Calculation of molecular number and brightness from fluorescence microscopy image series. The software was published in a 2016 paper. The seminal paper for the technique is Digman et al. 2008. A review of the technique was published in 2017.

If you’re not familiar with the number and brightness (N&B) technique, then you should familiarise yourself with it by reading the papers mentioned above before continuing with the nandb package. The nandb R package is not intended to introduce people to N&B, it’s for people who know about N&B and want to perform N&B calculations.

If you’re new to R and you’re here because you want to use nandb, be warned that you will need to learn some basic R first. I recommend reading the short book “Hands On Programming with R” by Grolemund. This is available for free at https://rstudio-education.github.io/hopr/. That should be enough but if you want further reading, check out “R for Data Science” which is available for free at https://r4ds.had.co.nz/.

This website gives an introduction to the nandb R package, assuming that the reader has a basic level of N&B and R knowledge.

Installation

You can install the release version of nandb from CRAN with:

You can install the (unstable) development version of nandb from GitHub with:

devtools::install_github("rorynolan/nandb")

I highly recommend using the release version. The dev version is just for the ultra-curious and should be thought of as unreliable.

Using nandb

There are two ways to use nandb.

  1. Interactively in the R session, playing with the image as a numeric array, dealing with one image at a time.
  2. In batch mode, having the software read TIFFs, perform the N&B calculations and then write the detrended TIFFs to disk when detrending is over. This method permits the user to use R as little as possible and is better for those who don’t intend to become bon a fide R users.

These are discussed in two articles. These articles deal with brightness; most people use N&B to calculate oligomeric state and hence brightness is the interesting quantity. This package also facilitates number calculations, which are done in the same way, replacing “brightness” with “number” in function names. For example, the “number” equivalent of brightness_timeseries() is number_timeseries(). These articles will use the “epsilon” definition of brightness, but you’re free to use the “B” definition if you prefer it.

  1. Brightness calculations on single images
  2. Brightness calculations on many images in batch mode

Both of these articles mention brightness timeseries. These are explained in the short article Brightness timeseries. N&B timeseries are a very nice feature of nandb, automating a common and otherwise laborious procedure.