Accounting for undetected compounds in statistical analyses of mass spectrometry ‘omic studies

Abstract

Mass spectrometry is an important high-throughput technique for profiling small molecular compounds in biological samples and is widely used to identify potential diagnostic and prognostic compounds associated with disease. Commonly, this data generated by mass spectrometry has many missing values resulting when a compound is absent from a sample or is present but at a concentration below the detection limit. Several strategies are available for statistically analyzing data with missing values. The accelerated failure time (AFT) model assumes all missing values result from censoring below a detection limit. Under a mixture model, missing values can result from a combination of censoring and the absence of a compound. We compare power and estimation of a mixture model to an AFT model. Based on simulated data, we found the AFT model to have greater power to detect differences in means and point mass proportions between groups. However, the AFT model yielded biased estimates with the bias increasing as the proportion of observations in the point mass increased while estimates were unbiased with the mixture model except if all missing observations came from censoring. These findings suggest using the AFT model for hypothesis testing and mixture model for estimation. We demonstrated this approach through application to glycomics data of serum samples from women with ovarian cancer and matched controls.

LC-MS Analysis of Phenolic Acids, Flavonoids and Betanin from Spineless <i>Opuntia ficus-indica</i> Fruits

“…The technical issues related to sample preparation or processing influence greatly the detection of its compounds. In all cases, a compound should be reported as a missing value in the resultant data set [16].…”

Multivariate two-part statistics for analysis of correlated mass spectrometry data from multiple biological specimens

“…Several statistical methods accommodate missing data and have been applied to MS data including accelerated failure time models (Tekwe et al, 2012), two-part models (Taylor and Pollard, 2009) and mixture models (Karpievitch et al, 2009;Taylor et al, 2013). The assumptions about the missing value mechanisms differ among these methods and as a consequence, missing values are modeled in different manners.…”

“…These models jointly test for a difference in the proportion of missing values and a difference in the means of the continuous components (Lachenbruch, 2001). Mixture models combine elements from survival analysis and two-part models by modeling missing values as a combination of censored values and the absence of a compound represented as a point-mass at 0 (Karpievitch et al, 2009;Taylor et al, 2013). These methods have only been used for single biospecimen analyses and few methods suitable for multivariate analysis of data with missing values have been developed.…”

Novel statistical approaches for missing values in truncated high-dimensional metabolomics data with a detection threshold.

“…biological samples. Raw data from a metabolomics experiment usually consist of the retention time (if liquid or gas chromatography is used for separation), the observed mass to charge ratio, and a measure of ion intensity (Taylor, Leiserowitz et al 2013). The ion intensity represents the measure of each metabolite's relative abundance whereas the mass-to-charge ratios and the retention times assist in identifying unique metabolites.…”

“…Missing values (MVs) in MS can occur from various sources both technical and biological. There are three common sources of missingness: (Taylor, Leiserowitz et al 2013) i) a metabolite could be truly missing from a sample due to biological reasons, ii) a metabolite can be present in a sample but at a concentration below the detection limit of the MS, and iii) a metabolite can be present in a sample at a level above the detection limit but fail to be detected due to technical issues related to sample processing.…”