Skip to contents

NICHESv2 1.0.0

First public release.

Main features

  • create_NICHESObject() scores ligand-receptor (LR) signaling between pairs of cells, in "spatial" mode (k-nearest neighbors or fixed radius) or "nonspatial" mode (type-stratified random sampling).
  • The 15-slot NICHESObject stores non-zero scores in sparse long-format data.tables, with print(), summary(), dim(), length() and [ methods.
  • aggregate_NICHESObject() summarizes each cell’s neighborhood by LR pair and, in spatial mode, by neighboring cell type.
  • load_LRM_database() provides ConnectomeDB2025 (default) and FANTOM5, both bundled for human, mouse, rat and pig, and OmniPath for human, mouse and rat.
  • extract_NICHESInputs_Seurat() and extract_NICHESInputs_AnnData() read Seurat V4 and V5 objects and .h5ad files. The AnnData reader doesn’t need Python.
  • export_to_TissuePlex() writes edges to a Parquet file for interactive viewing.
  • download_NICHESv2_data() downloads the two public datasets used in the website tutorials: a 10x Visium mouse brain section and interferon-beta stimulated PBMCs.

Changes for users of the 0.0.9000 development version

  • The license is now the GNU AGPL v3 instead of MIT. A commercial license is also available from the Raredon Lab.
  • [ now keeps each retained cell’s neighborhood entries and aggregation rows unchanged, so they still describe the full tissue. Previously, neighborhood slots kept the full object’s entries, aggregations were dropped, and aggregating a subset could give wrong means.
  • Recomputing aggregations on a subset now stops with an error. Aggregate the full object before subsetting, or run add_neighborhoods(overwrite = TRUE) first to analyze the subset as a separate tissue.
  • [ now keeps the edge.filter.mode and remove.autocrine attributes.
  • add_neighborhoods(overwrite = TRUE) now clears the stale $aggregations$neighborhood.edge.agg table.
  • Loading the package no longer changes data.table’s thread setting, and socket workers now stay at one data.table thread.
  • Functions with a seed argument now restore the caller’s random number state.
  • Mode "sampled" is now "nonspatial".
  • Aggregate_CellToCell_Edges() has been removed. Use aggregate_NICHESObject().
  • compute_PseudoBulk_Connectivity() and create_Edgelist_PseudoBulk() have been removed. To summarize by cell type, sum $edge.data using the cell types in $edge.meta.
  • add_cell_meta() now replaces $cell.meta in full. Use add_meta_column() to change one column.
  • The default database in load_LRM_database() is now ConnectomeDB2025.
  • In add_neighborhood_edge_agg(), the autocrine category is now split into self.autocrine and other.autocrine.
  • load_LRM_database("omnipath") now falls back to the OmniPath web service when OmnipathR can’t reach its ancillary services.