Filling missing loci

Chromatin tracing misses loci: a probe that did not hybridise, a spot that was not detected, a locus dropped by the aligner. A trace then has gaps — rows with no coordinates — and every distance that involves them is missing. uchrom.im.impute fills them, either from the trace itself or from the other traces of the population.

Method

impute_coordinates(cd, method=...)

What it uses

linear interpolation

"linear"

the neighbouring loci of the same trace

cubic spline

"cubic"

the same trace, a smooth curve through its loci

SnapFISH-IMPUTE (Yu et al.)

"snapfish"

traces of the population whose conformation is similar

 
from uchrom.im.impute import impute_coordinates

filled = impute_coordinates(cd, method="snapfish")       # also: uc.pp.impute(cd, method="snapfish")

The result is a new ChromData with the same spots and a spot column imputed marking the filled ones; a locus a method cannot fill stays missing (SnapFISH-IMPUTE, for instance, leaves a locus that no similar trace observed). Imputed coordinates are estimates: evaluate_imputation hides observed loci, fills them with each method and measures the error, as the tutorial does on Bintu et al. 2018 traces:

from uchrom.im.impute import evaluate_imputation

holdout = evaluate_imputation(cd, methods=("linear", "cubic", "snapfish"), frac=0.1, seed=0)
holdout.summary                                          # median error per method

Tutorial

API: uchrom.im.