Differential variation and expression analysis

Abstract

We propose an empirical Bayes approach using a three-component mixture model, the L 2 N model, that may be applied to detect both differential expression (mean) and variation. It consists of two log-normal components (L 2 ) for the differentially expressed (dispersed) features (one component for under-expressed [dispersed] and the other for over-expressed [dispersed] features), and a single normal component (N ) for the null features (i.e., non-differentially expressed [dispersed] features). Simulation results show that L 2 N can capture asymmetries in the numbers of over-and under-expressed (dispersed) features (e.g., genes) when they exist, can provide a better fit to data in which the distributions of the null and non-null features are not well-separated, but can also perform well under symmetry and separation. Thus the L 2 N model is particularly appealing when no a priori biological knowledge about the mixture density is available. The L 2 N model is implemented in an R package called DVX, for Differential Variation and eXpression analysis. The package also includes an implementation of differential expression analysis via the limma package, and a differential variation and expression analysis using a three-way normal mixture model. DVX is a user-friendly, graphical interface implemented via the 'Shiny' package [6], so that a user is not required to have R programming knowledge. It offers a set of diagnostics plots, data transformation tools, and report generation in Microsoft Excel-and Word-compatible formats. The package is available on the web, at https://haim-bar.uconn.edu/software/DVX/ .

Differential variation and expression analysis

“…To make the method easily accessible to biologists, we created a user‐friendly R Shiny interface called DVX (Bar & Schifano, ), which also includes the implementations of N 3 and limma. With all three models, it is possible to control for the effects of other factors and covariates and to set up linear contrasts, beyond the simple two‐group comparison.…”

“…The L 2 N model is implemented in an R‐driven, user‐friendly, graphical interface called DVX (Bar & Schifano, ), for differential variation and expression analysis. The software also includes an implementation of a three‐way normal mixture model, which we refer to as N 3 , similar to the one proposed in Bar et al ().…”

Detecting Differentially Variable MicroRNAs via Model-Based Clustering

“…The ten equal variance tests are (1) the F test (denoted as F), (2) Ahn and Wang's score test [ 12 ] (denoted as AW), (3) Phipson and Oshlack's AD test [ 13 ] (denoted as PO.AD), (4) Phipson and Oshlack's SQ test [ 13 ] (denoted as PO.SQ), (5) Levene's test [ 14 ] (denoted as L), (6) Brown and Forsythe's test [ 15 ] (denoted as BF), (7) trimmed-mean-based Levene's test [ 15 ] (denoted as Ltrim), (8) improved AW test based on Levene's test [ 16 ] (denoted as iL), (9) improved AW test based on the BF test [ 16 ] (denoted as iBF), and (10) improved AW test based on the trimmed-mean-based Levene's test [ 16 ] (denoted as iTrim). The remaining six methods are based on Bar et al's [ 7 ] N3 model and Bar and Schifano's [ 8 ] L 2 N model. Both N3 and L 2 N models have been implemented in the R package DVX [ 8 ].…”

“…The remaining six methods are based on Bar et al's [ 7 ] N3 model and Bar and Schifano's [ 8 ] L 2 N model. Both N3 and L 2 N models have been implemented in the R package DVX [ 8 ]. For both N3 and L 2 N, DVX outputs raw p values, q values, and posterior probabilities p gk that the probe g belongs to cluster k given its expression profile and estimated model parameters, k = 1, 2, 3.…”

“…Bar et al [ 7 ] introduced a bivariate model (N3) to account for both differential expression and differential variation in high-throughput data analysis, by assuming that both means and variances follow three-component mixture distributions. Bar and Schifano [ 8 ] proposed a unified three-component mixture model, the L 2 N model, that can be used to detect either differential expression (mean) or differential variation, by modeling the differences of means and variances (dispersions) between two groups of samples. In the L 2 N model, one log-normal component is used to fit underexpressed (dispersed) probes, one log-normal component is used to fit overexpressed (dispersed) probes, and one normal component is used to fit nondifferentially expressed (dispersed) probes.…”