Select cross-technology anchor features on a training pool — select_cross_tech_anchor

A selected anchor must be detected in every required technology class present in `required_technologies`. Selection is based only on `train_expr` and `train_meta`; held-out samples should never be passed here.

Usage

select_cross_tech_anchor_features(
  train_expr,
  train_meta,
  technology_col = "technology",
  required_technologies = c("NGS", "Toray", "qPCR", "NanoString"),
  min_detection_rate = 0.05,
  n_anchor = 20L
)

Arguments

train_expr: Numeric matrix, samples x features, training pool.
train_meta: Data frame aligned to `train_expr`.
technology_col: Column identifying technology class.
required_technologies: Technology classes required for selection.
min_detection_rate: Minimum per-technology detection rate.
n_anchor: Maximum number of anchors to select.

Value

Data frame with detection rates and `selected_yes`.