Downloading data from the website

The first step required to run HPAStainR is downloading HPA’s normal tissue staining data and their cancer data. While available online, HPAStainR has a function that can download and load the data for you.

library(HPAStainR)
#> Loading required package: dplyr
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
#> Loading required package: tidyr

HPA_data <- HPA_data_downloader(tissue_type = "both", save_file = FALSE)

The above function has downloaded both normal tissue and cancer data. save_file was set to FALSE, but if it were set to TRUE as there was no given argument for save location, both files being saved to the current working directory. The data has also been unzipped and loaded into the object HPA_dat as a list of data frames called hpa_dat and cancer_dat which hold the normal tissue and cancer tissue data respectively. If the code is run again it would redownload the files unless you had set save_file to TRUE, in which case it would just load said saved files.

The hpar package as a version controlled alternative

HPA_data_downloader provides the most up to date information on the Human Protein Atlas website. However for more consistent results with version control, feel free to use the data from the hpar BioConductor package using the following commands.

if (!requireNamespace("hpar", quietly = TRUE))
    BiocManager::install(hpar)
data(hpaNormalTissue, package = "hpar")
data(hpaCancer, package = "hpar")

Using the HPAStainR function

Now that the data is available you can now us the HPAStainR function. This requires a list of proteins or genes you are interested in. In this example, we’re going to use pancreatic enzymes PRSS1, PNLIP, CELA3A, and the hormone PRL.

gene_list = c("PRSS1", "PNLIP","CELA3A", "PRL")

stainR_out <- HPAStainR::HPAStainR(gene_list = gene_list,
          hpa_dat = HPA_data$hpa_dat,
          cancer_dat = HPA_data$cancer_dat,
          cancer_analysis = "both",
          stringency = "normal")


head(stainR_out, 10)
#> # A tibble: 10 x 11
#>    cell_type percent_high_ex… percent_medium_… percent_low_exp… percent_not_det…
#>    <chr>                <dbl>            <dbl>            <dbl>            <dbl>
#>  1 PANCREAS…             0.75             0                0                0.25
#>  2 PITUITAR…             0.25             0                0                0   
#>  3 DUODENUM…             0                0.25             0                0.75
#>  4 SMALL IN…             0                0.25             0                0.75
#>  5 COLON - …             0                0                0.25             0.75
#>  6 head and…             0                0.07             0                0.93
#>  7 pancreat…             0                0.02             0                0.98
#>  8 ADIPOSE …             0                0                0                1   
#>  9 ADRENAL …             0                0                0                1   
#> 10 APPENDIX…             0                0                0                1   
#> # … with 6 more variables: number_of_proteins <dbl>, tested_proteins <chr>,
#> #   detected_proteins <chr>, staining_score <dbl>, p_val <dbl>, p_val_adj <dbl>

The output of HPAStainR is a tibble with multiple columns. The basic columns include the following:

• cell_type: The cell types/cancers that are tested in the Human Protein Atlas.

• percent/count_high/medium/low_expression: Either the percent or count of genes from the list that stain either at high levels, medium levels or low levels.

• percent/count_not_detected: The number or percent of proteins that failed to stain the cell type.

• number of proteins: The number of proteins tested in a cell type.

• tested_proteins: A character string of proteins that were tested in the cell type as not all proteins are tested in every cell type.

• detected_proteins: A character string of proteins that were detected in each cell type.

• enriched_score: An arbitrary ranking value further explained below.

• p_val: A p-value denoting an enrichment of rarely staining proteins (stained in <29% of the cell types, see paper [cite] for further details).

• p_val_adjust: The previous p-value adjusted for multiple testing using “holm”

The staining score an arbitrary rank of staining weighted on how highly a protein stained. See the manual for the equation and further information.

Run the Shiny app

shiny_HPAStainR(hpa_dat = HPA_data$hpa_dat,
                cancer_dat = HPA_data$cancer_dat,
                cell_type_data = hpa_summary)