pRAIN — Parallel RAIN

pRAIN is a parallel-enabled fork of the RAIN R package (Thaben & Westermark, 2014) for detecting rhythmic signals in circadian and other periodic time-series data.

The original RAIN implements a non-parametric umbrella-alternatives test (Mack & Wolfe, 1981) extended for cyclic data. pRAIN adds a single new parameter (num.cores) that parallelises the two most expensive internal steps without changing any outputs.

---

What changed from the original

New parameter: num.cores

rain(..., num.cores=1L)

The default (1L) reproduces original RAIN behaviour exactly. Any value greater than 1 activates parallel execution across two independent layers:

1. Harding null-distribution precomputation — the dominant bottleneck for dense sampling (e.g. 1-hour intervals) or when the sampling duration is not a multiple of the period. All unique configurations are identified upfront; their exact null distributions are then computed simultaneously rather than sequentially.

2. Per-series evaluation loop — each time series (e.g. gene) is scored independently against the precomputed distributions and has its p-values adjusted. These are evaluated in parallel across all input columns.

Platform support

Platform	Precomputation	Evaluation loop
macOS / Linux	parallel::mclapply (fork-based)	parallel::mclapply
Windows	sequential	parallel::makeCluster + parLapply

The parallel package ships with base R meaning no additional dependencies are introduced.

Observed speedup

Example: 48-hour time-series, 2-hour sampling interval, 5 replicates, 10000 genes.

num.cores	Time
1	~403 s
2	~224 s
4	~136 s
8	~130 s

Speedup scales primarily with the number of unique null distributions that need to be computed, which grows with the number of replicates and (especially) when the sampling duration is not a multiple of the period.

---

Installation

# install.packages("pak")pak::pak("pascalnoser/prain")

You will also need the Bioconductor dependencies:

if (!requireNamespace("BiocManager", quietly=TRUE))
install.packages("BiocManager")
BiocManager::install(c("multtest", "gmp"))

---

Quick start

library(prain)
# Simulate data: 24 time points, 3 replicates, 200 genes
set.seed(42)
mat<-matrix(rnorm(24*3*200), nrow=24*3)
# Sequential (original behaviour)results_seq<- rain(mat, deltat=1, period=24, nr.series=3,
method="independent")
# Parallel — use all available coresresults_par<- rain(mat, deltat=1, period=24, nr.series=3,
method="independent",
num.cores=parallel::detectCores())
# Results are identical
all.equal(results_seq, results_par) # TRUE

See ?rain for the full parameter reference.

---

Credits

pRAIN is a thin wrapper around the original RAIN package. All statistical methodology and the core algorithm are the work of the original authors:

Paul F. Thaben and Pål O. Westermark
Detecting Rhythms in Time Series with RAIN.
Journal of Biological Rhythms, 29(6), 391–400 (2014).
doi: 10.1177/0748730414553029

If you use pRAIN in published work, please cite the paper above.

The original RAIN package is available on Bioconductor: https://www.bioconductor.org/packages/release/bioc/html/rain.html

---

License

GPL-2 (same as the original RAIN package).