Table 2 Description of multi-omics models for estimating variance components.

From: Integrative multi-omics analysis to gain new insights into COVID-19

Analysis scope

Model

Description

Model 1. Single-omics analysis: estimate variance components using omics layers separately

\(\mathbf{Y}=\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}\) is the vector of COVID-19 outcome, \(\mathbf{G}\) is the vector of genome effect, and \(\varvec{\upepsilon}\) is the vector of residuals, \(\mathbf{T}\) is the vector of transcriptome effect, \(\mathbf{M}\) is the vector metabolome effect, \(\mathbf{E}\) is the vector of exposome effect. G\(\mathbf{x}\mathbf{E}\) is the vector of gene by environment interaction, T\(\mathbf{x}\mathbf{E}\) is the vector of transcriptome by environment interaction, M\(\mathbf{x}\mathbf{E}\) is the vector of metabolome by environment interaction. r\(\mathbf{G},\mathbf{E}\) is the vector of covariance between genome and exposome, r\(\mathbf{T},\mathbf{E}\) is the vector of covariance between genome and exposome, and r\(\mathbf{M},\mathbf{E}\) is the vector of covariance between metabolome and exposome.

G\(\mathbf{x}\mathbf{T}\) is the vector of genome by transcriptome interaction, r\(\mathbf{G},\mathbf{T}\) is the vector of covariance between genome and transcriptome.

T\(\mathbf{x}\mathbf{M}\) is the vector of transcriptome by metabolome interaction, r\(\mathbf{T},\mathbf{M}\) is the vector of covariance between transcriptome and metabolome.

\(\mathbf{Y}=\mathbf{T}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\varvec{\upepsilon}\)

Model 2: Pairwise omics analysis: joint analysis of genome, transcriptome, and metabolome

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{M}+\mathbf{E}+\varvec{\upepsilon}\)

Model 3. Joint model fitting each omics data type with exposome

\(\mathbf{Y}=\mathbf{G}+\mathbf{E}+\mathbf{G}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{E}+\mathbf{T}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{M}+\mathbf{E}+\mathbf{M}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{E}+\mathbf{r}\mathbf{G},\mathbf{E}+\mathbf{G}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{E}+\mathbf{r}\mathbf{T},\mathbf{E}+\mathbf{T}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{M}+\mathbf{E}+\mathbf{r}\mathbf{M},\mathbf{E}+\mathbf{M}\mathbf{x}\mathbf{E}+\varvec{\upepsilon}\)

Model 4. Joint models fitting genome and transcriptome and their interplay

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\mathbf{G}\mathbf{x}\mathbf{T}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\mathbf{r}\mathbf{G},\mathbf{T}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\mathbf{G}\mathbf{x}\mathbf{T}+\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{G}+\mathbf{T}+\mathbf{G}\mathbf{x}\mathbf{T}+\mathbf{r}\mathbf{G},\mathbf{T}+\mathbf{E}+\varvec{\upepsilon}\)

Model 5. Joint models fitting transcriptome and metabolome and their interplay

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\mathbf{E}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\mathbf{T}\mathbf{x}\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\mathbf{r}\mathbf{T},\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\mathbf{T}\mathbf{x}\mathbf{M}+\varvec{\upepsilon}\)

\(\mathbf{Y}=\mathbf{T}+\mathbf{M}+\mathbf{T}\mathbf{x}\mathbf{M}+\mathbf{r}\mathbf{T},\mathbf{M}+\mathbf{E}+\varvec{\upepsilon}\)

  1. Note: An integrated omics model of the genome, transcriptome, and exposome of COVID-19 was analysed using data from 107,857 participants, but the sample size was reduced to 23,520 to account for the proportion of cases with metabolomics data.