In sigven/oncoEnrichR: Cancer-dedicated gene set interpretation

Gene fitness scores {.tabset}

The Cancer Dependency Map and Project Score have performed systematic genome-scale CRISPR/Cas9 drop-out screens in a large number of highly-annotated cancer models to identify genes required for cell fitness in defined molecular contexts
Here, we are showing, across the main human tissue types:
Genes in the query set that are annotated with a statistically significant effect on cell fitness in any of the screened cancer cell lines (fitness score is here considered a quantitative measure of the reduction of cell viability elicited by a gene inactivation, via CRISPR/Cas9 targeting). The fitness score is computed based on the BAGEL and CRISPRCleanR algorithms.
The underlying cell viability data comes from the integration of data in n = 912 cell lines, using both DepMap (2020_Q2) and Project Score (20210311), as published in Pacini et al., Nat Commun, 2021
Settings
Maximum loss-of-fitness score pr. gene (BAGEL - scaled Bayes Factor): r onc_enrich_report[['config']][['fitness']][['max_BF_score']]

Loss-of-fitness distribution

targets_wide <- as.data.frame(
  onc_enrich_report[['data']][['fitness']][['fitness_scores']][['targets']] |>
    dplyr::group_by(.data$symbol,
                    .data$n_gene_tissue,
                    .data$symbol_link_ps,
                    .data$n_gene,
                    .data$tissue) |>
    dplyr::summarise(cell_lines = paste(.data$model_name, collapse = ", "),
                     ps_link = paste(.data$model_link_ps, collapse = ", "),
                     .groups = "drop") |>
    dplyr::ungroup() |>
    dplyr::mutate(cell_lines = stringr::str_replace_all(.data$cell_lines, "\\.", "-"))
)

fitness_lof_oe_plot <- targets_wide |>
  dplyr::mutate(symbol = forcats::fct_reorder(symbol, n_gene)) |>
  ggplot2::ggplot(ggplot2::aes(x = symbol, y = n_gene_tissue, fill = tissue)) +
  ggplot2::geom_bar(stat = "identity") +
  ggplot2::coord_flip() +
  ggplot2::ylab("Number of cell lines with significant loss-of-fitness") +
  ggplot2::xlab("") +
  ggplot2::scale_fill_manual(values = pals::stepped(24)) +
  ggplot2::theme(
    panel.grid.minor = ggplot2::element_blank(),
    axis.text = ggplot2::element_text(family = "Helvetica", size = 11),
    legend.text = ggplot2::element_text(family = "Helvetica", size = 11),
    axis.title.x = ggplot2::element_text(family = "Helvetica", size = 12),
    legend.title = ggplot2::element_blank(),
    #set thickness of axis ticks
    axis.ticks = ggplot2::element_line(size = 0.2),
    #remove plot background
    plot.background = ggplot2::element_blank()
  )

plotly::ggplotly(fitness_lof_oe_plot)
rm(fitness_lof_oe_plot)

cat('<br>\n <ul><li>  <i> <span style="font-size: 105%; padding: 3px; background-color:#989898; color:white">&nbsp;&nbsp;<b>NO GENES</b> in the query set found with loss-of-fitness effect from CRISPR/Cas9 screens in DepMap/Project Score.&nbsp;&nbsp; </span></i></li></ul>',sep='\n')
cat('\n')
cat('<br><br>')

Loss-of-fitness table

A maximum of n = 2000 tissue-specific-cell-lines-per-gene are shown in the table below (the complete set of genes, including the loss-of-fitness scores per cell line, can be found in the Excel output of oncoEnrichR)

htmltools::br()

fitness_lof_oe_df <- onc_enrich_report[['data']][['fitness']][['fitness_scores']][['targets']] |>
  dplyr::group_by(.data$symbol,
                  .data$n_gene_tissue,
                  .data$symbol_link_ps,
                  .data$n_gene,
                  .data$tissue) |>
  dplyr::summarise(ps_link = paste(.data$model_link_ps, collapse = ", "),
                   .groups = "drop") |>
  dplyr::ungroup() |>
  dplyr::select(.data$symbol_link_ps, .data$tissue, .data$ps_link) |>
  dplyr::rename(cell_lines = .data$ps_link, symbol = .data$symbol_link_ps) |>
  dplyr::mutate(loss_of_fitness = T) |>
  head(2000)

fitness_lof_oe_table <- 
  crosstalk::SharedData$new(fitness_lof_oe_df)
rm(fitness_lof_oe_df)

crosstalk::bscols(
  list(
    crosstalk::filter_select("tissue", "Tissue", 
                             fitness_lof_oe_table, ~tissue)
  )
)

DT::datatable(
  fitness_lof_oe_table, 
  escape = F, 
  extensions = c("Buttons","Responsive"), 
  width = "100%",
  options = list(buttons = c('csv','excel'),
               dom = 'Bfrtip', 
               pagelength = 20)
  )

cat('<br>\n <ul><li>  <i> <span style="font-size: 105%; padding: 3px; background-color:#989898; color:white">&nbsp;&nbsp;<b>NO GENES</b> in the query set found with loss-of-fitness effect from CRISPR/Cas9 screens in DepMap/Project Score.&nbsp;&nbsp; </span></i></li></ul>',sep='\n')
cat('\n')
cat('<br><br>')

Target priority scores

Promising candidate therapeutic targets are indicated through target priority scores. Target priority scores are based on integration of CRISPR knockout gene fitness effects with genomic biomarker and patient data (Behan et al., Nature, 2019). All genes are assigned a target priority score between 0 – 100 from lowest to highest priority. In the heatmap shown below, genes in the query set are ranked according to their respective priority scores across all cancers (i.e. Pan-cancer), limited to the top 100 candidates.

oe_priority_matrix <- as.data.frame(
  onc_enrich_report[['data']][['fitness']][['target_priority_scores']][['targets']] |>
  dplyr::arrange(symbol) |>
  tidyr::pivot_wider(names_from = tumor_type,
                     values_from = priority_score)
)
rownames(oe_priority_matrix) <-
  oe_priority_matrix$symbol
oe_priority_matrix$symbol <- NULL
oe_priority_matrix <- as.matrix(oe_priority_matrix)

if(NROW(oe_priority_matrix) > 100){
  oe_priority_matrix <- oe_priority_matrix[1:100,]
}

oe_priority_matrix <- 
  oe_priority_matrix[nrow(oe_priority_matrix):1, ]

oe_target_priority_fig <- plotly::plot_ly(
  colors = "YlGn",
  width = 800,
  height = 400 + (14.67 * NROW(oe_priority_matrix))) |>
  plotly::add_heatmap(
    y = rownames(oe_priority_matrix),
    x = colnames(oe_priority_matrix),
    z = oe_priority_matrix,
    hovertext = "Priority score",
    yaxis = "y") |>
  plotly::layout(
    title = 'Target priority scores',
    xaxis = list(tickfont = list(size = 13, family = "Helvetica"), 
                 tickangle = -50),
    yaxis = list(tickfont = list(size = 12, family = "Helvetica")),
    margin = list(l = 75, r = 20, b = 150, t = 30, pad = 4)
  ) |>
  plotly::colorbar(
    nticks = 10, 
    title = list(text = "Priority score",side = "bottom"),
    limits = c(0, plyr::round_any(max(oe_priority_matrix), 10, ceiling)))

rm(oe_priority_matrix)
oe_target_priority_fig

cat('<br>\n <ul><li>  <i> <span style="font-size: 105%; padding: 3px; background-color:#989898; color:white">&nbsp;&nbsp;<b>NO / LIMITED GENES</b> in the query set nominated as prioritized targets in DepMap/Project Score.&nbsp;&nbsp; </span></i></li></ul>',sep='\n')
cat('\n')

{.unlisted .unnumbered .toc-ignore}

Citation Note : If you use the output of the Gene fitness scores module of oncoEnrichR in your research, please cite the following resources and tools as appropriate: