NSP

missionbio.demultiplex.protein.nsp.NSP

class NSP(jitter: float = 0.5, random_state: Optional[int] = None, sample_size: float = inf, max_zero_read_cells: float = 0.05)

Noise corrected and Scaled Protein counts (NSP)

This is a novel normalization method for Tapestri protein data. It is based on the assumption that the background and signal are linearly dependent on the total read counts. The method fits a linear model to the background and signal and then corrects the read counts for the linear dependence.

Functions

transform(reads[, scale])	Normalize read counts.
scaling_factor(reads[, jitter])	Identifies the appropriate scaling factor for oversequenced runs.
plot()	Plot the signal and background

param jitter:

float The standard deviation of the jitter to be added to the read counts before applying the normalization. The jitter is sampled from a normal distribution cenetered at 0. This is only applicable for NSP and asinh

param random_state:

int The random state to use for the NSP and asinh methods.

param sample_size:

int The number of cells to use to estimate the signal and background linear models. All cells are used when set to np.inf.

param max_zero_read_cells:

float The maximum percentage of cells that can have 0 reads after applying the scaling factor. If more than this percentage of cells have 0 reads, the scaling factor is considered too high and is by default set to 1.