← New search

Molecular Biomarker of Drug Resistance Developed From Patient-Derived Organoids Predicts Survival of Colorectal Cancer Patients.

Front Oncol · 2022

PubMed 35425715 ↗ DOI 10.3389/fonc.2022.855674 ↗ pmid-35425715

L1 53/100 ★ ☆ ★ ★ 3/4

Why this verdict

The main results reproduced, with only marginal, non-material deviations.

Reproduced on the brainbox compute

Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)

Content-critical question only partially held

+2 pts

From: Q5 · Derivability / plausibility 🟡

Content-critical question only partially held

+2 pts

From: Q7 · Core claim 🟡

Content-critical question only partially held

+2 pts

From: Q8 · Severity of the miss (overall human judgment) 🟡

Minor / cosmetic deviation

+1 pts

From: Q3 · Location of the main deviation 🟡

Minor / cosmetic deviation

+1 pts

From: Q6 · Severity of the deviation 🟡

Concordant (toward reproduced)

Code + data deposited & functional

-2 pts

From: Data & code availability —Available & functional

Total score +6

✓ What held up

✓Reported values were directly comparable

What did not (or only partly)

🟡Could not use the authors’ exact input data
🟡A deviation arose in the data or preprocessing
🔴A deviation was attributed to the published material
🟡Reported values were not (fully) derivable from the shared data
🟡The deviation was non-trivial in magnitude
🟡The central claim did not (fully) hold under reproduction
🟡Overall, the reproduction showed a material discrepancy

How its reproducibility compares

53/100

Reproducibility score

1.2 SD below mean

vs. all fields · 1187 studies

🎯 Scores higher than 14% of all assessed papers rank 1014 of 1187 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

▸Reproduction agent’s raw note

Described well enough to attempt a clean 1:1: the paper publishes the exact 5-gene Drug-Resistant Score (DRS) formula and validates it on the public GSE40967 cohort (the brief's data accession), reporting univariate Cox P=8e-04 for OS. We applied the published model to GSE40967 (GPL570) via GEOquery on «our HPC». RESULT = PARTIAL: the biomarker's DIRECTION reproduces in every configuration (high DRS -> worse OS, HR>1) and the association is significant under the paper's described z-scored + maxstat-dichotomized pipeline (log-rank P=0.0023-0.021), but the EXACT reported P=8e-04 is not matched and the verbatim continuous-DRS Cox is non-significant (P=0.44). The gap is well-explained by two undocumented choices: (1) the paper used n=233 whereas the public cohort has 573 OS-evaluable samples and the subset basis is not stated, and (2) 'gene expression level' normalization + multi-probe collapse are unspecified (coefficients fit on TCGA FPKM, applied across 4 array platforms). No fabrication indicated; this is a reproducibility/documentation gap. NOT attempted (out of scope / hard 20%): LASSO re-derivation on TCGA-CRC (coefficients already published), the other 3 GEO validation cohorts, all wet-lab organoid work, and the Fig-8 downstream analyses (GSEA/enrichplot, TMB, ESTIMATE immune/stromal, CIBERSORT). The brief's 'code' link (GuangchuangYu/enrichplot) is a third-party GSEA visualization tool, not an analysis repo; reproduction applied the paper's published model to public data per its prose Methods (P16).

💻 Code ↗ 🗄 Data: GSE40967

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

v1 current initial assessment Score 53

assessed: 2026-06-14 ⛓ 230618ac6839

✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced: 2026-06-14
Rubric version: v1.0
Assessed by: 🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15
no human curator yet
Last updated: 2026-09-19

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: sonnet

Founding hypothesis

A gene signature and drug-resistant score model built from 5-fluorouracil (5-Fu) sensitivity data of patient-derived colorectal cancer organoids can serve as a molecular biomarker to predict the survival of colorectal cancer patients.

Core claims

★ Patient-derived colorectal cancer organoids (CRCOs) can be established from surgical CRC tissue with an 82% success rate (41/50) resource
★ Organoid size change assay effectively differentiates 5-Fu sensitive (34.1%) versus resistant (65.9%) CRCOs, reflecting clinically observed heterogeneity in 5-Fu response finding
★ Transcriptome sequencing of CRCOs identified DEGs associated with 5-Fu resistance, including molecules and pathways previously reported in 5-Fu resistance finding
★ A five-gene drug-resistant score model (DRSM: CACNA1D, CIITA, PFN2, SEZ6L2, WDR78) was constructed via LASSO regression from organoid-derived DEGs in the TCGA-CRC cohort method
★ The drug-resistant score (DRS) is an independent prognostic factor for overall survival in CRC patients in the TCGA-CRC cohort (P < 0.001) finding
★ The DRSM was validated in four independent GEO cohorts and predicts survival across different patient subgroups finding
★ DRS-high and DRS-low patients differ in molecular pathways, tumor mutational burden, immune-related pathways, immune/stromal scores, and immune cell composition finding
Organoid size change is as effective as the CellTiter-Glo 3D cell viability assay for judging drug sensitivity, while being more economical and easier to use method

Experimental setups

Assay	System	Perturbation	Readout	Platform
organoid culture establishment	primary CRC tumor tissue (human surgical specimens)	none	organoid formation success rate	—
drug sensitivity assay (organoid size change)	colorectal cancer organoids (CRCOs)	10 μM 5-Fu treatment	organoid size change ratio (day24/day0)	ZEISS Vert.A1 microscope; Image-Pro Plus 6.0
bulk transcriptome sequencing (RNA-seq)	CRCOs (5-Fu sensitive vs resistant; pre- vs post-5-Fu treatment)	5-Fu treatment vs untreated	differentially expressed genes (DEGs)	Illumina Novaseq; NEBNext Ultra RNA Library Prep Kit
functional enrichment analysis (GSEA)	CRCO transcriptome data	5-Fu resistance status	enriched molecular pathways	clusterProfiler / enrichplot R packages
gene expression and survival cohort analysis	TCGA-COAD/READ CRC patient cohort	none (observational)	DRS, overall survival, TMB, immune score, stromal score, immune cell proportion	TCGAbiolinks
gene expression and survival validation	4 GEO CRC patient cohorts	none (observational)	survival prediction by DRSM	GEOquery
LASSO and Cox regression modeling	TCGA-CRC gene expression and survival data	none	gene coefficients / drug-resistant score model	glmnet (v4.0-2) R package; coxph

Key results

– 41 of 50 (82%) CRC tissues yielded viable organoid cultures 82%
– 14/41 (34.1%) CRCOs were 5-Fu sensitive and 27/41 (65.9%) were resistant
– DEGs identified in sensitive vs resistant untreated CRCOs and in pre- vs post-5-Fu-treatment CRCOs overlapped known 5-Fu resistance genes/pathways
– Five-gene DRSM (CACNA1D, CIITA, PFN2, SEZ6L2, WDR78) constructed from 26 candidate genes via LASSO with 5-fold cross-validation
– DRS is an independent predictor of overall survival in multivariate Cox analysis in the TCGA-CRC cohort P < 0.001
– DRSM predicted survival across four GEO validation cohorts and across different patient subgroups
– DRS-high vs DRS-low groups differed in TMB, immune-response pathways, immune score, stromal score, and immune cell proportions
– A 36.42% cutoff of organoid size change (day24/d0) was used to classify 5-Fu sensitivity, consistent with prior validated methodology 36.42%

Key statistics

pvalue P < 0.001 (DRS as independent prognostic factor for OS, multivariate Cox analysis, TCGA-CRC cohort)
count 41/50 (82%) (CRCO organoid culture establishment success rate)
count 14/41 (34.1%) sensitive, 27/41 (65.9%) resistant (5-Fu drug sensitivity classification of CRCO lines)
other 36.42% cutoff (size day24/d0) (cutoff for classifying 5-Fu sensitivity of CRCOs)
count 26 genes with non-zero coefficients (candidate genes retained after LASSO regression filtration)
count 5 genes with P < 0.05 (genes retained as independent prognostic factors for the DRSM)
pvalue P < 0.05 (significance threshold for DEG identification via limma)
other DRS = GEL(CACNA1D)×-0.0563 + GEL(CIITA)×-0.0356 + GEL(PFN2)×0.0332 + GEL(SEZ6L2)×0.0378 + GEL(WDR78)×-0.0386 (drug-resistant score (DRS) model formula)

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

The study classified 41 patient-derived colorectal cancer organoids as 5-Fu sensitive (n=14) or resistant (n=27) using organoid size change, then identified DEGs by transcriptome sequencing analyzed with the R package limma. DEGs associated with both 5-Fu resistance and patient survival in the TCGA-CRC cohort were progressively filtered by univariate and multivariate Cox regression, then by LASSO regression with 5-fold cross-validation, yielding a 5-gene Drug-Resistant Score Model (DRSM). The model was evaluated using Kaplan-Meier curves with log-rank tests and validated across four independent GEO cohorts, with multivariate Cox regression confirming DRS as an independent prognostic factor for overall survival.

Replicationbiological Sample size50 surgically resected tumor tissues yielding 41 organoid lines (82% success rate); 14 classified sensitive, 27 resistant; TCGA-CRC cohort used for model development; four GEO cohorts used for external validation; individual cohort sample sizes not stated in available text Groups5-Fu sensitive vs resistant CRCOs (organoid level); pre-treatment vs post-treatment surviving CRCOs; DRS-high vs DRS-low CRC patients in TCGA and GEO cohorts Pairingmixed Randomization/blindingnot stated DispersionSEM Exact p-valuesno Effect sizesyes Multiplicity correctionBenjamini-Hochberg FDR

Statistical tests used

Test	Applied to	n	Assumptions
limma empirical Bayes moderated t-statistic	DEG identification: (1) 5-Fu sensitive vs resistant untreated CRCOs; (2) pre-treatment CRCOs vs post-treatment surviving CRCOs	41 organoid lines total (14 sensitive, 27 resistant); n per group for pre/post comparison not stated	not stated
Gene Set Enrichment Analysis (GSEA) with Benjamini-Hochberg FDR correction	Functional pathway enrichment of transcriptomic data; 1,000 permutations	—	not stated
Univariate Cox proportional hazards regression	Screening each DEG for association with overall survival in TCGA-CRC cohort; P < 0.05 threshold for retention	—	not stated
Multivariate Cox proportional hazards regression	Identification of independent prognostic factors from LASSO-retained genes; confirmation of DRS as independent prognostic factor in TCGA-CRC and GEO cohorts	—	not stated
LASSO regression (glmnet v4.0-2) with 5-fold cross-validation	Feature selection from univariate Cox-filtered genes to construct DRSM; lambda selected by minimizing cross-validated error	—	not stated
Log-rank test	Kaplan-Meier survival curve comparisons between DRS-high and DRS-low groups in TCGA and GEO cohorts	—	not stated
MaxStat maximum rank statistic	Optimal cutpoint selection to dichotomize patients into DRS-high and DRS-low groups	—	not stated
Wilcoxon rank-sum test	Comparison of two groups (specific comparisons not fully specified in available text)	—	not stated
Two-sided Fisher exact test	Analysis of contingency tables (specific comparisons not fully specified in available text)	—	not stated
Spearman correlation and distance correlation	Correlation coefficient analyses (specific variable pairs not fully specified in available text)	—	not stated

Approaches that could also have been used

limma (an empirical Bayes method originally designed for microarray data) was applied to RNA-seq FPKM values for DEG identification

↳ Could also: DESeq2 or edgeR, which model raw RNA-seq read counts with a negative binomial distribution, could also be used for DEG analysis — DESeq2 and edgeR are purpose-built for count-based RNA-seq data and explicitly model count overdispersion; they are among the most extensively benchmarked tools for this data type and are frequently recommended when count-level data are available as the analysis input
An optimal data-driven cutpoint (MaxStat maximum rank statistic) was used on the TCGA training cohort to dichotomize patients into DRS-high and DRS-low groups

↳ Could also: The median DRS, pre-specified quartiles, or retaining DRS as a continuous predictor in Cox regression could also be used — Data-driven cutpoint optimization on the same dataset used for evaluation can inflate apparent group separation and reduce external generalizability; continuous modeling or a pre-specified split preserves statistical power and avoids this dependency
Univariate Cox regression (P < 0.05) was used as a pre-screening step before LASSO feature selection, running a separate test for each DEG without a multiplicity correction at that stage

↳ Could also: Applying LASSO directly to all DEGs without a prior univariate screen, or applying Benjamini-Hochberg FDR to the univariate p-values before retaining genes, could also be used — Running many uncorrected univariate tests before penalized regression can admit marginally significant or correlated genes; FDR-controlled pre-screening or direct penalized regression makes the selection process more self-contained and reproducible
Organoid size change data were summarized with SEM (standard error of the mean) from 8 replicates

↳ Could also: Standard deviation (SD) or a 95% confidence interval could also be used to describe the spread of replicate measurements — For a small number of replicates, SD conveys the observed biological or technical variability in the measurements directly, whereas SEM reflects precision of the mean estimate; both are informative, and the choice affects how readers interpret the magnitude of variability
5-fold cross-validation was used within the TCGA training cohort to select the optimal LASSO penalty (lambda)

↳ Could also: Bootstrap resampling or leave-one-out cross-validation (LOOCV) could also estimate prediction error and select lambda — Bootstrap resampling provides lower-variance optimism-corrected estimates of predictive performance, particularly in moderate-sized cohorts; LOOCV is another established alternative; the choice of internal validation method can affect the stability of the selected gene set and reported performance metrics
Survival differences between DRS groups were evaluated with the log-rank test, which implicitly assumes proportional hazards

↳ Could also: A Cox model with DRS as a continuous predictor, or a restricted mean survival time (RMST) analysis, could also quantify survival differences — Treating DRS continuously in Cox regression avoids information loss from dichotomization; RMST does not require the proportional hazards assumption and provides a directly interpretable difference in survival time, which may be informative when proportional hazards has not been formally verified

Software: R/limma · R/clusterProfiler · R/glmnet 4.0-2 · R/survival (coxph) · R/MaxStat · R/pheatmap · R/enrichplot · R/TCGAbiolinks · R/GEOquery · R/forestplot · Image-Pro Plus 6.0

Citation network

Where this publication sits in the reproducibility-weighted citation graph — what it is built on, and what is built on it. Citation data from OpenAlex.

Authors · 8

1Lifeng Chen 2Bo Tian 3Wen Liu 4Wen Liu 5Hai‐tao Liang 6Yong You 7Weizhen Liu 8Weizhen Liu

Citations

Impact: low

Foundation confidence

None of its references are in our reproducibility record yet — its foundation cannot be assessed.

Topics

Oncology Drug resistance Organoid Biomarker

No assessed neighbours yet — the network grows as more papers are assessed.

Data lineage

The datasets this paper uses (text-mined from the full text via Europe PMC), and which other assessed papers stand on the same data. A shared dataset is a factual link — not a judgement.

GSE17538 GEO in Results (http://purl.org/orb/Results)

also used by 2 papers:

Indoleamine 2, 3-Dioxygenase 1 and CD8 Express...⚑ L1 48/100 ⚑
Enhancing chemotherapy response prediction via... L1 78/100

GSE14333 GEO in Results (http://purl.org/orb/Results)

also used by 1 paper:

Multi-omics analyses identify mannose phosphat... L1 61/100

GSE87211 GEO in Results (http://purl.org/orb/Results)

also used by 1 paper:

Enhancing chemotherapy response prediction via... L1 78/100

10.17632/rnrmjkvjjc.2 DOI in Data Availability (http://purl.obolibrary.org/obo/IAO_0000611)