylation, digalactosylation, total galactosylation, monosialylation, total sialylation, bisecting GlcNAc and sialylation without bisecting GlcNAc. Pearson correlation coefficients were the lowest for fucosylation (0.34≤r≤0.46), while galactosylation had the highest correlation values (0.95≤r≤0.96) across different pre-processing and harmonization methods. Conclusion: All normalisation and harmonisation procedures performed comparably well. We decided to use total area normalisation without taking into account the response factor or subclass concentration given that total area normalization is currently widely used in the field.

IgG glycans ; normalization ; GWAS ; meta-analysis

nije evidentirano