Demonstration of various classes in R

R_Bioc_tidy_data.R

# dataframes vs lm S3 vs Bioc S4
# Michael Love
# Nov 1 2023

dat <- data.frame(genotype=c("wt","wt","mut","mut"),
                  count=c(10,20,30,40),
                  score=c(-1.2,0,3.4,-5),
                  gene=c("Abc","Abc","Xyz","Xyz"))
library(tibble)
dat |> as_tibble()
dat <- dat |> as_tibble()

# linear regression
lm(score ~ count, data=dat)
fit <- lm(score ~ count, data=dat)
class(fit)
methods(class="lm")

names(fit)
fit$coefficients
summary(fit)

library(broom)
fit |> tidy()
fit |> glance()

library(dplyr)
library(tidyr)
library(purrr)
sim <- tibble(g=rep(1:3,each=30),y=rnorm(90),x=rnorm(90))
sim |>
  nest(.by=g) |>
  mutate(fit = map(data, \(d) lm(y ~ x, data=d)),
         stats = map(fit, glance)) |>
  unnest(stats)

# Bioconductor objects have:
# slots, validation, special methods (dispatch)

library(SummarizedExperiment)

x <- list(counts=matrix(1:4 * 10, ncol=2),
          score=matrix(-2:1, ncol=2))
rowdata <- data.frame(gene=c("Abc","Xyz"))
coldata <- data.frame(genotype=c("wt","mut"))

rownames(x[[1]]) <- rownames(x[[2]]) <- rowdata$gene
colnames(x[[1]]) <- colnames(x[[2]]) <- coldata$genotype
se <- SummarizedExperiment(assays=x, rowData=rowdata, colData=coldata)

assay(se, "counts")
rowData(se)
colData(se)

class(se)
methods(class = class(se))
slotNames(se)

se_sub <- se[ , se$genotype == "wt" ]
assay(se_sub, "counts")

# validation
rowdata2 <- rowdata
rownames(rowdata2) <- c("Xyz","Abc")
SummarizedExperiment(assays=x, rowData=rowdata2, colData=coldata)

# special methods
cbind(se, se)