https://github.com/riborings/mediation

https://github.com/riborings/mediation

Last synced: 7 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/riborings/mediation
• Owner: RiboRings
• Created: 2024-02-22T19:00:54.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-03-10T17:31:29.000Z (8 months ago)
• Last Synced: 2024-03-11T16:55:29.026Z (8 months ago)
• Language: HTML
• Size: 1.24 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Mediation Analysis in the miaverse

There are two functions available in the funs.R script:

- `mediate_assay` perfoms mediation analysis for every taxon in an assay of choice or for every component of a reducedDim of choice

- `mediate_coldata` performs mediation analysis for a variable in the colData

In either case, the treatment and the outcome variables are taken from the colData.

## Prepare for examples

The following code can also be found in the tutorial.R script. Some example output (both statistical tables and plots) are stored in

the results folder.

```

library(mia)

library(dplyr)

library(tidyr)

library(mediation)

library(scater)

library(ggplot2)

library(ggrepel)

source("./funs.R")

data("enterotype", package = "mia")

tse <- enterotype

tse <- tse[ , tse$ClinicalStatus %in% c("healthy", "obese")]

tse <- subsetByPrevalentFeatures(tse,

                                 prevalence = 0.1)

tse <- transformAssay(tse, method = "relabundance")

```

## Example: alpha diversity as mediator

Here the question is: does shannon diversity mediate the effect of age on clinical status?

```

tse <- estimateDiversity(tse, index = "shannon")

med_out <- mediate_coldata(tse,

                           outcome = "ClinicalStatus",

                           treatment = "Age",

                           mediator = "shannon",

                           family = binomial("logit"),

                           boot = TRUE, sims = 1000)

summary(med_out)

plot(med_out)

```

## Example: reduced dimension as mediator

Here the question is: does any of the reduced components mediate the effect of gender on clinical status?

```

tse <- runMDS(tse,

              assay.type = "relabundance",

              ncomponents = 5)

med_df <- mediate_assay(tse,

                        outcome = "ClinicalStatus",

                        treatment = "Gender",

                        dim.type = "MDS",

                        family = binomial("logit"),

                        boot = TRUE, sims = 300)

p <- ggplot(med_df, aes(x = ACME_estimate,

                        y = -log10(ACME_pval))) +

  geom_point() +

  scale_x_continuous(limits = c(-0.11, 0.11)) +

  geom_hline(yintercept = -log10(0.05),

             linetype = "dashed",

             colour = "red") +

  geom_text_repel(data = subset(med_df, ACME_pval < 0.05),

                  aes(label = Mediator),

                  colour = "black") +

  labs(x = "Mediated Effect",

       y = "Negative Decadic Logarithm of P-value") +

  theme_classic()

p

```

## Example: taxa in assay as mediator

Here the question is: does any of taxa mediate the effect of age on clinical status?

```

med_df <- mediate_assay(tse,

                        outcome = "ClinicalStatus",

                        treatment = "Age",

                        assay.type = "relabundance",

                        family = binomial("logit"),

                        boot = TRUE, sims = 100)

p <- ggplot(med_df, aes(x = ACME_estimate,

                        y = -log10(ACME_pval))) +

  geom_point() +

  scale_x_continuous(limits = c(-0.05, 0.05)) +

  geom_hline(yintercept = -log10(0.05),

             linetype = "dashed",

             colour = "red") +

  geom_text_repel(data = subset(med_df, ACME_pval < 0.05),

                  aes(label = Mediator),

                  colour = "black") +

  labs(x = "Mediated Effect",

       y = "Negative Decadic Logarithm of P-value") +

  theme_classic()

p

```