Abstract
Tumors comprise related clones whose evolutionary structure and clone-specific transcriptional programs are obscured in bulk sequencing data. While clonal architecture can often be inferred from bulk DNA sequencing, recovering clone-resolved transcriptional programs from bulk RNA-seq has remained largely inaccessible. We present PICTographPlus, a probabilistic framework that reconstructs clone-resolved transcriptomes by integrating DNA-inferred clonal phylogenies with bulk RNA-seq. PICTographPlus fits a phylogeny-regularized mixture model that aligns DNA-derived clone proportions with observed expression, enabling inference of clone-specific gene expression and localization of pathway gains and losses to specific evolutionary transitions. Using single-cell derived clone structures and simulated bulk mixtures, we demonstrate robust recovery of clone-level gene set regulation across tumor purities and sampling densities. Applications to lung and pancreatic cancer cohorts reveal clone-restricted transcriptional programs associated with tumor suppressor loss and metastatic progression. PICTographPlus transforms widely available bulk assays into evolutionary, clone-resolved transcriptional maps, enabling retrospective and cohort-scale analyses without specialized experimental data.