training_module: training_module

In ThomasDCY/AdaLiftOver: Adaptive liftOver

training_module

Description

This function takes GRangesList with 'epigenome' and 'grammar' meta columns and evaluate the predictive power of overlapping with the true orthologous genomic regions.

Usage

training_module(
  gr_candidate,
  gr_true,
  max_filter_proportion = 0.4,
  interaction = FALSE
)

Arguments

gr_candidate	A GRangesList object with 'epigenome' and 'grammar' meta columns.
gr_true	A GRanges object as the true orthologous genomic regions that we aim to derive.
max_filter_proportion	The maximum proportion of candidate target regions that are allowed to be filtered out. Default is 0.4.
interaction	If we include the interaction term in the logistic regression. Default is FALSE.

Value

A data.table with the following columns:

auroc	The optimal AUROC.
aupr	The optimal AUPR.
b.intercept	The estimated regression coefficient for the intercept.
b.epigenome	The estimated regression coefficient for the epigenome signal.
b.grammar	The estimated regression coefficient for the sequence grammar.
b.interaction	The estimated regression coefficient for the interaction term if included.
max_filter_proportion	The input parameter max_filter_proportion.
threshold	The suggested score threshold.
precision	The precision associated with the suggested score threshold.

Author(s)

Chenyang Dong [e-mail]

Examples

data('training_module_example')
res <- training_module(gr_candidate, gr_true)

ThomasDCY/AdaLiftOver documentation built on June 2, 2025, 12:35 p.m.