br_seed(), br_seed_screen(), br_seed_select(), and br_seed_model()
implementing the SEED (Selection of Essential prognostic genes from Expression Data)
pipeline for identifying cancer prognosis target genes without control samples (#57).
Based on Yang et al. (2025) Cancer Letters, DOI: 10.1016/j.canlet.2025.217960.
The four-step pipeline includes:
br_seed_screen): Univariate screening via Cox/logistic regression or
Spearman correlation against one or more clinical indicators, with p-value
threshold filtering and significance intersection across indicators.enrich argument).br_seed_select): Sequential gene selection through Lasso
regularization (glmnet), multivariate regression, and stepwise selection
with reproducibility via seed.br_seed_model): Risk score model construction with coefficient
extraction and optional z-score expression scaling.
All functions return S3 list objects with cli-based print methods, reuse the
existing br_pipeline() engine for batch regression, and support mirai
parallelization via n_workers.br_get_model_stats() returning a tidy data.frame of per-model summary
statistics (N, events, C-index, AIC, LR test p, PH test p for Cox; N, AIC,
deviance for GLM; N, R-squared for LM), eliminating the need for manual
supplementary model fitting to obtain diagnostic metrics (#68).br_diagnose().FGFR3::TACC3, EGFR-AS1,
1p/2q, gene name, R reserved words (if, TRUE, NA), and
backtick-quoted user input in x, x2, and y variables (#69). Refactored
repair_names() with remove_backticks() for robust quoting, and applied
quoting to response variables in br_set_y().biocViews field to DESCRIPTION for Bioconductor Suggests compatibility.br_compare_models() and br_show_forest_comparison() (#54).qs with qs2.br_show_fitted_line() and br_show_fitted_line_2d() documentation
for compatibility with visreg 3.0, which removes the gg argument and
renames line.par, fill.par, points.par to line, fill, points (#65).devtools::check(env_vars = c('_R_CHECK_DEPENDS_ONLY_' = "true")) and fixed reported check issues.br_pipeline() to reduce running time.br_show_coxph_diagnostics().Enhancements & New Features:
br_show_forest_circle()for circular forest plots.br_show_coxph_diagnostics().br_show_nomogram()for clinical prediction modeling.dry_runoption to br_pipeline()for pipeline validation.Fixes & Improvements:
br_show_nomogram().br_show_residuals() to show residuals vs fitted plot for regression models.br_set_x() and br_set_x2() to properly handle different input
types, including valid column names (in R), invalid column names, and
model formula terms.br_get_model_names() and br_rename_models() to get and set model names.run_parallel in br_run().br_get_model() by merging it with br_get_models().options().exponentiate option appropriately in br_run().br_avail_methods_use_exp() to include c("poisson", "quasipoisson").log_first option to br_show_forest().: for interaction term combinations throughout the package.br_show_forest().br_get_model() couldn't properly process multiple string inputs for the idx parameter.