PICTographPlus is a computational tool that integrates bulk DNA and RNA sequencing data to:
- Reconstruct Clone-Specific Transcriptomic Profiles
- Infer Tumor Evolution
- Identify Transcriptional Transitions Between Clones
The tool infers tumor clonal evolution from single or multi-region sequencing data by modeling the uncertainty of mutation cellular fraction (MCF) in small somatic mutations (SSMs) and copy number alterations (CNAs). Using a Bayesian hierarchical model, it assigns SSMs and CNAs to subclones, reconstructing tumor evolutionary trees that adhere to principles of lineage precedence, sum condition, and optional constraints based on sample presence. For deconvolution, PICTographPlus integrates tumor clonal tree structures with clone proportions across samples to resolve bulk gene expression data. It optimizes an objective function that minimizes discrepancies between observed and predicted sample-level gene expression while imposing a smoothness penalty, ensuring that closely related clones display greater gene expression similarity. Lastly, the tool conducts pathway enrichment analysis to identify statistically significant alterations in pathways connecting tumor clones.
runPictograph– Tumor evolution inference using genomic datarunDeconvolution– Bulk RNA expression deconvolution based on tumor evolutionrunGSEA– Gene Set Enrichment Analysis (GSEA) for transcriptomic differences between clones
- Uses Bayesian hierarchical modeling to infer tumor clonal evolution.
- Deconvolves bulk gene expression data using tumor clonal tree structures with 7 model variants.
- Performs pathway enrichment analysis to highlight significant transcriptomic alterations.
| Scenario | Recommended model | Why |
|---|---|---|
| Default — matched normal sample available | elastic_net (λ=0.01) |
Best synthetic F1 (0.347) and Sensitivity (0.368). |
| With-normal, if interpretability favoured | tree_delta (λ=0.05) |
Nearly-tied F1 (0.339) with an explicit tree-structured prior; also strongest on with_extnorm. |
| With-normal, prioritise precision / low-FDR | adaptive (λ=0.50) |
Highest MCC in with_normal (0.248). |
| Tumor-only (no normal reference) | adaptive_v2 (λ=0.50) |
Best F1 (0.348), Sensitivity (0.360), MCC (0.256). |
| External (population-average) normal only | tree_delta (λ=0.05) |
Best F1 (0.293) and Sensitivity (0.276). |
| Highest absolute expression recovery (Pearson r) | plain (λ=0.10) |
Top star-topology Pearson r (0.942) among 7 models. |
JAGS must be installed separately. Download it from: https://mcmc-jags.sourceforge.io
Run the following command in R:
# Install from GitHub
install.packages("devtools")
devtools::install_github("KarchinLab/pictographPlus", build_vignettes = TRUE)PICTographPlus was developed under R (4.4.2). All package versions during development can be found at installed_packages.csv
Detailed tutorial can be accessed through vignette.
library(pictographPlus)
vignette("pictographPlus", package = "pictographPlus")If you use PICTographPlus in your research, please cite the archived software release:
Lai J, Yang Y, Karchin R (2026). pictographPlus: Reconstructing Clone-Resolved Transcriptional Programs from Bulk Tumor Sequencing. R package version 1.1.1. Zenodo. https://doi.org/10.5281/zenodo.19896732
You can also retrieve a bibentry from R with:
citation("pictographPlus")Processed data generated in the companion study — clone-level expression matrices, inferred clonal trees, and edge-level GSEA results — are available on Mendeley Data at https://doi.org/10.17632/cv66sgfcfn.2.
The benchmarks and applications in the manuscript use the following public datasets:
| Dataset | Accession |
|---|---|
| wellDR-seq scDNA/scRNA co-profile (benchmarking) | GEO: GSE261713 |
| IPMN WES + RNA-seq | dbGaP: phs002225.v3.p1 |
| TRACERx NSCLC WES | EGA: EGAD00001009825 |
| TRACERx NSCLC RNA-seq | EGA: EGAD00001009862 |
| PanCuRx PDAC WGS | EGA: EGAD00001004551 |
| PanCuRx PDAC RNA-seq | EGA: EGAD00001004548 |
All datasets were used in accordance with the respective data access agreements.