A new pipeline for the normalization and pooling of metabolomics data

• DOI: 10.1101/2021.07.16.452593
• Publication Date: 2021-07-17T00:15:32Z
• Authors: Vivian Viallon; Mathilde His; Sabina Rinaldi; Marie Breeur; Audrey Gicquiau; Bertrand Hemon; Kim Overvad; Anne Tjønneland; Agnetha Linn Rost...; Joseph A. Rothwell; Lucie Lecuyer; Gianluca Severi; Rudolf Kaaks; Theron Johnson; Matthias B. Schulze; Domenico Palli; Claudia Agnoli; Salvatore Panico; Rosario Tumino; Fulvio Ricceri; Monique Verschuren; Peter Engelfriet; Charlotte Onland; Roel Vermeulen; Therese Haugdahl ...; Ilona Urbarova; Raul Zamora-Ros; Miguel Rodriguez-...; Pilar Amiano; José Maria Huerta; Eva Ardanaz; Olle Melander; Filip Ottoson; Linda Vidman; Matilda Rentoft; Julie A Schmidt; Ruth C Travis; Elisabete Weiderpass; Mattias Johansson; Laure Dossus; Mazda Jenab; Marc J Gunter; Lorenzo Bermejo; Dominique Scherer; Reza M Salek; Pekka Keski-Rahkonen; Pietro Ferrari
• Keywords: Pooling; Normalization; Data set

Abstract Pooling metabolomics data across studies is often desirable to increase the statistical power of analysis. However, this can raise methodological challenges as several preanalytical and analytical factors could introduce differences in measured concentrations variability between datasets. Specifically, different may use variable sample types (e.g., serum versus plasma) collected, treated stored according protocols, assayed laboratories using instruments. To address these issues, a new pipeline was developed normalize pool through set sequential steps: (i) exclusions least informative observations metabolites removal outliers; imputation missing data; (ii) identification main sources PC-PR2 analysis; (iii) application linear mixed models remove unwanted variability, including samples’ originating study batch, preserve biological variations while accounting for potential residual variances studies. This applied targeted acquired Biocrates AbsoluteIDQ kits eight case-control nested within European Prospective Investigation into Cancer Nutrition (EPIC) cohort. Comprehensive examination measurements indicated that improved comparability Our be adapted other molecular data, biomarkers well proteomics used pooling datasets, example international consortia, limit biases introduced by inter-study variability. versatility makes our work interest epidemiologists.