Interpretation

Helpers for interpreting model predictions via in silico mutagenesis (ISM) and enhancer-to-gene (E2G) scoring:

• ISM: ism_variants lists every single-base substitution in an interval, and ism_matrix turns per-variant scores from DnaClient.score_ism_variants into a position-by-base matrix.

• Enhancer scoring: tile_interval, score_enhancers, get_active_enhancers, and assign_consensus_enhancers tile genomic intervals, score candidate elements from predict_interval_gradients() (or predict_interval_gradients_batch()) outputs, call active enhancer-gene links with modality-calibrated thresholds, and cluster called peaks into cross-biosample consensus intervals.

ISM

interpretation.ism.ism_variants(interval, ...)	Create a list of all possible single nucleotide variants for an interval.
interpretation.ism.ism_matrix(...[, ...])	Construct the ISM (position, base) matrix from individual ISM scores.

Enhancer scoring

Call and score candidate cis-regulatory enhancers from predict_interval_gradients() (or predict_interval_gradients_batch()) outputs and chromatin accessibility tracks. See Mapping lineage-specific enhancers at the BCL11A locus for an end-to-end walkthrough.

interpretation.enhancer_scoring.tile_interval(...)	Partitions interval into overlapping sliding-window tiles.
interpretation.enhancer_scoring.score_enhancers(...)	Scores candidate enhancers for the target gene of an input gradient.
interpretation.enhancer_scoring.get_active_enhancers(...)	Calls and merges active enhancer-to-gene links from input gradients.
interpretation.enhancer_scoring.assign_consensus_enhancers(...)	Assigns cross-biosample consensus peak bounds via reciprocal-overlap NMS.

alphagenome.interpretation.enhancer_scoring.DEFAULT_GRADIENT_SCORE_THRESHOLDS: Mapping[OutputType, float]

Calibrated default per-tile gradient_score thresholds by OutputType on 1 Mb genomic intervals.