In isglobal-brge/dsExposomeClient: DataSHIELD client site Exposome functions

When the ExposomeSet object is on the study the server, the function ds.exwas is in charge of performing the ExWAS. As an example let's present a situation where we are interested in associating each exposure with blood_pre phenotype (e.g outcome) using the sex as a covariate (e.g. adjusted model). This study would be done as follows

exwas_results <- ds.exwas("blood_pre ~ sex", Set = "exposome_object", family = "gaussian")
head(exwas_results$exwas_results)

This previous formula will fit linear models between the exposures and phenotype as follows:

    phenotype ~ exposure_1 + covar1 + ... + covarN
    phenotype ~ exposure_2 + covar1 + ... + covarN
    phenotype ~ exposure_3 + covar1 + ... + covarN
    ...
    phenotype ~ exposure_M + covar1 + ... + covarN

The model is written as a string, where the left side term is the phenotype, and the right term are the covariates (e.g. variables to be adjusted for). In the case of having more covariates proceed as: phenotype ~ cov1 + cov2 + ... + covN. A crude model is fitted using phenotype ~ 1.

It's important to note that if the family argument does not match with the nature of the data availalbe in the phenotype, the ExWAS will fail (Example: family argument should be binomial for binary outcomes and gaussian for continuous).

To visualize the results from the ExWAS, the function ds.plotExwas() can be used. It takes the output of ds.exwas and creates two different visualizations depending on the argument type.

• A Manhattan-like plot with the p-values of the association between each exposure and the outcome colored by families of exposures. The vertical red line stands for the significant level corrected by the effective number of tests.

ds.plotExwas(exwas_results, type="manhattan")

• A plot showing the effects (beta values) and their confidence intervals of a given model can be obtained with:

ds.plotExwas(exwas_results, type="effect")

Inverse ExWAS

We can also carry out an 'inverse' ExWAS where exposures are regressed on a given phenotype and some adjusting covariates. In other words, we are estimating an inverse/reverse model as in a traditional ExWAS:

    exposure_1 ~ phenotype + covar1 + ... + covarN
    exposure_2 ~ phenotype + covar1 + ... + covarN
    exposure_3 ~ phenotype + covar1 + ... + covarN
    ...
    exposure_M ~ phenotype + covar1 + ... + covarN

The function to perform this analysis is ds.invExWAS(). In that case the formula is not used and the models can be fitted by

inverse_exwas_results <- ds.invExWAS("exposome_object", phenotype =  "blood_pre", covariables = "sex")
head(inverse_exwas_results$exwas_results)

Crude models are obtained by removing the covariables argument and models adjusted for more than one covariable are fitted by writing covariables = c("covar1", "covar2", ..., "covarN").

The same plotting functions can be used with the object returned by the inverse ExWAS analysis.

ds.plotExwas(inverse_exwas_results, "manhattan")

isglobal-brge/dsExposomeClient documentation built on March 5, 2024, 12:26 p.m.