pathfindR Analysis for non-Homo-sapiens organisms

Ege Ulgen

2021-01-04

As mentioned in the vignette Introduction to pathfindR, enrichment analysis with pathfindR is not limited to the built-in data. The users are able to utilize custom protein-protein interaction networks (PINs) as well as custom gene sets. These abilities to use custom data naturally allow for performing pathfindR analysis on non-Homo-sapiens input data. In this vignette, we’ll try to provide an overview of how pathfindR analysis using Mus musculus data can be performed.

Preparation of Necessary Data

As of v1.5, pathfindR offers utility functions for obtaining organism-specific PIN data and organism-specific gene sets data via get_pin_file() and get_gene_sets_list(), respectively. See the vignette Obtaining PIN and Gene Sets Data for detailed information on how to gather PIN and gene sets data (for any organism of your choice) for use with pathfindR.

For performing non-human active-subnetwork-oriented enrichment analysis, the user needs the following resources:

After obtaining and processing these data for use, the user can run pathfindR using custom parameters.

Important Note: Because the non-human organism-specific PIN will likely contain less interactions than the Homo sapiens PIN, pathfindR may result in less (or even no) enriched terms.

Obtain Organism-specific Gene Sets

We can obtain the up-to-date M.musculus (KEGG identifier: mmu) KEGG Pathway Gene Sets using the function get_gene_sets_list():

If using another organism, all you have to do is to replace “mmu” with the KEGG organism code in the related arguments in this vignette.

gsets_list <- get_gene_sets_list(source = "KEGG",
                                 org_code = "mmu")

This returns a list containing 2 objects named: gene_sets containing sets of genes of each pathway and desriptions containing the description of each pathway.

The M.musculus KEGG gene set data mmu_kegg_genes and mmu_kegg_descriptions are already provided in pathfindR. For other organisms, the user may wish to save the data as RDS files for future use:

mmu_kegg_genes <- gsets_list$gene_sets
mmu_kegg_descriptions <- gsets_list$descriptions

## Save both as RDS files for later use
saveRDS(mmu_kegg_genes, "mmu_kegg_genes.RDS")
saveRDS(mmu_kegg_descriptions, "mmu_kegg_descriptions.RDS")

These can be later loaded via:

mmu_kegg_genes <- readRDS("mmu_kegg_genes.RDS")
mmu_kegg_descriptions <- readRDS("mmu_kegg_descriptions.RDS")

The function get_gene_sets_list() can also be used to obtain gene sets data from other sources. See the vignette Obtaining PIN and Gene Sets Data for more detail.

Obtain Organism-specific Protein-protein Interaction Network

You may use the function get_pin_file() to obtain organism-specific BioGRID PIN data (see the vignette Obtaining PIN and Gene Sets Data)

Note that BioGRID PINs are smaller for non-H.sapiens organisms and this, in turn, results in less or no significantly enriched terms with pathfindR analysis.

Here, we demonstrate obtaining the organism-specific protein-protein interaction network (PIN) from STRING. You may choose the organism of your choice and find the PIN on the downloads page with the description “protein network data (scored links between proteins)”. When processing, we recommend filtering the interactions using a link score threshold (e.g. 800).

Regardless of the resource, the raw PIN data should be processed to a SIF file, each interactor should be specified with their gene symbols. The first 3 interactions from an example SIF file is provided below:

C2cd2 pp Ints2
Apob pp Gpt
B4galnt1 pp Mettl1

Notice there are no headers and each line contains an interaction in the form GeneA pp GeneB, separated by tab (i.e. \t) with no row names and no column names.

Below we download process the STRING PIN for use with pathfindR:

## Downloading the STRING PIN file to tempdir
url <- "https://stringdb-static.org/download/protein.links.v11.0/10090.protein.links.v11.0.txt.gz"
path2file <- file.path(tempdir(check = TRUE), "STRING.txt.gz")
download.file(url, path2file)

## read STRING pin file
mmu_string_df <- read.table(path2file, header = TRUE)

## filter using combined_score cut-off value of 800
mmu_string_df <- mmu_string_df[mmu_string_df$combined_score >= 800, ]

## fix ids
mmu_string_pin <- data.frame(Interactor_A = sub("^10090\\.", "", mmu_string_df$protein1),
                             Interactor_B = sub("^10090\\.", "", mmu_string_df$protein2))
head(mmu_string_pin, 2)
Interactor_A Interactor_B
ENSMUSP00000000001 ENSMUSP00000017460
ENSMUSP00000000001 ENSMUSP00000039107

Since the interactors are Ensembl peptide IDs, we’ll need to convert them to MGI symbols for use with pathfindR. This can be achieved via biomaRt or any other conversion method you prefer:

library(biomaRt)

mmu_ensembl <- useMart("ensembl", dataset = "mmusculus_gene_ensembl")

converted <- getBM(attributes = c("ensembl_peptide_id", "mgi_symbol"),
                   filters = "ensembl_peptide_id",
                   values = unique(unlist(mmu_string_pin)),
                   mart = mmu_ensembl)
mmu_string_pin$Interactor_A <- converted$mgi_symbol[match(mmu_string_pin$Interactor_A, converted$ensembl_peptide_id)]
mmu_string_pin$Interactor_B <- converted$mgi_symbol[match(mmu_string_pin$Interactor_B, converted$ensembl_peptide_id)]
mmu_string_pin <- mmu_string_pin[!is.na(mmu_string_pin$Interactor_A) & !is.na(mmu_string_pin$Interactor_B), ]
mmu_string_pin <- mmu_string_pin[mmu_string_pin$Interactor_A != "" & mmu_string_pin$Interactor_B != "", ]

head(mmu_string_pin, 2)
Interactor_A Interactor_B
Gnai3 Ppy
Gnai3 Ccr3

Next, we remove self interactions and any duplicated interactions, format the data frame as SIF:

# remove self interactions
self_intr_cond <- mmu_string_pin$Interactor_A == mmu_string_pin$Interactor_B
mmu_string_pin <- mmu_string_pin[!self_intr_cond, ]

# remove duplicated inteactions (including symmetric ones)
mmu_string_pin <- unique(t(apply(mmu_string_pin, 1, sort))) # this will return a matrix object

mmu_string_pin <- data.frame(A = mmu_string_pin[, 1],
                             pp = "pp",
                             B = mmu_string_pin[, 2])

Finally, we save the gene symbol PIN as a SIF file named “mmusculusPIN.sif” under the temporary directory (i.e. tempdir()):

path2SIF <- file.path(tempdir(), "mmusculusPIN.sif")
write.table(mmu_string_pin,
            file = path2SIF,
            col.names = FALSE,
            row.names = FALSE,
            sep = "\t",
            quote = FALSE)
path2SIF <- normalizePath(path2SIF)

We’ll use this path to the custom sif for analysis with run_pathfindR().

The STRING Mus musculus PIN created above is available in pathfindR and can be used via setting pin_name_path = "mmu_STRING" in run_pathfindR().

Running pathfindR on non-Homo sapiens data

Input Data

The data used in this vignette (myeloma_input) is the data frame of differentially-expressed genes along for the GEO dataset GSE99393. The RNA microarray experiment was perform to detail the global program of gene expression underlying polarization of myeloma-associated macrophages by CSF1R antibody treatment. The samples are 6 murine bone marrow derived macrophages co-cultured with myeloma cells (myeloma-associated macrophages), 3 of which were treated with CSF1R antibody (treatment group) and the rest were treated with control IgG antibody (control group). In myeloma_input, 45 differentially-expressed genes with |logFC| >= 2 and FDR <= 0.05 are presented.

knitr::kable(head(myeloma_input))
Gene_Symbol FDR
Aoah 8.23e-05
AW112010 8.23e-05
F13a1 8.23e-05
Pde3b 8.23e-05
P2ry14 8.23e-05
Fcgrt 8.23e-05

Executing run_pathfindR()

After obtaining the necessary PIN and gene sets data, you can then perform pathfindR analysis by setting these arguments: - convert2alias = FALSE: alias conversion only works on H.sapiens genes - pin_name_path = path2SIF: as we’re using a non-built-in PIN, we need to provide the path to the mmu sif file - gene_sets = "Custom: as we’re using a non-built-in source for gene sets - custom_genes = mmu_kegg_genes - custom_descriptions = mmu_kegg_descriptions

myeloma_output <- run_pathfindR(input = myeloma_input,
                                convert2alias = FALSE,
                                gene_sets = "Custom",
                                custom_genes = mmu_kegg_genes,
                                custom_descriptions = mmu_kegg_descriptions,
                                pin_name_path = path2SIF)

knitr::kable(myeloma_output)
ID Term_Description Fold_Enrichment occurrence support lowest_p highest_p Up_regulated Down_regulated
mmu04060 Cytokine-cytokine receptor interaction 12.061372 1 0.1666667 0.0000066 0.0000066 Ccl8, Ccl12, Cxcl10, Il15, Il1b, Tnfsf10, Lifr
mmu04668 TNF signaling pathway 16.895070 1 0.1666667 0.0000217 0.0000217 Ccl12, Cxcl10, Il1b, Il15
mmu04625 C-type lectin receptor signaling pathway 13.257936 1 0.0833333 0.0001550 0.0001550 Il1b, Clec4n, Clec4e
mmu00230 Purine metabolism 7.231602 1 0.0833333 0.0002845 0.0002845 Gda, Pde3b
mmu04217 Necroptosis 6.867420 1 0.0833333 0.0003326 0.0003326 Il1b, Tnfsf10
mmu04623 Cytosolic DNA-sensing pathway 16.179177 1 0.0833333 0.0004891 0.0004891 Cxcl10, Il1b
mmu05164 Influenza A 11.500861 1 0.2500000 0.0005685 0.0005685 Il1b, Ccl12, Cxcl10, Tnfsf10
mmu05132 Salmonella infection 4.481556 1 0.0833333 0.0012058 0.0012058 Il1b, Tnfsf10
mmu05323 Rheumatoid arthritis 16.649502 1 0.1666667 0.0015326 0.0015326 Il15, Il1b, Ccl12
mmu04657 IL-17 signaling pathway 16.088282 1 0.2500000 0.0016999 0.0016999 Cxcl10, Ccl12, Il1b
mmu04620 Toll-like receptor signaling pathway 9.740525 1 0.1666667 0.0022732 0.0022732 Il1b, Cxcl10
mmu00760 Nicotinate and nicotinamide metabolism 12.238095 1 0.0833333 0.0075047 0.0075047 Art2b
mmu04630 JAK-STAT signaling pathway 5.785281 1 0.0833333 0.0108629 0.0108629 Il15, Lifr
mmu05134 Legionellosis 8.089588 1 0.0833333 0.0173113 0.0173113 Il1b
mmu04621 NOD-like receptor signaling pathway 11.431993 1 0.0833333 0.0195262 0.0195262 Il1b, Ccl12, Gbp2, Gbp7
mmu05143 African trypanosomiasis 13.636735 1 0.0833333 0.0299877 0.0299877 Il1b
mmu04012 ErbB signaling pathway 5.820557 1 0.1666667 0.0335621 0.0335621 Nrg1
mmu04672 Intestinal immune network for IgA production 11.363946 1 0.0833333 0.0433333 0.0433333 Il15

Because we used a very strict cut-off (logFC >= 2 + FDR <= 0.05), there were only 18 enriched KEGG pathways. However, the pathways identified here are significantly related to the pathways identified in the original publication by Wang et al.1.

Built-in Mus musculus Data

As aforementioned, for Mus musculus (only), we have provided the necessary PIN (mmu_STRING) and gene set data (mmu_KEGG) so you can also run:

myeloma_output <- run_pathfindR(input = myeloma_input,
                                convert2alias = FALSE,
                                gene_sets = "mmu_KEGG",
                                pin_name_path = "mmu_STRING")

  1. Wang Q, Lu Y, Li R, et al. Therapeutic effects of CSF1R-blocking antibodies in multiple myeloma. Leukemia. 2018;32(1):176-183.↩︎