cpam (changepoint additive models)

An R package for omics time series analysis.

Read the methods paper here

Application of cpam to RNA-seq time series of Arabidopsis plants treated with excess-light.

Key features

• Changepoint detection: Identify sharp transitions in expression.
• smooth trends: Model expression as a smooth function of time.
• Shape-constrained trends: Cluster targets into biologically meaningful temporal shape classes.
• Quantification uncertainty: Account for uncertainty in expression estimates.
• Transcript-level analysis
  • Perform gene- or transcript-level inferences.
  • Aggregate $p$-values at the gene level for improved power.
• Case-only or case-control time series: Analyse time series data with or without controls.
• User-friendly: Sensible defaults and an interactive shiny interface.

Our new package cpam provides a comprehensive framework for analysing time series omics data. The method uses modern statistical approaches while remaining user-friendly, through sensible defaults and an interactive interface. Researchers can directly address key questions in time series analysis—when changes occur, what patterns they follow, and how responses are related. While we have focused on transcriptomics, the framework is applicable to other high-dimensional time series measurements, making it a valuable addition to the omics analysis toolkit.

This is the initial development version of cpam and we are actively seeking community feedback. If you encounter issues or have suggestions for improvements, please open an issue. We welcome questions and discussion about using cpam for your research through Discussions. Our goal is to work with users to make cpam a robust and valuable tool for time series omics analysis. We can also be contacted via the email addresses listed in our paper here.

Installation

# Installation code here
remotes::install_github("l-a-yates/cpam")

Usage

Step 1: Load the package

library(cpam)

Step 2: Create a tibble for the experimental design.

In this Arabidopsis thaliana time series example, we used the software kallisto to generate counts from RNA-seq data. To load the counts, we provide the file path for each kallisto output file (alternatively you can provide the counts directly as count matrix, or use other quantification software)

exp_design_path
#> # A tibble: 50 × 4
#>    sample  time path                                condition
#>    <chr>  <dbl> <chr>                               <chr>    
#>  1 JHSS01     0 output/kallisto/JHSS01/abundance.h5 treatment
#>  2 JHSS02     0 output/kallisto/JHSS02/abundance.h5 treatment
#>  3 JHSS03     0 output/kallisto/JHSS03/abundance.h5 treatment
#>  4 JHSS04     0 output/kallisto/JHSS04/abundance.h5 treatment
#>  5 JHSS05     0 output/kallisto/JHSS05/abundance.h5 treatment
#>  6 JHSS06     5 output/kallisto/JHSS06/abundance.h5 treatment
#>  7 JHSS07     5 output/kallisto/JHSS07/abundance.h5 treatment
#>  8 JHSS08     5 output/kallisto/JHSS08/abundance.h5 treatment
#>  9 JHSS09     5 output/kallisto/JHSS09/abundance.h5 treatment
#> 10 JHSS10     5 output/kallisto/JHSS10/abundance.h5 treatment
#> # ℹ 40 more rows

Step 3: Obtain a table with the transcript-to-gene mapping

N.B. This is not needed if your counts are aggregated at the gene level, but transcript-level analysis with aggregation of $p$-values to the gene level is recommended. E.g., for Arabidopsis thaliana:

t2g_arabidopsis
#> # A tibble: 50 × 2
#>    target_id   gene_id  
#>    <chr>       <chr>    
#>  1 AT1G01010.1 AT1G01010
#>  2 AT1G01020.2 AT1G01020
#>  3 AT1G01020.6 AT1G01020
#>  4 AT1G01020.1 AT1G01020
#>  5 AT1G01020.4 AT1G01020
#>  6 AT1G01020.5 AT1G01020
#>  7 AT1G01020.3 AT1G01020
#>  8 AT1G03987.1 AT1G03987
#>  9 AT1G01030.2 AT1G01030
#> 10 AT1G01030.1 AT1G01030
#> # ℹ 40 more rows

Step 4: Run cpam

  cpo <- prepare_cpam(exp_design = exp_design_path,
                      count_matrix = NULL,
                      t2g = t2g_arabidopsis,
                      model = "case-only",
                      import_type = "kallisto",
                      num_cores = 5)
  cpo <- compute_p_values(cpo) 
  cpo <- estimate_changepoint(cpo) 
  cpo <- select_shape(cpo) 

Step 5: Visualise the results

Load the shiny app for an interactive visualisation of the results:

  visualise(cpo) # not shown in vignette

Or plot one gene at a time:

  plot_cpam(cpo, gene_id = "AT3G23280")

Isoform 1 (AT3G23280.1) has a changepoint at 67.5 min and has a monotonic increasing concave (micv) shape. Isoform 2 (AT3G23280.2) has no changepoint and has an unconstrained thin-plate (tp) shape.

We can generate a results table which has $p$-values, shapes, log-fold changes and counts with many optimal filters (see tutorials):

  results(cpo)
#> # A tibble: 15,279 × 25
#>    target_id   gene_id     p    cp shape lfc.0 lfc.5 lfc.10 lfc.20 lfc.30 lfc.45
#>    <chr>       <chr>   <dbl> <dbl> <chr> <dbl> <dbl>  <dbl>  <dbl>  <dbl>  <dbl>
#>  1 AT1G01910.1 AT1G01…     0     0 micv      0 1.01   1.70   2.38   2.60   2.73 
#>  2 AT1G01910.2 AT1G01…     0    10 cv        0 0      0      0.553  0.775  0.790
#>  3 AT1G01910.5 AT1G01…     0    10 cx        0 0      0     -3.20  -4.57  -4.82 
#>  4 AT1G02610.1 AT1G02…     0    45 mdcx      0 0      0      0      0      0    
#>  5 AT1G02610.2 AT1G02…     0    10 cx        0 0      0     -0.645 -1.16  -1.71 
#>  6 AT1G02610.3 AT1G02…     0    10 mdcx      0 0      0     -1.48  -2.11  -2.25 
#>  7 AT1G04080.1 AT1G04…     0    10 cv        0 0      0      2.75   3.85   3.97 
#>  8 AT1G04080.2 AT1G04…     0    45 micv      0 0      0      0      0      0    
#>  9 AT1G04080.3 AT1G04…     0     0 micv      0 0.268  0.445  0.603  0.638  0.656
#> 10 AT1G04080.5 AT1G04…     0    10 cx        0 0      0     -2.17  -3.04  -3.10 
#> # ℹ 15,269 more rows
#> # ℹ 14 more variables: lfc.60 <dbl>, lfc.90 <dbl>, lfc.180 <dbl>,
#> #   lfc.240 <dbl>, counts.0 <dbl>, counts.5 <dbl>, counts.10 <dbl>,
#> #   counts.20 <dbl>, counts.30 <dbl>, counts.45 <dbl>, counts.60 <dbl>,
#> #   counts.90 <dbl>, counts.180 <dbl>, counts.240 <dbl>

Tutorials

For a quick-to-run introductory example, we have provided a small simulated data set as part of the package.

• Introductory Example

The following two tutorials use real-world data to demonstrate the capabilities of the cpam package. In addition, they provide code to reproduce the results for the case studies presented in the manuscript accompanying the cpam package.

• Arabidopsis Case Study
• Human Embryo Case Study