‘multiOmics data analysis, integration, and visualisation protocol

• Data integration and modelling with R
• Systems biology analysis and visualization pipeline in R

Data management framework

• pISA-tree on GitHub
  • Petek, M., Zagorščak, M., Blejec, A. et al. pISA-tree - a data management framework for life science research projects using a standardised directory tree. Sci Data 9, 685 (2022). https://doi.org/10.1038/s41597-022-01805-5

Expected measurements

• one or multiple genotypes
• under single and multiple abiotic/biotic stressors
• experiment duration: XY hours, days, ... : time-series experimental design
• tissue: single or multiple
• Omics' strategies:
  • Hormonomics
  • Transcriptomics
  • Proteomics
  • Metabolomics
  • Phenomics

Analysis steps:

1. Design Phenodata, a master experimental design table describing samples for analysis, prior to sample collection according to good data management practice
2. Data preprocessing and overall inspection

• detection of outliers and faulty measurements

  • {gplot2}
  • {rgl}
  • {raster}
  • {multcompView}
  • {gridExtra}

• data transformation (if needed)

  • {multimode}
  • {fitdistrplus}
  • {caret}
  • {glmnet}
  • {MASS}
  • {BIGL}
  • {robustbas}
  • {preprocessCore}
  • {compositions}
  • {mgcv}

• interpolation

  • point-to-point
    • approxfun {stats} - Returns a list of points which linearly interpolate given data points, or a function performing the linear (or constant) interpolation.
  • polynomial
    • predict {stats} - A generic function for predictions from the results of various model fitting functions
    • boxplot.stats {grDevices} - Box Plot Statistics
    • mad {BiocGenerics} - Compute the median absolute deviation for a vector
    • aq.plot {mvoutlier} - Adjusted quantile plots for multivariate outlier detection

• extrapolation

  

• imputation

  • for qPCR see Baebler, Š., Svalina, M., Petek, M. et al. quantGenius: implementation of a decision support system for qPCR-based gene quantification. BMC Bioinformatics 18, 276 (2017). https://doi.org/10.1186/s12859-017-1688-7

3. Statistical analysis of individual omics data layers

   • ggplot {ggplot2} - various plots, https://r-graphics.org/chapter-ggplot2
   • corr.test {psych} - Find the correlations, sample sizes, and probability values between elements of a matrix or data.frame
   • cor.plot {psych} - Create an image plot for a correlation or factor matrix
   • pairs.panels {psych} - SPLOM, histograms and correlations for a data matrix
   • rcorr {Hmisc} - Matrix of Correlations and P-values
   • heatmaply_cor {heatmaply} - Cluster heatmap based on plotly
   • corrplot {corrplot} - A visualization of a correlation matrix
   • pheatmap {pheatmap} - A function to draw clustered heatmaps
   • t_test {rstatix}
   • ggdotplot, ggviolin {ggpubr}
   • metaMDS {vegan} - Nonmetric Multidimensional Scaling with Stable Solution from Random Starts, Axis Scaling and Species Scores
   • {limma} for e.g. non-targeted Proteomics, RNA-seq, ..
     • limma::lmFit - Linear Model for Series of Arrays
     • limma::makeContrasts - Construct Matrix of Custom Contrasts
     • limma::contrasts.fit - Compute Contrasts from Linear Model Fit
     • limma::eBayes - Empirical Bayes Statistics for Differential Expression
     • limma::decideTests - Multiple Testing Across Genes and Contrasts
     • limma::topTable - Table of Top Genes from Linear Model Fit

4. Correlation based network inference within each omics level

   • Leave-One-Out graphs
     • qgraph {qgraph}
     • igraph
   • Lioness

   Since results from both methods heavily depend on selected thresholds, Lioness node and edge selection using FDR being even more sensitive on correlation difference cut-off, we suggest to use an automated graph thresholding approach.

5. Integration across different omics datasets

• Canonical Correlation Analysis
• N-Integration Discriminant Analysis with DIABLO
  • {mixOmics}
    • block.splsda {mixOmics} N-integration and feature selection with Projection to Latent Structures models (PLS) with sparse Discriminant Analysis
    • plotDiablo {mixOmics} Graphical output for the DIABLO framework
    • plotVar {mixOmics} Plot of Variables
    • plotIndiv {mixOmics} Plot of Individuals (Experimental Units)
    • plotArrow {mixOmics} Arrow sample plot
    • circosPlot {mixOmics} circosPlot for DIABLO
    • cimDiablo {mixOmics} Clustered Image Maps (CIMs) ("heat maps") for DIABLO
    • network {mixOmics} Relevance Network for (r)CCA and (s)PLS regression
• Leave-One-Out graphs
  • qgraph {qgraph}
  • igraph

6. Integration of data with prior knowledge
   • igraph
   • DiNAR subApps
   • Cytoscape
   • https://skm.nib.si/

Start of the analysis:

• Data is expected to be arranged within data management framework, with complete and descriptive metadata files, including Phenodata file.
• 'Omics files are expected to be preprocessed (see suggestions in Step 2).
• Minimal input files can be found within './input' directory.
• For Step 3: Statistical analysis of individual omics data layers run script 01_Step3.Rmd
• For Step 4: Correlation based network inference within/between each omics level run script 02_Step4.Rmd
• For Step 5: Integration across different omics datasets run script 03_Step5.Rnw

For more info see multiOmics_data_analysis_Protocol