Tool:pathfindR - Enrichment Analysis using Active Subnetworks

Question

[egeulgen]

Hello all,

I would like to introduce our group's active-subnetwork-oriented enrichment R package: pathfindR (latest release is v2.0.0).

This tool is designed to improve enrichment analysis by firstly identifying active subnetworks in differential expression/methylation data using a protein-protein interaction network. It then performs enrichment analysis (Over-Representation Analysis). By utilizing the interaction information, pathfindR identifies most of the key involved processes.

• pathfindR provides the human gene sets KEGG, Reactome, BioCarta, GO-MF, GO-BP, GO-CC and GO-All (BP+MF+CC) and the M.musculus KEGG mmu_KEGG gene set for enrichment analysis. The user can also use any other custom gene sets.
• pathfindR also enables creating pathway diagrams for hsa-KEGG with the involved genes colored by change values or visualizations of interactions of genes involved in the pathways and interaction graphs for other gene sets, displaying the interactions of involved genes
• The package allows for clustering of enriched terms and establishment of representative terms. This allows for further abstraction and reduces the complexity of analysis. The available clustering approaches are hierarchical and fuzzy.
• pathfindR provides functionality to score terms for individual samples and plot a heat map of the scores. This allows the user to investigate how a gene set is altered in a given sample (or in cases vs. controls).
• pathfindR provides functionality to combine and compare 2 pathfindR enrichment results
• the package also offers numerous visualization options, including term-gene interaction graph, term-gene interaction heatmap, UpSet plots and graphs of identified subnetworks.

---

You can install the released version of pathfindR from CRAN via:

install.packages("pathfindR")

And the development version from GitHub via:

install.packages("devtools") # if you have not installed this
devtools::install_github("egeulgen/pathfindR")

---

You can read more about the method, its example applications on 3 different datasets and comparison with other available methods in our 2019 article.

Find out more in the pathfindR website and check out the pathfindR tutorial.

[Zhilong Jia]

It's useful.
1. Why not to bioconductor, containing lots of bioinformatics R packages instead of CRAN?
2. no examples of pathmap in the vignette.
3. An enhancement: Visualise the results of run_pathfindR, besides the table show.

Thank you.

[egeulgen]

Thank you for the feedback! To answer them:

1. We did not think it would make any difference
2. We will try to include an example for pathmap in the next version
3. The pathways are already visualized and can be viewed through the HTML reports. Would you expand on how we could provide more visualizations?

Thanks again,

-E

[Zhilong Jia]

1. Precision advertising. The targeting users of pathfindR are bioinformaticians, which are the users of Bioconductor, while the users of CRAN is R users. Getting more chance to encounter target audiences of pathfindR via Bioconductor. Anyway, not a problem.
2. I did not check the HTML report as I cannot run the examples after installing pathfindR and filed an issue at Github. I mean the pathway enrichment itself, not visualization for each pathway based on pathview.

[egeulgen]

We’re looking into your issue on Github. We will try to include a visualization of the enrichment results in the next version (considering a bubble chart but feel free to share any ideas of what you had in mind).

[jared.andrews07]

I will second the idea of a bubble chart. This is one of the easier/better GO packages to use that I've tried, but it lacks the visualization options of some others. Keep up the good work.

[kalyanimeha]

Can you please give a tutorial by taking some examples of a gene.

[egeulgen]

You can see an example application in the vignette (browseVignettes("pathfindr_vignette")). We also have a wiki.

[amgodogma]

HI This is a great package. I'm having a little trouble with the pathway scoring part of the tutorial. I'm not sure how to create the RA_exp_mat for my data?

thanks again

[egeulgen]

The expression matrix for the pathway scores is the same matrix as the one you use for differential expression analysis. Columns are samples, rows are genes and the values are expression values.

[andrea.curtabbi]

Hi the package seems to be exactly what I need but I have an issue in install it, I try with the pak command: pak::pkg_install("pathfindR") It gives me always this error: Error: callr subprocess failed: Failed to download class from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/class_7.3-17.tgz. Failed to download cluster from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/cluster_2.1.0.tgz. Failed to download codetools from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/codetools_0.2-16.tgz. Failed to download DEoptimR from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/DEoptimR_1.0-8.tgz. Failed to download diptest from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/diptest_0.75-7.tgz. Failed to download doParallel from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/doParallel_1.0.15.tgz. Failed to download flexmix from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/flexmix_2.3-15.tgz. Failed to download foreach from https://cran.rediris.es//bin/macosx/el-capitan/contrib/3.6/foreach_1.5.0.tgz. Failed to download fpc from `https://cran.rediris.es//bin/

I am using R 3.6.3, in a mac Os x Catallina. May it be a Catalina-related problem?

Thanks

[egeulgen]

hey, I think it's an issue with pak. install.packages("pathfindR") should work. You may try with devtools for the development version:

install.packages("devtools") # if you have not installed "devtools"
devtools::install_github("egeulgen/pathfindR")

[manojkamalanathan711]

Hi there! I have tried both pak::pkg_install("egeulgen/pathfindR") and devtools::install_github("egeulgen/pathfindR"), both didn't work:

With pak::pkg_install("egeulgen/pathfindR"), I get the following error:

Error: callr subprocess failed: Failed to download DBI from https://cran.rstudio.com//bin/windows/contrib/3.6/DBI_1.1.0.zip. Failed to download DEoptimR from https://cran.rstudio.com//bin/windows/contrib/3.6/DEoptimR_1.0-8.zip. Failed to download MASS from https://cran.rstudio.com//bin/windows/contrib/3.6/MASS_7.3-51.6.zip. Failed to download Matrix from https://cran.rstudio.com//bin/windows/contrib/3.6/Matrix_1.2-18.zip. Failed to download R.methodsS3 from https://cran.rstudio.com//bin/windows/contrib/3.6/R.methodsS3_1.8.0.zip. Failed to download R.oo from https://cran.rstudio.com//bin/windows/contrib/3.6/R.oo_1.23.0.zip. Failed to download R.utils from https://cran.rstudio.com//bin/windows/contrib/3.6/R.utils_2.9.2.zip. Failed to download RCurl from https://cran.rstudio.com//bin/windows/contrib/3.6/RCurl_1.98-1.2.zip. Failed to download RSQLite from https://cran.rstudio.com//bin/windows/contrib/3.6/RSQLite_2.2.0.zip. Failed to download XML from `https://cran.rstudi

And with devtools::install_github("egeulgen/pathfindR"), I get the following error:

Skipping 1 packages not available: pathfindR.data Installing 35 packages: pathfindR.data, org.Hs.eg.db, backports, generics, tidyselect, tweenr, polyclip, RcppEigen, dplyr, tidyr, RcppArmadillo, modeltools, mclust, DEoptimR, R.methodsS3, jsonlite, openssl, XVector, zlibbioc, bitops, tinytex, ggforce, ggrepel, tidygraph, graphlayouts, flexmix, prabclus, diptest, robustbase, kernlab, R.oo, Biostrings, XML, graph, RCurl Installing packages into ‘C:/Users/manojka/Documents/R/win-library/3.6’ (as ‘lib’ is unspecified) Error: Failed to install 'pathfindR' from GitHub: (converted from warning) package ‘pathfindR.data’ is not available (for R version 3.6.2)

Is there anyway around this? Thanks!

[egeulgen]

hey,

package ‘pathfindR.data’ is not available (for R version 3.6.2)

You need to have R version 4 installed.

[manojkamalanathan711]

Thanks you, that works!

[jungjoon.kim]

What a great package! Thanks!

Answer 1

[Francesco]

Hi thanks a lot for your work! I am currently using your package to extract further information from an RNAseq experiment. I have two questions on the output from running "run_pathfindR", notably on the plots that are returned in the folders. I wasn't able to find answers in supporting material.

1) Running on KEGG and on Reactome I obtain two different kind of plots, what is the difference? 2) The plots are given only for significantly enriched pathways?

Here's my code: output_df <- run_pathfindR(input = pathdata, gene_sets = "Reactome", p_val_threshold = 0.05, output_dir = "pathfindR_Results_Reactome", max_to_plot = 400, adj_method = "fdr")

output_df_new <- run_pathfindR(input = pathdata, gene_sets = "KEGG", p_val_threshold = 0.05, output_dir = "pathfindR_Results_KEGG", max_to_plot = 400, adj_method = "fdr")

Thank you again, Francesco

[egeulgen]

Hey Francesco, thank you for using pathfindR.

1) For KEGG, the pathway diagrams from (hsa) KEGG are downloaded and the involved genes are colored. For Reactome, interactions of affected genes within the Protein-Protein Interaction Network (by default BioGRID) are determined (via a shortest-path algorithm) and are visualized (colored by change status) per affected pathway.

2) Yes, the plots are only for significantly enriched pathways.

Best,

-E

[Francesco]

Dear Ege, I'm sorry to bother you furhter but I still can't find myself in these results. Looking at the kegg plots for some specific pathway I find several genes depicted as up or downregulated even if they are not in the input genes I am providing to the PathfindR tool. Can you explain me why? With my very best regards, Francesco.

Answer 2

[Deepblue]

Hi Ege. I'd like to change the color of enrichment plot from orange to another colour. In layers of the plot, I did not find any color name, except for ticks, legend, etc. with black, grey20, grey30 colors. Is there any way to modify the color of the circles? Where is it stored? Thanks