Normalization and integration of large-scale metabolomics data using support vector regression

Abstract

Introduction Untargeted metabolomics studies for biomarker discovery often have hundreds to thousands of human samples. Data acquisition of large-scale samples has to be divided into several batches and may span from months to as long as several years. The signal drift of metabolites during data acquisition (intra-and inter-batch) is unavoidable and is a major confounding factor for largescale metabolomics studies. Objectives We aim to develop a data normalization method to reduce unwanted variations and integrate multiple batches in large-scale metabolomics studies prior to statistical analyses. Methods We developed a machine learning algorithmbased method, support vector regression (SVR), for largescale metabolomics data normalization and integration. An R package named MetNormalizer was developed and provided for data processing using SVR normalization.Results After SVR normalization, the portion of metabolite ion peaks with relative standard deviations (RSDs) less than 30 % increased to more than 90 % of the total peaks, which is much better than other common normalization methods. The reduction of unwanted analytical variations helps to improve the performance of multivariate statistical analyses, both unsupervised and supervised, in terms of classification and prediction accuracy so that subtle metabolic changes in epidemiological studies can be detected. Conclusion SVR normalization can effectively remove the unwanted intra-and inter-batch variations, and is much better than other common normalization methods.

Development of LC‐QTOF‐MS method for human lung tissue fingerprinting. A preliminary application to nonsmall cell lung cancer

“…20% loss in features . On the other hand during long runs the signal intensity of metabolites drifts due to MS instrument contamination, matrix effects, or LC column degradation . Consequently, long method limits significantly a number of samples that can be analyzed in a single batch.…”

“…Consequently, long method limits significantly a number of samples that can be analyzed in a single batch. Although different strategies for integration of several analytical batches of fingerprinting data were proposed , signal variation between batches makes integration of data challenging and great care for experimental design, data acquisition, quality control, and subsequent data analysis is required . Therefore, the concept of the present study was to develop a methodology for lung tissue fingerprinting with single analytical platform (LC‐QTOF‐MS) that can detect metabolites from several classes and at the same time is relatively short to cover a large number of samples in a single batch.…”

Trends in the application of high-resolution mass spectrometry for human biomonitoring: An analytical primer to studying the environmental chemical space of the human exposome

“…For example, in the identification of novel per- and polyfluorinated compounds in human serum, HRMS data was normalized using an internal standard peak area ( 13 C 4 PFOS) and total area sums (Rotander et al, 2015). The latter approach involves the use of a pooled QC sample across the batch and applies computational models to correct the abundance deviation of a molecular feature in a sample according to its performance in the neighboring QC run using a locally estimated scatterplot smoothing (LOESS) method (Shen et al, 2016b). Data scaling and transformation compares molecular features within and between samples, and is considered a between chromatograms or column-wise correction (van den Berg et al, 2006).…”

Review of emerging metabolomic tools and resources: 2015–2016

“…MetNormalizer , is an R package developed to reduce confounding factors and unwanted variations, such as intra‐ and interbatch variability, using the machine learning algorithm‐based method, support vector regression normalization .…”