An Introduction to the OmicsLonDA Package

Ahmed A. Metwally

2021-10-26


OmicsLonDA (Omics Longitudinal Differential Analysis) is a statistical framework that provides robust identification of time intervals where omics features are significantly different between groups. OmicsLonDA is based on 5 main steps:

  1. Adjust measurements based on each subject’s specific baseline
  2. Global testing using linear mixed-effect model to select candidate features and covariates for time intervals analysis
  3. Fitting smoothing spline regression model
  4. Monte Carlo permutation to generate the empirical distribution of the test statistic
  5. Inference of significant time intervals of omics features.

Getting Started

Prerequisites

Installation

Install the latest version of OmicsLonDA from Bioconductor:

Example

The measurment matrix represents count/intensity of features from an omic experiment. Columns represent various samples from different subjects longitudinally. Rows represent various features. Here is an example:

The metadata dataframe contains annotations for each sample. Most impotantly it should have at least: (a) “Subject”: which denote from which subject this sample is coming from, (b) “Group”: which represents which group this sample is from (eg., healthy, disease, etc), (c) “Time”: which represents the collection time of the corresponding sample. Here is an example:

Create SummarizedExperiment object

se_ome_matrix = as.matrix(omicslonda_data_example$ome_matrix)
se_metadata = DataFrame(omicslonda_data_example$metadata)
omicslonda_se_object = SummarizedExperiment(assays=list(se_ome_matrix),
                                            colData = se_metadata)

Adjust for baseline using CLR

omicslonda_se_object_adjusted = adjustBaseline(se_object = omicslonda_se_object)

Measurments after baseline adjustment

assay(omicslonda_se_object_adjusted)[1:5, 1:5]
#>           Sample_2   Sample_3   Sample_4   Sample_5  Sample_6
#> Feature_1        0 -0.5513775 -0.5004965 -0.6113461 0.5933772
#> Feature_2        0 -0.3562154 -0.3074287 -0.3833650 0.4981392
#> Feature_3        0 -0.2574379 -0.4259317 -0.2364567 0.5127220
#> Feature_4        0 -0.4224943 -0.4646099 -0.3163138 0.4839143
#> Feature_5        0 -0.6515421 -0.3709867 -0.5105080 0.4529296

Visualize first feature

omicslonda_test_object = omicslonda_se_object_adjusted[1,]
visualizeFeature(se_object = omicslonda_test_object, text = "Feature_1",
                 unit = "days", ylabel = "Normalized Count", 
                 col = c("blue", "firebrick"), prefix = tempfile())
Visualize first feature

Visualize first feature

Specify interval bounds

points = seq(1, 500, length.out = 500)

Run OmicsLonDA on the first feature

res = omicslonda(se_object = omicslonda_test_object, n.perm = 10,
                 fit.method = "ssgaussian", points = points, text = "Feature_1",
                 parall = FALSE, pvalue.threshold = 0.05, 
                 adjust.method = "BH", time.unit = "days",
                 ylabel = "Normalized Count",
                 col = c("blue", "firebrick"), prefix = tempfile())

Visualize fitted spline of the first feature

visualizeFeatureSpline(se_object = omicslonda_test_object, omicslonda_object = res, fit.method = "ssgaussian",
                        text = "Feature_1", unit = "days",
                        ylabel = "Normalized Count", 
                        col = c("blue", "firebrick"),
                        prefix = "OmicsLonDA_example")