Counting missing values in a metabolite-intensity data set for measuring the analytical performance of a metabolomics platform.

Metabolomics requires quantitative comparison of individual metabolites present in an entire sample set. Unfortunately, missing intensity values in one or more samples are very common. Because missing values can have a profound influence on metabolomic results, the extent of missing values found in a metabolomic data set should be treated as an important parameter for measuring the analytical performance of a technique. In this work, we report a study on the scope of missing values and a robust method of filling the missing values in a chemical isotope labeling (CIL) LC-MS metabolomics platform. Unlike conventional LC-MS, CIL LC-MS quantifies the concentration differences of individual metabolites in two comparative samples based on the mass spectral peak intensity ratio of a peak pair from a mixture of differentially labeled samples. We show that this peak-pair feature can be explored as a unique means of extracting metabolite intensity information from raw mass spectra. In our approach, a peak-pair peaking algorithm, IsoMS, is initially used to process the LC-MS data set to generate a CSV file or table that contains metabolite ID and peak ratio information (i.e., metabolite-intensity table). A zero-fill program, freely available from MyCompoundID.org , is developed to automatically find a missing value in the CSV file and go back to the raw LC-MS data to find the peak pair and, then, calculate the intensity ratio and enter the ratio value into the table. Most of the missing values are found to be low abundance peak pairs. We demonstrate the performance of this method in analyzing an experimental and technical replicate data set of human urine metabolome. Furthermore, we propose a standardized approach of counting missing values in a replicate data set as a way of gauging the extent of missing values in a metabolomics platform. Finally, we illustrate that applying the zero-fill program, in conjunction with dansylation CIL LC-MS, can lead to a marked improvement in finding significant metabolites that differentiate bladder cancer patients and their controls in a metabolomics study of 109 subjects.