detectPanel: Leakage-Aware Discovery of Small Biomarker Panels
Discovers small binary-classification biomarker panels from
count or expression matrices while prioritizing detectability, expression
stability, and univariate discrimination. Candidate filtering and panel
selection can be repeated inside nested cross-validation to reduce
information leakage. The package provides shared resampling splits,
exhaustive small-panel search, logistic model fitting with an automatic
ridge fallback for unstable separation-prone fits, out-of-fold evaluation,
selection-frequency summaries, and optional 'DESeq2'
differential-expression support. The nested model-selection workflow
follows Varma and Simon (2006) <doi:10.1186/1471-2105-7-91>, and the
optional differential-expression analysis uses Love, Huber, and Anders
(2014) <doi:10.1186/s13059-014-0550-8>.
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=detectPanel
to link to this page.