Synapse-level BANC table with neuropil / NT enrichment — banc_synapses

Returns a lazy arrow::open_dataset() handle pointing at the per-synapse enriched parquet for BANC v888. The table carries ~169 M rows with neuropil membership, region, side and full per-transmitter Eckstein et al. (2024) prediction probabilities for each synapse. Apply further dplyr::filter() calls before dplyr::collect() — predicate pushdown skips parquet row groups that don't match.

Usage

banc_synapses_enriched(version = c("v2", "v3"), overwrite = FALSE)

Arguments

version: "v2" (default; paper synapses) or "v3" (updated synapses, still in testing).
overwrite: Logical. If TRUE, re-download the cached parquet even if it already exists.

Value

An arrow_dplyr_query lazy handle. Chain dplyr::filter() and dplyr::collect() to materialise a data frame.

Details

Source: gs://lee-lab_brain-and-nerve-cord-fly-connectome/compiled_data/banc_888/banc_888_synapses_<version>_enriched.parquet. The file is large (~9.6 GB for v2, ~15 GB for v3); the first call downloads it under tools::R_user_dir("bancr", "cache") and subsequent calls reuse the cache. Use overwrite = TRUE to force a refresh.

There is no CAVE equivalent of this enriched table (CAVE exposes the raw synapses_v<version> table without neuropil / region / NT enrichment), so there is no source / fallback argument.

Examples

if (FALSE) { # \dontrun{
library(dplyr)

# Lazy handle against the cached parquet
syn <- banc_synapses_enriched()

# Pull synapses in the right mushroom body calyx
mb_ca_r <- syn %>%
  filter(neuropil == "MB_CA_R") %>%
  collect()

# Synapses involving a specific neuron
me <- "720575941633499884"
my_syn <- syn %>%
  filter(pre_root_id == me | post_root_id == me) %>%
  collect()
} # }