Analyze What Papers Actually Reported

Data sources to use (public + paper-linked)

• DepMap / PRISM drug–sensitivity & gene-expression matrices (use DepMap/PRISM to correlate ABCB1 with replication-competent virus sensitivity proxies and with compounds selective for MDR1-high cells) — linked in the PAC-1 study DepMap/PRISM used to associate ABCB1 with drug sensitivity
• GEO RNA-Seq dataset GSE89307 (MDR1 expression in tumor-specific CD8 and related samples) for efflux/transport expression patterns; paper describes MDR1 dynamics and provides a pipeline example GSE89307 RNA-Seq dataset for MDR1 dynamics
• Preclinical OV MDR1-driven oncolytic adenovirus data (Ad5/3MDR1E1) showing MDR1 promoter activity drives replication in MDR1-high ovarian cells; use as a validation/biological prior MDR1 promoter drives oncolytic adenovirus replication in ovarian cancer

Primary analysis questions (operational)

1. Is ABCB1 (MDR1) mRNA/protein expression in ovarian cell lines / tumors quantitatively correlated with proxies for viral/oncolytic replication (e.g., adenoviral E4 qPCR replication, permissiveness assays) or with replication-sensitive compound responses in PRISM?
2. Does MDR1 expression predict replication of MDR1-promoter driven oncolytic constructs versus CMV-driven controls (using experimental fold-change replication metrics reported in the 2011 study)?
3. Which covariates confound the relationship (e.g., prior chemotherapy exposure, tumor histology, ABCB1 genomic fusions, ABCB1 copy number, other ABC transporters)?

Stepwise bioinformatics & coding plan (reproducible pipeline)

1) Acquire & harmonize datasets

• Download DepMap expression (logTPM/TPM) and PRISM viability matrix; extract ovarian lineage cell lines and all cell lines for comparison (DepMap release used in PAC-1: DepMap PRISM 24Q4) DepMap/PRISM as source
• Download GEO GSE89307 (for transcript-level MDR1 patterns and normalization examples) GSE89307 RNA-Seq
• If available, obtain primary OV patient tumor RNA-seq and clinical chemo-exposure metadata (TCGA-OV or institutional cohorts) — use to stratify by prior chemo exposure (MDR1 tends to be higher in pretreated tumors per the 2011 study).

2) Preprocess & QC

• Expression: use TPM or log2(TPM+1) harmonized across datasets; remove genes with low counts; for PRISM link gene symbols to DepMap gene IDs using HGNC mapping.
• Metadata harmonization: map cell lines to tissue (OV), prior drug-resistance annotations (e.g., paclitaxel-resistant OVCAR8-PTX), ABCB1 copy-number, and fusion events when available (SLC25A40-ABCB1 fusion reported in PAC-1 study increased ABCB1 expression dramatically) ABCB1 fusion raises expression

3) Primary statistical analyses

1. Correlation analysis: compute Spearman rho between ABCB1 expression and PRISM viability AUC or IC50 for (a) replication-sensitive agents (if available) and (b) PAC-1 as positive control (replicate PAC-1–ABCB1 correlation reported) — correct p-values for multiple testing (Benjamini-Hochberg).
2. OV subset: repeat correlations within ovarian lineage only (greater biological relevance to OV replication hypotheses and MDR1 promoter activity observed in pretreated OV tumors) MDR1-high in pretreated ovarian tumors
3. Regression models: linear models (or robust regression) predicting replication-readout (or drug-sensitivity) from ABCB1 expression, controlling for covariates: prior resistance status, CNV, key co-expressed transporters (ABCC1/ABCG2), TP53 or other driver mutations, and lineage. Use mixed-effects models if multiple assays per cell line exist.
4. Threshold analysis: derive cutpoints (e.g., upper quartile ABCB1) and compare replication outcomes (Wilcoxon/Mann-Whitney) to mimic MDR1 promoter–driven vs CMV-driven replication comparisons from the 2011 oncolytic study.

4) Validation & mechanistic cross-checks

• Validate with ex vivo OCMs if accessible (authors of PAC-1 and OV oncolytic work used ex vivo OCMs) — correlate ABCB1 expression with replication or cytotoxic readouts. Use the 2011 adenovirus replication metrics (E4 qPCR) as a target metric for what replication fold-change to expect.
• Sensitivity analyses: (a) remove fusion/high-copy outliers to test robustness; (b) stratify by prior chemo exposure (pretreated vs naive) since MDR1 is often elevated after prior chemotherapy in OV datasets reported in 2011.

5) Visualization & reporting (Plotly-ready)

• Scatter plots: ABCB1 (x-axis) vs replication readout or PRISM AUC (y-axis) with LOWESS trendline, color-coded by lineage (ovarian vs others) and marker-size by ABCB1 CNV; provide Spearman rho and adjusted p-value in the panel caption.
• Boxplots: replication readout split by ABCB1 tertiles/quartiles; include pairwise tests and effect sizes (Cliff’s delta).
• Forest plot: regression coefficients (ABCB1 and covariates) with 95% CI to show adjusted association strength.

6) Reproducible code & workflow

• Workflow manager: snakemake or Nextflow for reproducibility (data download, preprocessing, analysis, figures, report). Use conda or mamba environments for dependencies (pandas, numpy, statsmodels, scikit-learn, plotly, scipy, anndata if single-cell needed).
• Version control: GitHub with DOI via Zenodo; provide a README with dataset versions (DepMap release) and exact commands.

Minimal viable analysis (MVA) — runnable in ~1 day

1. Download DepMap expression + PRISM (cell-line metadata filtered to ovarian lineage).
2. Compute Spearman correlation ABCB1 vs PRISM AUC for PAC-1 (positive control) and any oncolytic/replication-relevant compound if present.
3. Create three Plotly figures (scatter + trendline; box by ABCB1 quartile; regression coefficient barplot) and a short report summarizing effect sizes and p-values.

Potential pitfalls, biases, and blindspots

• Correlation ≠ causation: MDR1 expression may co-vary with prior chemo exposure or other resistance mechanisms; regression + stratification reduces confounding but cannot fully prove causality.
• Dataset differences: in vitro cell lines and ex vivo OCMs differ from in vivo tumor microenvironment; oncolytic replication measured in 2011 used specific adenoviral constructs and MOIs — replication readouts may not be directly present in PRISM.
• Measurement error: ABCB1 mRNA may not perfectly reflect functional protein/efflux activity; where possible include functional efflux proxies (rhodamine, rhodamine-123 efflux assays) or protein data.
• Publication/selection bias: MDR1-high examples may be enriched in literature; include negative-control analyses (other ABC transporters) to check specificity.

Validation experiments to follow bioinformatics findings

1. In vitro: infect matched OV cell lines with MDR1-promoter-driven adenovirus vs CMV-driven control; measure E4 copies by qPCR at multiple timepoints and compare replication fold-change across ABCB1 expression strata (mimic 2011 methods) Replication metrics in oncolytic adenovirus study
2. Functional: measure rhodamine-123 or MTG efflux to quantify functional MDR1 efflux activity and test whether efflux, rather than mRNA alone, better predicts replication or oncolytic killing (supported by MDR1 efflux assays in the T cell study) MTG/DCV efflux assays for MDR1 function