SDAMS

msqrob2 provides a robust linear mixed model framework for assessing differential abundance in MS-based Quantitative proteomics experiments. Our workflows can start from raw peptide intensities or summarised protein expression values. The model parameter estimates can be stabilized by ridge regression, empirical Bayes variance estimation and robust M-estimation. msqrob2's hurde workflow can handle missing data without having to rely on hard-to-verify imputation assumptions, and, outcompetes state-of-the-art methods with and without imputation for both high and low missingness. It builds on QFeature infrastructure for quantitative mass spectrometry data to store the model results together with the raw data and preprocessed data.

A streamlined tool provides a graphical user interface for quality control based signal drift correction (QC-RFSC), integration of data from multi-batch MS-based experiments, and the comprehensive statistical analysis in metabolomics and proteomics.

The package provides methods of combining the graph structure learning and generalized least squares regression to improve the regression estimation. The main function sparsenetgls() provides solutions for multivariate regression with Gaussian distributed dependant variables and explanatory variables utlizing multiple well-known graph structure learning approaches to estimating the precision matrix, and uses a penalized variance covariance matrix with a distance tuning parameter of the graph structure in deriving the sandwich estimators in generalized least squares (gls) regression. This package also provides functions for assessing a Gaussian graphical model which uses the penalized approach. It uses Receiver Operative Characteristics curve as a visualization tool in the assessment.

The Power Law Global Error Model (PLGEM) has been shown to faithfully model the variance-versus-mean dependence that exists in a variety of genome-wide datasets, including microarray and proteomics data. The use of PLGEM has been shown to improve the detection of differentially expressed genes or proteins in these datasets.

Latent variable modeling with Principal Component Analysis (PCA) and Partial Least Squares (PLS) are powerful methods for visualization, regression, classification, and feature selection of omics data where the number of variables exceeds the number of samples and with multicollinearity among variables. Orthogonal Partial Least Squares (OPLS) enables to separately model the variation correlated (predictive) to the factor of interest and the uncorrelated (orthogonal) variation. While performing similarly to PLS, OPLS facilitates interpretation. Successful applications of these chemometrics techniques include spectroscopic data such as Raman spectroscopy, nuclear magnetic resonance (NMR), mass spectrometry (MS) in metabolomics and proteomics, but also transcriptomics data. In addition to scores, loadings and weights plots, the package provides metrics and graphics to determine the optimal number of components (e.g. with the R2 and Q2 coefficients), check the validity of the model by permutation testing, detect outliers, and perform feature selection (e.g. with Variable Importance in Projection or regression coefficients). The package can be accessed via a user interface on the Workflow4Metabolomics.org online resource for computational metabolomics (built upon the Galaxy environment).

This package unifies access to Statistal Modeling of Omics Data. Across linear modeling engines (lm, lme, lmer, limma, and wilcoxon). Across coding systems (treatment, difference, deviation, etc). Across model formulae (with/without intercept, random effect, interaction or nesting). Across omics platforms (microarray, rnaseq, msproteomics, affinity proteomics, metabolomics). Across projection methods (pca, pls, sma, lda, spls, opls). Across clustering methods (hclust, pam, cmeans). Across survival methods (coxph, survdiff, coin). It provides a fast enrichment analysis implementation.