BGC Extraction

Algorithm to smooth contiguous gene cluster predictions into single regions.

class gecco.refine.ClusterRefiner(object)

A post-processor to extract contiguous clusters from CRF predictions.

__init__(threshold: float = 0.8, criterion: str = 'gecco', n_cds: int = 5, n_biopfams: int = 5, average_threshold: float = 0.6, edge_distance: int = 0) → None

Create a new ClusterRefiner instance.

Parameters:

• threshold (float) – The probability threshold to use to consider a protein to be part of a gene cluster.

• criterion (str) – The criterion to use when checking for cluster validity.

• n_cds (int) – The minimum number of genes a gene cluster must contain to be considered valid. If criterion is gecco, then this is the minimum number of annotated CDS.

• n_biopfams (int) – The minimum number of biosynthetic Pfam domains a gene cluster must contain to be considered valid (only when the criterion is antismash).

• average_threshold (int) – The average probability threshold to use to consider a gene cluster valid (only when the criterion is antismash).

• edge_distance (int) – The minimum distance from the edge the gene cluster must be located (it may start at an edge, but must span for longer than edge_distance), in number of annotated genes (only when the criterion is gecco).

iter_clusters(genes: List[Gene]) → Iterator[Cluster]

Find all clusters in a table of CRF predictions.

Parameters:

genes (list of Gene) – A list of genes with probability annotations estimated by ClusterCRF.

Yields:

gecco.model.Cluster – Valid clusters found in the input with respect to the postprocessing criterion given at initialisation.