Abundance-ranked centroid clustering with likelihood-based best-parent selection and a distance-aware count-ratio merge guard. Supports both Levenshtein (handles indels) and Hamming distances.
Usage
super_cluster2(
input_path,
distance = 3,
method = c("lv", "hamming"),
barcode_col = "barcode",
counts_col = "counts",
output_dir = NULL,
file_pattern = "\\.csv$",
verbose = TRUE,
use_cpp = TRUE,
use_kmer_filter = TRUE,
kmer_size = 5L,
min_shared_kmers = 2L,
merge_ratio = 20,
error_rate = 0.005
)Arguments
- input_path
Character string or data.frame.
- distance
Numeric. Maximum edit distance. Default: 3.
- method
Character string. "lv" (Levenshtein, default) or "hamming".
- barcode_col
Character string. Default: "barcode".
- counts_col
Character string. Default: "counts".
- output_dir
Character string or NULL. Default: NULL.
- file_pattern
Character string. Default: "\.csv$".
- verbose
Logical. Default: TRUE.
- use_cpp
Logical. Default: TRUE.
- use_kmer_filter
Logical. Default: TRUE.
- kmer_size
Integer. Seed size for LV index. Default: 5.
Integer. Kept for API compatibility. Default: 2.
- merge_ratio
Numeric. Base count-ratio for the distance-aware merge guard. Effective ratio increases with distance. Default: 20.
- error_rate
Numeric. Approximate per-base error rate for likelihood scoring. Default: 0.005.
Value
A tibble with columns:
cluster_id, central_barcode, all_barcodes, all_counts, sum_counts.