score_annotation: Score an annotation source

Description Usage Arguments Value Note

View source: R/annotation_analysis.R

Description

This function takes MPRA data and an annotation source (or alternatively pre-fit priors) and checks how well the conditional prior improves upon the marginal prior by a prior ratio at the maximum likelihood estimates. A conditional:marginal ratio > 1 indicates that the conditional prior does better. A higher fraction of alleles for which this is true indicates an improved conditional prior.

Usage

1
2
3
4
5
6
7
score_annotation(
  mpra_data,
  annotations,
  nb_params = NULL,
  conditional_prior = NULL,
  marginal_prior = NULL
)

Arguments

mpra_data

a data frame of MPRA data

annotations

a data frame containing annotations of the same alleles in mpra_data

nb_params

an optional data frame of pre-fit maximum likelihood estimates for allele parameters

conditional_prior

an optional data frame containing pre-fit conditional priors

marginal_prior

an optional data frame containing a pre-fit marginal prior

Value

A fraction between 0 and 1 indicating the number of parameter estimates that have higher density under the conditional prior than under the marginal prior.

Note

Note that even with near-perfect annotations or conditional priors, the fraction of parameter estimate ratios above 1 will not approach 1. It will instead approach some difficult-to-determine limit defined by the noise in the assay. Thus this functionality should only be used to COMPARE annotation sources, rather than make discrete claims / measurements about individual annotation sources.


andrewGhazi/malacoda documentation built on Aug. 2, 2020, 12:54 a.m.