olink_anova_posthoc {OlinkAnalyze}

Function which performs an ANOVA posthoc test per protein.

Description

Performs a post hoc ANOVA test using emmeans::emmeans with Tukey p-value adjustment per assay (by OlinkID) for each panel at confidence level 0.95. See olink_anova for details of input notation.

The function handles both factor and numerical variables and/or covariates. The posthoc test for a numerical variable compares the difference in means of the outcome variable (default: NPX) for 1 standard deviation difference in the numerical variable, e.g. mean NPX at mean(numerical variable) versus mean NPX at mean(numerical variable) + 1*SD(numerical variable).

Usage

olink_anova_posthoc(
  df,
  olinkid_list = NULL,
  variable,
  covariates = NULL,
  outcome = "NPX",
  model_formula,
  effect,
  effect_formula,
  mean_return = FALSE,
  post_hoc_padjust_method = "tukey",
  verbose = TRUE
)

Arguments

df

NPX data frame in long format with at least protein name (Assay), OlinkID, UniProt, Panel and a factor with at least 3 levels.

olinkid_list

Character vector of OlinkID's on which to perform post hoc analysis. If not specified, all assays in df are used.

variable

Single character value or character array. Variable(s) to test. If length > 1, the included variable names will be used in crossed analyses . Also takes ':' notation.

covariates

Single character value or character array. Default: NULL. Covariates to include. Takes ':' or '*' notation. Crossed analysis will not be inferred from main effects.

outcome

Character. The dependent variable. Default: NPX.

model_formula

(optional) Symbolic description of the model to be fitted in standard formula notation (e.g. "NPX~A*B"). If provided, this will override the outcome, variable and covariates arguments. Can be a string or of class stats::formula().

effect

Term on which to perform post-hoc. Character vector. Must be subset of or identical to variable.

effect_formula

(optional) A character vector specifying the names of the predictors over which estimated marginal means are desired as defined in the emmeans package. May also be a formula. If provided, this will override the effect argument. See ?emmeans::emmeans() for more information.

mean_return

Boolean. If true, returns the mean of each factor level rather than the difference in means (default). Note that no p-value is returned for mean_return = TRUE and no adjustment is performed.

post_hoc_padjust_method

P-value adjustment method to use for post-hoc comparisons within an assay. Options include tukey, sidak, bonferroni and none.

verbose

Boolean. Default: True. If information about removed samples, factor conversion and final model formula is to be printed to the console.

Value

A "tibble" of posthoc tests for specified effect, arranged by ascending adjusted p-values. Columns include:

• Assay: "character" Protein symbol

• OlinkID: "character" Olink specific ID

• UniProt: "character" UniProt ID

• Panel: "character" Name of Olink Panel

• term: "character" term in model

• contrast: "character" the groups that were compared

• estimate: "numeric" difference in mean NPX between groups

• conf.low: "numeric" confidence interval for the mean (lower end)

• conf.high: "numeric" confidence interval for the mean (upper end)

• Adjusted_pval: "numeric" adjusted p-value for the test

• Threshold: "character" if adjusted p-value is significant or not (< 0.05)

Examples

library(dplyr)

npx_df <- npx_data1 %>% filter(!grepl('control',SampleID, ignore.case = TRUE))

#Two-way ANOVA, one main effect (Site) covariate.
#Results in model NPX~Treatment*Time+Site.
anova_results <- olink_anova(df = npx_df,
                             variable=c("Treatment:Time"),
                             covariates="Site")

#Posthoc test for the model NPX~Treatment*Time+Site,
#on the interaction effect Treatment:Time with covariate Site.

#Filtering out significant and relevant results.
significant_assays <- anova_results %>%
filter(Threshold == 'Significant' & term == 'Treatment:Time') %>%
select(OlinkID) %>%
distinct() %>%
pull()

#Posthoc, all pairwise comparisons
anova_posthoc_results <- olink_anova_posthoc(npx_df,
variable=c("Treatment:Time"),
covariates="Site",
olinkid_list = significant_assays,
effect = "Treatment:Time")


#Posthoc, treated vs untreated at each timepoint, adjusted for Site effect
anova_posthoc_results <- olink_anova_posthoc(npx_df,
model_formula = "NPX~Treatment*Time+Site",
olinkid_list = significant_assays,
effect_formula = "pairwise~Treatment|Time")