In ismms-himc/LuminexTools:

knitr::opts_chunk$set(echo = TRUE)
library(readxl)
library(tidyverse)
library(magrittr)
library(tidyr)
library(SummarizedExperiment)
library(patchwork)
library(LuminexTools)

a. read file name

folder/ data/ file1 file2 Luminex_QA_temp.Rmd

f.list <- list.files("data", full.names = T)
names(f.list) <- f.list

f.list <- lapply(f.list, LuminexTools::read_lmx, analyt_start_row = 28)

b.normalized by bdg_norm_lmx_multi(bridge.str, data.ls, ref_bach)

assign common reference samples and inspect for

commonsample <- c("QC", "HD")
names(f.list)

c. Normalization

Normalization do not take account of plate-wise global median normalization, for the reason that luminex is quantitive assay.
+ use case 1: ref_batch = NULL, global plate median of commonsample normalization will be applied.
+ use case 2: ref_batch = some_names(f.list), one of the file name need to be used as the anchor plate. inspect file name using names(f.list).
+ use case 3: bridge.str = commonsample, commonsample will be used to compute normalization factor.

f.list <- bdg_norm_lmx_multi(bridge.str = commonsample, data.ls = f.list)
cmb <- LuminexTools::cmb_lmx_se(f.list)

d. save file output

data_default: data as is.
data_imputed: below detection is replaced by LOD.
normed: normalized.(option).

# save file 
LuminexTools::save_lmx_csv(cmb, pre_fix = ".", assay2save = c("data_default", "data_imputed", "normed"))

QA.

prepare data for plot.

lod <- data.frame(analyt = rownames(cmb),
                  lod = cmb@elementMetadata$LOD)

hod <- data.frame(analyt = rownames(cmb),
                  hod = cmb@elementMetadata$HOD)

##--combined plates
plotdf <- data.frame(Sample = cmb$Sample,
                     plate = gsub("(.*Plate|_Results.*)", "", cmb$File),
                     t(cmb@assays@data$data_imputed))%>%
  mutate(type = case_when(
    grepl("HD", Sample)  ~ "HD",
    grepl("QC", Sample) ~ "QC",
    TRUE ~ "StudySample"
  ))%>%
  mutate(sample = gsub("(^THMA02_|_R|_extra| extra)", "", Sample),
         timepoint = gsub(".*_", "", sample),
         sample = gsub("_.*", "", sample))


# for plot
plotdf <- plotdf %>%
  gather(-Sample, -sample, -type, -plate, -timepoint, key = "analyte", value = "value")

e. plot for QC, HD, Studysamply by type

plot.ls <- lapply(unique(plotdf$analyte), function(x){

  tit = x
  plotdf%>%
  #filter(Sample != "HDPlasma_1017")%>%
  dplyr::filter(analyte == x)%>%
  ggplot(aes(type, log1p(as.numeric(value))))+
  geom_boxplot(aes(color = plate))+
  geom_point(aes(color = plate),
             position = position_jitterdodge(dodge.width = 0.75, jitter.width = 0.2))+
  ggrepel::geom_text_repel(aes(label = Sample), size = 2.5)+
  #geom_hline(yintercept = lod$lod[lod$analyt == x]%>%log1p(), color = "blue", alpha = 0.5)+
  #geom_hline(yintercept = hod$hod[lod$analyt == x]%>%log1p(), color = "red", alpha = 0.5)+
  geom_segment(aes(x = 0.5, xend = 3.5, y = log1p(lod), yend = log1p(lod)),
               color = "blue",
               data = lod %>% dplyr::filter(analyt == x))+
  geom_segment(aes(x = 0.5, xend = 3.5, y = log1p(hod), yend = log1p(hod)),
             color = "red",
             data = hod %>% dplyr::filter(analyt == x))+
  labs(title = tit, y = "log1p Conc.", x = "")+
  #labs(y = "log1p Conc.", x = "")+
  theme_bw(base_size = 12)+
  #facet_grid(analyte ~ plate)+
  theme(axis.text.x = element_text(angle = 90, hjust = 0.5, vjust = 0.5, size = 8))
})

# per 4x6 panel plot
for(i in 0 : (nrow(cmb)/24-1)) {
  p <- wrap_plots(plot.ls[(i*24+1) : min(nrow(cmb), (i*24+24))], ncol = 6)
  p_name <- paste0("trace_plot_", i, ".png" )
  ggsave(filename =p_name, plot = p, width = 15, height = 7, dpi = 750) # change size res as needed
}

f. plate out of range sample count summary

sub_tit = paste("total sample on plates:", ncol(cmb))

data.frame(Sample = cmb$Sample,
                     t(cmb@assays@data$data_default))%>%
  gather(-Sample, key = "analyte", value = "value")%>%
  group_by(analyte)%>%
  summarise(out_of_range_count = sum(!is.finite(value)))%>%
  ggplot(aes(analyte, out_of_range_count))+
  geom_bar(stat = "identity", alpha = 0.7)+
  geom_text(aes(label = out_of_range_count))+
  coord_flip()+
  labs(subtitle = sub_tit)+
  theme_bw()

g. timepoint plot

plot.ls <- lapply(unique(plotdf$analyte), function(x){

  tit = x
  plotdf%>%
  dplyr::filter(type != "HD", type != "QC")%>%
  dplyr::filter(analyte == x)%>%
  ggplot(aes(timepoint, log1p(as.numeric(value))))+
  geom_point()+
  geom_line(aes(group = sample))+
  labs(title = tit, y = "log1p Conc.", x = "")+
  #labs(y = "log1p Conc.", x = "")+
  theme_bw(base_size = 12)+
  #facet_grid(analyte ~ plate)+
  theme(axis.text.x = element_text(angle = 90, hjust = 0.5, vjust = 0.5, size = 8))
})

# per 4x6 panel plot
for(i in 0 : (nrow(cmb)/24-1)) {
  p <- wrap_plots(plot.ls[(i*24+1) : min(nrow(cmb), (i*24+24))], ncol = 6)
  p_name <- paste0("trace_plot_", i, ".png" )
  ggsave(filename =p_name, plot = p, width = 15, height = 7, dpi = 750) # change size res as needed
}

h. PCA

mat <- t(scale(t(cmb@assays@data$data_imputed[ ,!grepl("(QC|HD1017)", cmb$Sample)])))

colnames(mat)
hmat <- mat%>% 
  data.frame()%>%
  mutate_all(log10)%>%
  mutate_all(scale)%>%
  as.matrix()
hmat[is.na(hmat)] <- 0
summary(as.numeric(hmat))
#mat[mat > 2] <- 2
pheatmap::pheatmap(t(hmat))

fit <- prcomp(t(mat))
data.frame(sample = colnames(mat),
           fit$x)%>%
  ggplot(aes(PC1, PC2))+
  geom_point()+
  theme_bw()

write.csv(data.frame(
  sample = mat$Sample,
  hmat
), file = "~/Downloads/clusterg.csv")