Table 1.
Summary of samples and characteristics of the datasets used.
Fig 1.
Analysis of the parameters on the glioma dataset.
A1 and A2: A contour plot of the median of the inner AVEresult of an SRGCCA with different tau values for each block (GE, gene expression of the host transcriptome, CGH (comparative genomic hybridization) for the copy number variation and y for the location). Higher tau normally increases the inner AVE, Schäfer’s approximation is marked with the red vertical line. B: First two dimensions of the superblock on the glioma dataset. The first two components of the superblock within the best model, according to the inner AVE from the glioma dataset. C: First dimensions of the host transcriptome and the CGH block of models on the glioma dataset are represented. Comparison of the different models by visualizing the first components of the host transcriptome gene expression (GE) and the copy number variation (CGH) blocks from the glioma dataset. Each point represents a sample (colored by location). Cort: supratentorial, dipg: brain stem, midl: central nuclei.
Fig 2.
First dimensions of the host transcriptome and the microbiome block of models on the Crohn’s disease ulcerative colitis/ dataset.
Comparison of the models that better explained the interaction between the microbiome and the host transcriptome data on the CD/UC dataset. Each point represents a sample colored according to a characteristic: A) samples are colored by disease type, CD Crohn’s disease, Ctrl, control; DCtrl diseased control, inflamed but not from IBD patients, UC ulcerative Colitis; and B, by location, colon or ileum, on the first components of the host transcriptome and the microbiome. Better models separate samples by tissue location using the host transcriptome component.
Fig 3.
First dimensions of the host transcriptome and the microbiome block of models on the hematopoietic stem cell transplant Crohn’s disease dataset.
Comparison of the models that better explained the interaction between the microbiome and the host transcriptome data on the HSCT CD dataset. Each point represents a sample (colored by disease status): A, non-CD (Control) or CD; and B, by location, colon or ileum, on the first components of the host transcriptome and the microbiome. Better models separate samples by tissue location by the host transcriptome component and the diseased and controls samples by the microbiome component.
Fig 4.
First dimensions of the host transcriptome and the microbiome block of models on the pouchitis dataset.
Comparison of the models vis-à-vis on the pouchitis dataset by the first component of the host transcriptome and the microbiome from the HSCT CD dataset. Each point represents a sample colored by sex (A), where females are in red and males in blue, and by location (B), where the pouch is the red, and PPI is the pre-pouch ileum. The samples do not show a sex-specific pattern but on the best models the host transcriptome partially separates pouch and pre-pouch ileum samples.
Fig 5.
UpSet plot of the of the models on the hematopoietic stem cell transplant Crohn’s disease dataset.
The heights of the bars represent the genes (A) or OTUs (B) shared between the models selected by the points; 30 intersections are shown.
Fig 6.
Bootstrap results of three models on the hematopoietic stem cell transplant Crohn’s disease dataset.
Variance of AVE using the same samples on three models with the HSCT CD dataset. Each point shows the AVE for each analysis performed. The brighter colors reflect the result of this model on the original data (including all samples). Dispersion on the bootstrapped samples is reduced as a model more accurately represents the relationships present on the dataset.
Table 2.
Bootstrapped mean and standard deviation of inner and outer AVE values on the HSCT CD dataset.
Table 3.
Bootstrapped mean and standard deviation of inner and outer AVE values on the pouchitis dataset.
Fig 7.
Multiple co-inertia analysis and area under the curve for the location of the Crohn’s disease/ulcerative colitis, the hematopoietic stem cell transplant Crohn’s disease and pouchitis dataset.
A, B, C plots are the results of applying multiple co-inertia analysis (MCIA) where the horizontal and vertical axis represent the synthetic variable 1 and 2 respectively. D, E, F plots are the area under the curve (AUC) for all the methods applied on this dataset. The first row (A, D) is the analysis of CD/UC dataset, the second one (B, E) the HSCT CD dataset, and the third one (C, F) the pouchitis dataset.