In espors/idepGolem: Integrated Differential Expression and Pathway Analysis

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

# library(idepGolem)
devtools::load_all()

# Make data for enrichment functions
idep_data <- get_idep_data()

DATABASE <- Sys.getenv("GE_DATABASE")[1]
YOUR_DATA_PATH <- paste0(DATABASE, "data_go/BcellGSE71176_p53.csv")
YOUR_EXPERIMENT_PATH <- paste0(DATABASE, "data_go/BcellGSE71176_p53_sampleInfo.csv")

expression_file <- data.frame(
  datapath = YOUR_DATA_PATH
)
experiment_file <- data.frame(
  datapath = YOUR_EXPERIMENT_PATH
)

load_data <- input_data(
  expression_file = expression_file,
  experiment_file = experiment_file,
  go_button = FALSE,
  demo_data_file = idep_data$demo_data_file,
  demo_metadata_file = idep_data$demo_metadata_file
)

converted <- convert_id(
  query = rownames(load_data$data),
  idep_data = idep_data,
  select_org = "BestMatch"
)

all_gene_info <- gene_info(
  converted = converted,
  select_org = "BestMatch",
  idep_data = idep_data
)

converted_data <- convert_data(
  converted = converted,
  no_id_conversion = FALSE,
  data = load_data$data
)

gene_names <- get_all_gene_names(
  mapped_ids = converted_data$mapped_ids,
  all_gene_info = all_gene_info
)

processed_data <- pre_process(
  data = converted_data$data,
  missing_value = "geneMedian",
  data_file_format = 1,
  low_filter_fpkm = NULL,
  n_min_samples_fpkm = NULL,
  log_transform_fpkm = NULL,
  log_start_fpkm = NULL,
  min_counts = .5,
  n_min_samples_count = 1,
  counts_transform = 1,
  counts_log_start = 4,
  no_fdr = NULL
)

One important aspect of iDEP is the ability to take a subset of genes and perform pathway analysis to determine the pathways that are significantly enriched with different testing effects. The iDEP data base has biological pathways for each species in the database, and with each pathway there is a list of gene IDs that correspond with that process. We then calculate a p-value for the proportion of genes from a pathway in the subset to the proportion of genes subsetted to the entire background. The link below explains the most basic and popular method of determining overlap and enrichment. There are two functions to perform pathway analysis. The first one will be described in the section below.

Enrichment Link https://tv.qiagenbioinformatics.com/video/19605716/understanding-the-p-value-of

Get Gene Sets

Performing pathway analysis for the subsetted list of genes requires only the pathways that have at least one gene in their gene list. We use a database query to ensure that all the pathways that have genes from the subset will be returned in the gene sets. Creating a query requires a subset of the gene_names data frame. For this example, we will simply use the first 100 rows of the data. iDEP has many different methods to create interesting subsets, including differential expression analysis. Depending on the structure of the subsetted genes, use the code chunk below to filter gene_names.

# gene_names_query <- gene_names[gene_names$ensembl_ID == rownames(example_data), ]
gene_names_query <- gene_names[gene_names$ensembl_ID == rownames(processed_data$data), ]

The converted parameter will be the return from the convert_id function call. The input for go determines which portion of the pathway database to use for the analysis. To determine the choices for this input, we can run the function gmt_category. This will return the portions of the database that have pathways corresponding to the matched species. The first two parameters are returned objects in the "Load_Data" instruction. The next is the organism that the expression data is for, an input that has been used in previous functions. gmt_file is a datapath to a gmt file, but only if the species is new and not in the iDEP database. Lastly, the data from the get_idep_data call is filled in. An example call for the demonstration data is in the code chunk below.

gmt_choices <- gmt_category(
  converted = converted,
  converted_data = converted_data$data,
  select_org = "BestMatch",
  gmt_file = NULL,
  idep_data = idep_data
)

Search this list for the desired portion, and specify it with either gmt_choices[[2]] or the string denoting the section. For gmt_choices[[2]] it would be "GOBP". The next parameter, select_org, should be the same as the chunk above. The input gmt_file is only used if you are working with a species that is not in the iDEP database. If this is the case, and there is a GMT file that goes with the species, the input will be the datapath stored in a data frame. For details on this, see the first instruction Load_Data. The next inputs are objects that are also covered in the first instruction. With all the correct inputs, we are now going to actually subset gene_names and create a query. We will use the process_heatmap_data function which is described in the "Clustering" instruction. This final workflow is demonstrated in the chunk below.

heatmap_data <- process_heatmap_data(
  data = processed_data$data,
  n_genes_max = 150,
  # n_genes_min = 0,
  gene_centering = TRUE,
  gene_normalize = TRUE,
  sample_centering = FALSE,
  sample_normalize = FALSE,
  all_gene_names = gene_names,
  select_gene_id = "symbol"
)

gene_names_query <- gene_names[gene_names$symbol %in% rownames(heatmap_data), ]

gene_sets <- read_pathway_sets(
  all_gene_names_query = gene_names_query,
  converted = converted,
  go = "GOBP",
  select_org = "BestMatch",
  gmt_file = NULL,
  idep_data = idep_data,
  gene_info = all_gene_info
)

Calculating Overlap

To use the find_overlap function, we are going to work with the returned object gene_sets from above. This function performs the operation to statistically determine the pathways that are enriched according to the gene query that was submitted. The function above returns a list with objects that will be inputted into parameters of the find_overlap function. The parameter pathway_table is found at gene_sets$pathway_table. This input is a table that is returned from the read_pathway_sets function. It contains information on the pathways that were found by querying the database. The parameter query_set is the vector of ensembl IDs that were used as the query subset. The next input parameter total_genes, is the total number of genes in the database for the matched species. processed_data is the input parameter for the data matrix that has gone through the pre-processing functions. The gene_info parameter takes in the all_gene_info list that was create in the "Load_Data" instruction. The input parameter go takes in the selection for the portion of the database. This input should be the same as what was inputted for go in the function read_pathway_sets.

The next parameter takes in the idep_data list that is created from the iDEP database. use_filtered_background is TURE/FALSE for which gene set to use as the background when calculating overlap. For TRUE, the background gene set will be the genes in the processed data matrix that passed the filtering. If the value is set to FALSE, the background will be all the gene IDs that were found in the iDEP database for the matched species. It is recommended to use the value TRUE, as this will give the most accurate p-value for which pathways are enriched. The video in the introduction gives details about the meaning of enriched. select_org is the input that specifies the species that the expression data is for. Finally, the last input parameter is reduced and can be set to FALSE, or a decimal value. If a decimal value is inputted, this will filter out all the pathways that have a greater proportion of genes in common with another pathway than the inputted decimal. If FALSE is inputted, no pathways will be removed with redudant gene sets. This results of this function be examined in the next section.

pathway_info <- find_overlap(
  pathway_table = gene_sets$pathway_table,
  query_set = gene_sets$query_set,
  total_genes = gene_sets$total_genes,
  processed_data = processed_data$data,
  gene_info = all_gene_info,
  go = "GOBP",
  idep_data = idep_data
  use_filtered_background = TRUE,
  select_org = "BestMatch",
  reduced = .75
)

Examining Results

The table below is the first four columns of the results data frame that was created in the example enrichment analysis. The fifth column is difficult to put into a table because it contains the gene sets for all the significant pathways that were found. The first column in this table is the adjusted p-value for the significance of the overlap calculated for the pathway. The next column denotes the amount of genes from the query that were associated with the given pathway. The next column gives the total number of genes that were associated with the pathway. Finally, the last column is the functional category or description of what the pathway is biologically responsible for. To see the genes for the functional category from the query, run pathway_info[, 5]. If there is no significantly enriched pathways, the data frame will say "No significant enrichment found!"

DT::datatable(
  pathway_info[, 1:4],
  options = list(dom = "ft"),
  rownames = FALSE,
  selection = "single"
)

Conclusion

This instruction covered the steps needed to perform enrichment analysis on a subset of genes. Using these steps it is possible to determine the pathways that are enriched based off the ratio of genes in the subset corresponding to the pathway and genes in the background corresponding to the pathway. The goal is to determine the pathways that are being affected by the different treatments and effects. For troubleshooting, all functions have documentation and the code is available on Github. (https://github.com/gexijin/idepGolem)

espors/idepGolem documentation built on Oct. 27, 2024, 4:56 a.m.