cpa.ComPertAPI.compute_uncertainty#

ComPertAPI.compute_uncertainty(covs, pert, dose, thrh=30)[source]#

Compute uncertainties for the queried covariate+perturbation combination. The distance from the closest condition in the training set is used as a proxy for uncertainty.

Parameters:

covs list of strings: Covariates (eg. cell_type) for the queried uncertainty
pert string: Perturbation for the queried uncertainty. In case of combinations the format has to be ‘pertA+pertB’
dose string: String which contains the dose of the perturbation queried. In case of combinations the format has to be ‘doseA+doseB’

Returns:

: min_cos_dist: float

Minimum cosine distance with the training set.

min_eucl_dist: float: Minimum euclidean distance with the training set.
closest_cond_cos: string: Closest training condition wrt cosine distances.
closest_cond_eucl: string: Closest training condition wrt euclidean distances.