Skip to contents

Filter proximity scores

FilterProximityScores.Rd

At least one of the thresholds must be set. The function will filter out protein pairs that do not meet the specified thresholds.

All filters can be used together or separately, but it's recommended to only use the background_threshold_pct filter. This filter is applied to the fraction of UMI counts which does not depend on the cell size (total number of molecules).

The filters are applied in the following order:

  1. Remove protein pairs with a minimum UMI count fraction below background_threshold_pct

  2. Remove protein pairs with a minimum UMI count below background_threshold_count

  3. Remove protein pairs detected in fewer than min_cells_count cells

Usage

FilterProximityScores(object, ...)

# S3 method for class 'tbl_lazy'
FilterProximityScores(
  object,
  background_threshold_pct = NULL,
  background_threshold_count = NULL,
  min_cells_count = NULL,
  name = "filtered_proximity",
  ...
)

# S3 method for class 'data.frame'
FilterProximityScores(
  object,
  background_threshold_pct = NULL,
  background_threshold_count = NULL,
  min_cells_count = NULL,
  ...
)

Arguments

object

An tbl_df or tbl_lazy object with proximity scores.

...

Additional arguments. Currently not used.

background_threshold_pct

The background abundance level given as a fraction total UMI counts. For example, 0.05 means that the background is set at 5% of the total UMI counts for each cell. A protein pair must have at least this fraction for both proteins to be kept in the table.

background_threshold_count

The background abundance level given as a UMI count. For example, 30 means that the background is set at 30 UMI counts. A protein pair must have at least this UMI count for both proteins to be kept in the table. Note that this puts a hard cutoff on the UMI counts, which may not be desired if the cells have very different total UMI counts.

min_cells_count

The minimum number of cells in which a protein pair must be detected to be kept in the table. This is useful to remove protein pairs that are detected in very few cells, which may be due to noise or low expression levels.

name

Name of the temporary table in the database containing the filtered proximity scores.

Value

A tbl_df or tbl_lazy with filtered proximity scores.

Examples

library(pixelatorR)

pxl_file <- minimal_pna_pxl_file()
se <- ReadPNA_Seurat(pxl_file)
#> ✔ Created a <Seurat> object with 5 cells and 158 targeted surface proteins
proximity_table <- ProximityScores(se, add_marker_proportions = TRUE)
#> ! Setting `add_marker_counts = TRUE` which is required when `add_marker_proportions = TRUE`.

# Filter scores
proximity_table_filtered <- FilterProximityScores(
  proximity_table,
  background_threshold_pct = 0.001
)
#> Warning: Missing values are always removed in SQL aggregation functions.
#> Use `na.rm = TRUE` to silence this warning
#> This warning is displayed once every 8 hours.

# Rows kept
pct_rows_kept <- round(nrow(proximity_table_filtered) / nrow(proximity_table) * 100, digits = 2)
glue::glue("Fraction of rows kept: {pct_rows_kept}%")
#> Fraction of rows kept: 11.13%