Luminex

Luminex measures several analytes in one well with bead sets. Each bead region carries a distinct fluorescence signature and one antibody, so the instrument can tell fifty or a hundred analytes apart in a single sample. Two lasers do the reading: one classifies the bead region, and one measures the reporter signal on the bead. The raw number per analyte per well is a median fluorescence intensity, and concentrations come from a standard curve fitted per run.

The files

The instrument’s own record is a binary .lxb file per well, which the acquisition software reads. The file that reaches the analyst is the export from that software. xPONENT writes a batch CSV, and Bio-Plex Manager, Bio-Rad’s branded build of the same platform, writes its own CSV or XML. The layouts differ, but the content is the same long table: one row per well per analyte, with the bead count and the fluorescence summary.

The example reads a synthetic export in the xPONENT shape: a seven-point standard series in a three-fold dilution, eight unknown samples, three analytes.

library(readr)

lux <- read_csv("/opt/data/xponent_export.csv", show_col_types = FALSE)
dim(lux)
head(lux, 4)

[1] 45  7
# A tibble: 4 × 7
  Well  Sample Analyte   `Bead Count` `Median FI` `Net MFI` `Dilution Factor`
  <chr> <chr>  <chr>            <dbl>       <dbl>     <dbl>             <dbl>
1 A01   Std1   IL-6                69       18321     18276                 1
2 A01   Std1   IL-8                48       22708     22663                 1
3 A01   Std1   TNF-alpha          108       15179     15134                 1
4 A02   Std2   IL-6                52        6037      5992                 3

Every row is one well read for one analyte. Bead Count is how many beads of that region the instrument counted, Median FI is the median reporter fluorescence across those beads, and Net MFI subtracts the background.

Bead count first

The bead count is the quality check that comes with the file. Too few beads means the median rests on too little evidence, and the common threshold is 35 beads. Check it before touching the fluorescence values.

library(dplyr)

# Below 35 beads, the median is unreliable.
lux %>% filter(`Bead Count` < 35)

lux_clean <- lux %>% filter(`Bead Count` >= 35)
cat(nrow(lux) - nrow(lux_clean), "row(s) dropped on bead count\n")

# A tibble: 1 × 7
  Well  Sample    Analyte `Bead Count` `Median FI` `Net MFI` `Dilution Factor`
  <chr> <chr>     <chr>          <dbl>       <dbl>     <dbl>             <dbl>
1 C05   Sample_05 IL-8              12        2374      2329                NA
1 row(s) dropped on bead count

One well in the fixture falls below the threshold: sample 5 read for IL-8. The row is dropped, and any concentration derived from it would have been wrong silently.

The standard curve

The standards travel in the same file as the samples. Each dilution of the standard gives a fluorescence value per analyte, and the run’s software fits a five-parameter logistic to convert sample fluorescence into concentration. On a log-log plot the standards should fall on a clean monotone curve.

library(ggplot2)

standards <- lux_clean %>% filter(grepl("^Std", Sample))

standards %>%
  group_by(Analyte, `Dilution Factor`) %>%
  summarise(median_fi = median(`Median FI`), .groups = "drop") %>%
  arrange(Analyte, `Dilution Factor`) %>%
  print(n = 21)

curve_plot <- ggplot(standards,
                     aes(x = `Dilution Factor`, y = `Median FI`,
                         colour = Analyte)) +
  geom_point(size = 2) +
  scale_x_log10() +
  scale_y_log10() +
  labs(x = "Dilution factor (log scale)",
       y = "Median fluorescence intensity")

dir.create("outputs", showWarnings = FALSE, recursive = TRUE)
ggsave("outputs/luminex_standard_curve.png", plot = curve_plot,
       width = 7, height = 4, dpi = 110, bg = "white")

# A tibble: 21 × 3
   Analyte   `Dilution Factor` median_fi
   <chr>                 <dbl>     <dbl>
 1 IL-6                      1     18321
 2 IL-6                      3      6037
 3 IL-6                      9      1991
 4 IL-6                     27       661
 5 IL-6                     81       211
 6 IL-6                    243        72
 7 IL-6                    729        24
 8 IL-8                      1     22708
 9 IL-8                      3      7284
10 IL-8                      9      2440
11 IL-8                     27       850
12 IL-8                     81       274
13 IL-8                    243        94
14 IL-8                    729        32
15 TNF-alpha                 1     15179
16 TNF-alpha                 3      5029
17 TNF-alpha                 9      1657
18 TNF-alpha                27       601
19 TNF-alpha                81       176
20 TNF-alpha               243        65
21 TNF-alpha               729        21

The fixture’s curve is straight by construction; a real run bends at both ends, which is why the fit is a five-parameter logistic rather than a line. The fitting lives in the acquisition software, and in R the drLumi and nCal packages implement it for xPONENT and Bio-Plex exports. This page stops at the file: the fluorescence values, the bead counts, and the standards.

Summary

• Luminex reports a median fluorescence intensity per analyte per well, from a bead count of a few dozen beads.
• The analyst’s file is the software export, a long table of well by analyte. The per-well binary .lxb stays with the instrument.
• Check the bead count before anything else. Below 35 beads, the median is not to be trusted.
• Concentrations come from a per-run standard curve, a five-parameter logistic fitted to the dilution series in the same file.