reluxr

Implements the deconvolution algorithm developed in Mauri, Vecchione, and Fritz (2019) which enables deconvolution of luminscence readings for experimental culture plates. {reluxr} provides functions for calculating the ‘best’ deconvolution matrix from a calibration plate, and enables usage of this calibration matrix (or one calculated previously) to adjust luminescent experimental values from a plate reader. This is an R-based implementation of the MatLab workflow from the original paper paper titled 1Deconvolution of Luminescence Cross-Talk in High-Throughput Gene Expression Profiling’ (Mauri, Vecchione, and Fritz 2019)

Installation

The package is not currently available on CRAN. Install the released version from r-universe with the following code:

install.package("reluxr", repos = "https://bradyajohnston.r-universe.dev")

Example

This is a basic example which shows you how to solve a common problem:

library(reluxr)

fl <- system.file(
   "extdata",
   "calibrate_tecan",
   "calTecan1.xlsx",
   package = "reluxr"
 )

Create a Deconvolution Matrix

The deconvolution matrix is created from a calibration plate, which contains a single well with luminescent bacteria, with all other wells being empty. The cross-talk when the plate-reader measures the wells can then be calculated and used to create a deconvolution matrix which will remove the crosstalk from the measured values.

dat <- plate_read_tecan(fl)

dat
#> # A tibble: 23,040 × 5
#>    cycle_nr time_s signal well   value
#>       <dbl>  <dbl> <chr>  <chr>  <dbl>
#>  1        1      0 OD600  A01   0.0450
#>  2        1      0 OD600  A02   0.0452
#>  3        1      0 OD600  A03   0.0453
#>  4        1      0 OD600  A04   0.0453
#>  5        1      0 OD600  A05   0.0453
#>  6        1      0 OD600  A06   0.0452
#>  7        1      0 OD600  A07   0.0458
#>  8        1      0 OD600  A08   0.0456
#>  9        1      0 OD600  A09   0.0455
#> 10        1      0 OD600  A10   0.0451
#> # … with 23,030 more rows

Regardless of how you read in the required data, it needs to be (and should be regardless) in a tidy format, with each row being an observation and each column a variable. In the case above we have columns for the cycle_nr, time_s, signal, well and value. While it would be better to have a column for the OD600 and for the LUMI data, the time points do not match and aren’t currently pivotable.

To have a look at the final data, we can plot the plate based on log-transformed luminescence values. We can see the very bright single well than contains the bacteria, and the bleed-through signal that is around it.

dat_fil <- dat |> 
  dplyr::filter(signal == "LUMI", time_s > 500)


dat_fil |> 
  dplyr::group_by(well) |> 
  dplyr::summarise(value = mean(value)) |> 
  rl_plot_plate(value)

We can use the rl_calc_decon_matrix() function which will calculate a deconvolution matrix, reducing the background bleed-through from the plate to below a noise threshold. The noise threshold should be the instrument’s background noise, which is defined as three times the standard deviation of a blank well. The lower the noise threshold, the harder it will be to calculate a deconvolution matrix which works with the data.

We can also quickly look at the resulting deconvolution matrix (mat_d_best) itself.

mat_d_best <- rl_calc_decon_matrix(
  data = dat_fil,
  value = value,
  time = time_s,
  ref_well = "E05",
  b_noise = 20
)

image(log10(mat_d_best))

Deconvoluting the Data

Now that we have the matrix, we can use it to adjust the data.

We do so using the rl_adjust_plate() funciton, which takes a dataframe, the name of the column you which to adjust, the deconvolution matrix (in this case mat_d_best, and the name ofthe column which stores the time data).

The returned dataframe will have the value column adjusted and deconvoluted using the deconvolution matrix supplied.

In the examples below we first plot all of the values without deconvolution, then apply the deconvolution matrix and plot the values again.

rl_plot_time <- function(data, time, value, group = "well") {
  
  data <- dplyr::mutate(
    data, 
    time = {{ time }}, 
    value = {{ value }}, 
    group = {{ group }}
  )
  
  plt <- ggplot2::ggplot(
    data, 
    mapping = ggplot2::aes(
      x = time,
      y = value, 
      group = group
    )
  ) + 
    ggplot2::geom_line() + 
    ggplot2::scale_y_log10() + 
    ggplot2::theme_bw()
  
  plt
}

Raw Values

dat |>
  dplyr::filter(signal == "LUMI") |> 
  rl_plot_time(time_s, value, well) + 
  ggplot2::labs(
    x = "Time (s)", 
    y = "LUM"
  )
#> Warning in self$trans$transform(x): NaNs produced
#> Warning: Transformation introduced infinite values in continuous y-axis
#> Warning: Removed 25 row(s) containing missing values (geom_path).

Deconvoluted Values

dat |>
  dplyr::filter(signal == "LUMI") |> 
  
  rl_adjust_plate(value, mat_d_best, time = time_s) |> # deconvolute the values
  
  rl_plot_time(time_s, value, well) + 
  ggplot2::labs(
    x = "Time (s)", 
    y = "LUM"
  )
#> Warning in self$trans$transform(x): NaNs produced
#> Warning: Transformation introduced infinite values in continuous y-axis
#> Warning: Removed 657 row(s) containing missing values (geom_path).

We can also replot the plate from earlier, with the newly deconvoluted values.

dat_fil |> 
  rl_adjust_plate(value, mat_d_best, time = time_s) |> # deconvolute the values
  dplyr::group_by(well) |> 
  dplyr::summarise(value = mean(value)) |> 
  rl_plot_plate(value) + 
  ggplot2::scale_fill_viridis_c(
    "log10(LUMI)",
    breaks = 1:5, 
    limits = c(1, NA)
  )
#> Scale for 'fill' is already present. Adding another scale for 'fill', which
#> will replace the existing scale.
#> Warning in FUN(X[[i]], ...): NaNs produced

Mauri, Marco, Stefano Vecchione, and Georg Fritz. 2019. “Deconvolution of Luminescence Cross-Talk in High-Throughput Gene Expression Profiling.” ACS Synthetic Biology 8 (6): 1361–70. https://doi.org/10.1021/acssynbio.9b00032.