Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

lncEvo: automated identification and conservation study of long noncoding RNAs.

BMC Bioinformatics · 2021
L1 61/100 3/4
Why this verdict

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

Reproduced on the brainbox compute brainarbeit.com
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: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +6
✓ What held up
  • Any deviation was negligible
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🟡Reported values were only indirectly comparable
  • 🟡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 central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
61/100
Reproducibility score
0.7 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 22% of all assessed papers rank 906 of 1173 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

lncEvo (Bryzghalov et al. 2021; authors' Nextflow pipeline; brief names STAR + SRR4421334). Described well enough to reproduce, with two minor repo-portability fixes documented (legacy python2.7 filter scripts; report-only multiqc/sweetviz steps incompatible with the pinned env). Reproduced THREE ways on «our HPC»: (1) C4 -- genuine fastp+STAR (2.7.3a, exact main.nf params) on the paper's own human run SRR4421334 -> 91.65% uniquely mapped, BIT-FOR-BIT identical to an independent earlier run (fully deterministic); (2) C1d -- FULL end-to-end lncEvo assembly+prediction module re-run from scratch on the paper's 3 human runs -> 11645 human lncRNA transcripts vs the authors' deposited 10716 (+8.7%, within-tol; difference mechanistically explained by Ensembl reference version drift, since lncEvo fetches the current release not the paper's pinned r103); (3) deposit re-audit (C1/C3) -- dog 5268 ~ Fig6 ~5000 (within-tol), human/chimp/horse counts consistent with 'intermediate'. NOT 1:1 but a strong, honest reproduction of the pipeline-derived results. NOT verifiable: the mouse pieces (Fig6 ~25000; headline 550/569 human-mouse reversal numbers) -- mmusculus is ENTIRELY ABSENT from the public deposit, flagged as not-substantiated-by-shipped-data (fabrication-watch); a human reviewer must confirm whether mouse results exist elsewhere. Full 5-species conservation re-run was out-of-scope by design.

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.

  1. v1 current initial assessment Score 85
    assessed: 2026-06-18 ⛓ c46242884d94
✎ 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.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

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
2026-06-24
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-18
no human curator yet
Last updated
2026-08-05

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

Because lncRNAs generally lack the strong sequence conservation from purifying selection seen in protein-coding genes, software designed for evolutionary analysis of protein-coding genes/transcripts is not applicable to lncRNAs, motivating a dedicated integrated pipeline for lncRNA identification and cross-species conservation study.

Core claims
  • lncEvo is an integrated Nextflow/Docker pipeline combining transcriptome assembly, lncRNA identification, and cross-species conservation analysis into a single workflow. resource
  • Software developed for evolutionary analysis of protein-coding genes is not applicable to lncRNAs because they lack strong sequence conservation from purifying selection. finding
  • lncRNAs display diverse conservation patterns (positional/syntenic, sequence-orthologous, and locus-only conservation) reflecting different modes of action. finding
  • The pipeline's redesigned cross-species genome alignment procedure significantly reduces computational time while maintaining comparable sensitivity relative to prior methodology. method
  • Individual pipeline modules (transcriptome assembly, lncRNA prediction) can be run independently without invoking the conservation module. method
  • lncEvo supports scalable cloud execution via AWS Batch with AWS Athena integration for ad hoc SQL querying of results. resource
  • More than 70 percent of lincRNAs cannot be linked to homologs in species diverged more than 50 million years ago. finding
  • lncEvo was validated by assembling transcriptomes and identifying/comparing lncRNAs across five species (human, mouse, chimp, horse, dog) using public RNA-seq data. method
Experimental setups
Assay System Perturbation Readout Platform
bulk RNA-Seq transcriptome assembly human, mouse, chimp, horse, dog (public SRA RNA-seq runs) none assembled transcriptome (GTF/FASTA), transcript expression (TPM) STAR, StringTie, Salmon
lncRNA identification/filtering assembled transcriptomes from the five test species none classification of transcripts as lncRNA vs coding based on length, ORF content, coding potential, and expression TransDecoder, Coding Potential Calculator (CPC), Cuffcompare
comparative genome alignment / conservation analysis pairwise species comparisons (human as reference vs mouse/chimp/horse/dog) none orthologous/syntenic lncRNA counterparts, exonic sequence identity LAST (lastal, last-train), axtChain, chainNet, slncky
RNA-seq quality control raw FASTQ reads from all five species none read quality metrics, adapter/contaminant content FastQC, fastp/bbduk, multiQC
rRNA read filtering RNA-seq reads none removal of rRNA-derived reads via mapping to rRNA reference Bowtie 2
Key results
  • Redesign of cross-species genome alignment generation reduced required computational time compared to prior methodology while keeping sensitivity comparable
  • lncEvo successfully executed the full workflow (assembly, lncRNA prediction, conservation analysis) across five species using public RNA-seq datasets
Key statistics
  • other >70 percent (lincRNAs (cited, Hezroni et al.) lacking detectable homologs in species diverged >50 million years ago)
  • count 5 species (number of species used to test lncEvo (human, mouse, chimp, horse, dog))
  • other 1 TPM (minimum expression threshold in at least one sample required to retain a candidate lncRNA)
  • count 2500 transcripts per batch (batching size for slncky ortholog search on reference lncRNA sets)
  • other 128 cores (m5.2xlarge: 8 vCPU/32GiB RAM; m5.8xlarge: 32 vCPU/128GiB RAM) (AWS Batch compute environments used for performance testing)
  • other ≥500 GB EBS volume, ulimit -n 20,000 (recommended AWS AMI configuration for running lncEvo)

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.

lncEvo is a bioinformatics software paper describing a computational pipeline; it contains no formal inferential hypothesis testing. Statistical content is limited to exploratory data analysis (EDA) with visual summary statistics generated automatically via SweetViz, and threshold-based filtering rules (e.g., transcript length ≥200 nt, expression ≥1 TPM in at least one sample, ORF length <100 aa). Conservation between species is characterised by exonic sequence identity scores computed by slncky, and top-ranked orthologs are selected by this metric. No p-values, confidence intervals, or formal tests are reported.

Replicationunclear Sample sizeThree RNA-Seq runs per species were used for pipeline testing (SRA accessions listed per species); no power calculation or sample-size justification is provided GroupsFive species (human, mouse, chimp, horse, dog) used for pipeline demonstration; pairwise conservation analyses performed with human as reference Pairingna Randomization/blindingnot stated Dispersionnone
Approaches that could also have been used
  • Expression quantification is performed with Salmon, and a threshold of ≥1 TPM in at least one sample is used to retain transcripts
    Could also: Kallisto or RSEM could also be used for expression quantification; the threshold could alternatively be framed as a minimum mean or median TPM across replicates, or as a counts-per-million cutoff applied after edgeR/DESeq2 normalisation — Alternative quantification tools and threshold strategies allow sensitivity analysis of how the filtering criterion affects the retained lncRNA set, and replicate-aware thresholds reduce the risk of retaining transcripts detected in only one noisy sample
  • Conservation between species is ranked and selected by exonic sequence identity (a single continuous metric computed by slncky)
    Could also: Additional conservation metrics such as synteny block score, percent coverage of aligned exons, or phyloP/phastCons base-level constraint scores could also be incorporated as complementary ranking criteria — Multiple orthogonal conservation metrics can distinguish lncRNAs that share positional conservation from those with true sequence similarity, reflecting the heterogeneous conservation modes described in the paper's own background section
  • Exploratory data analysis is summarised with automated HTML reports (SweetViz), presenting distributions and summary statistics visually
    Could also: Explicit reporting of central tendency and dispersion (e.g., median ± IQR or mean ± SD) for key output variables such as transcript length, exonic identity, and TPM could also be included in the text — Numeric summaries complement visual reports and allow readers to make direct comparisons with other studies without reproducing the figures
  • Pipeline performance is demonstrated on three RNA-Seq runs per species without a formal benchmark against alternative pipelines using shared metrics
    Could also: A benchmarking design using simulated RNA-Seq data with known ground-truth lncRNA sets (e.g., from LNCipedia or GENCODE) and standardised metrics (sensitivity, precision, F1) could also be applied — Ground-truth benchmarks allow quantitative comparison of lncRNA identification performance across pipelines and make sensitivity/specificity trade-offs explicit
  • Filtering thresholds (≥200 nt length, ≥1 TPM, ORF <100 aa) are applied as fixed, literature-derived cutoffs
    Could also: Sensitivity analyses varying each threshold independently could also be reported to show how the size and composition of the retained lncRNA set changes — Threshold sensitivity analyses help users understand how conservative or permissive the default settings are and guide parameter tuning for their own datasets
  • The comparison between lncRNA sets from two species is described qualitatively and via EDA visualisations
    Could also: Overlap statistics such as Jaccard similarity or hypergeometric enrichment tests could also be used to quantify and assess the significance of the overlap between species-specific and conserved lncRNA sets — Formal overlap statistics provide a concise, interpretable summary of cross-species concordance and allow assessment of whether observed overlaps exceed chance expectation
Software: Nextflow · Python/pandas · SweetViz (patched) · Salmon · StringTie · STAR · slncky · LAST (lastal/last-train) · TransDecoder · CPC2 (Coding Potential Calculator) · FastQC/MultiQC · fastp · Bowtie 2

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-33563213 (lncEvo)

Paper: Bryzghalov O, Makałowska I, Szcześniak MW. lncEvo: automated identification and conservation study of long noncoding RNAs. BMC Bioinformatics 2021;22:59. PMID 33563213 · PMCID PMC7871587 · DOI 10.1186/s12859-021-03991-2.

What lncEvo is. A Nextflow (+Docker) pipeline with three modules: (1) transcriptome assembly from RNA-Seq, (2) lncRNA prediction, (3) conservation study (genome-wide comparison of lncRNA transcriptomes between two species).

Code (corrected). The brief's code_url (github.com/alexdobin/STAR) is a screening artifact — STAR is only ONE internal tool. The actual pipeline repo is

Internal tools / exact params (from main.nf, environment.yml): QC FastQC 0.11.8 + fastp 0.20.0 (default) + multiQC 1.8 · rRNA removal Bowtie2 2.3.5 (--fast -X 1000 --un-conc) · align STAR 2.7.3a (full param string in main.nf star_align, twopass Basic, --outFilterMultimapNmax 20, etc.) · assembly StringTie 2.0 (--rf) + StringTie --merge (-T 0 -F 0) · Cuffcompare (cufflinks 2.2.1, removes class codes c,e,p,s via filter_class_codes.py) · ORF TransDecoder 5.5.0 (-m 100 -S) · coding potential CPC2 2.0b · expression Salmon 0.14.2 (-l A --validateMappings, threshold ≥1 TPM in ≥1 sample) · conservation LAST 1061 + axtChain (UCSC 366) + slncky 1.0. Reference: Ensembl (repo pinned to release 103).

In scope (pipeline-derived, attempted)

# Reported result Paper loc Pipeline Approach
C1 lncRNA counts per species (dog ~5000, mouse ~25000, others intermediate) Fig 6 assembly+prediction module (a) audit deposited *_lncrnas_data.txt; (b) re-run human assembly on «our HPC» and recount
C2 Positionally conserved lncRNAs, human↔mouse: 550 (human query) / 569 (mouse query) Results text conservation module audit deposit (mouse absent → uncheckable); re-run not attempted (needs 2 full assemblies + LAST)
C3 Per-species conservation hit counts (human→chimp/horse/dog) Fig 7 / deposit conservation module audit deposited *.orthologs.top.final.txt
C4 STAR alignment of the paper's own run SRR4421334 runs with lncEvo params Methods align step actual «our HPC» compute → uniquely-mapped % (data-usability point)
C5 Compute cost (Table 1: human 134.1 CPU‑h, 6h14m, 39.07 GB) Table 1 whole assembly hardware-dependent; record our wall/CPU as non-1:1 context only

Out of scope / not attempted

  • Wet-lab / manual interpretation — none (fully computational paper).
  • Full 5-species re-run end-to-end (mouse, chimp, horse, dog assemblies + all pairwise conservation): enormous compute, low marginal evidential value beyond the human re-run + deposit audit. Not attempted; stated honestly.
  • slncky/LAST conservation re-run: depends on two completed assemblies; deferred.

Datasets relied upon

RNA-seq runs (SRA), 3–4 per species, human as conservation reference: human SRR4421334/SRR4421792/SRR4421350; mouse SRR7771840/42/43/46; chimp SRR1602576/SRR1758919/SRR1758927; horse SRR9133801/SRR10140550/SRR10205788; dog SRR10915304/05/07. Plus the deposited result files (S3) which act as the authors' ground-truth outputs for the audit.

Figures / tables: Fig 6Fig 7Table
C4
Reported
STAR alignment of paper's own run SRR4421334 with lncEvo params (no per-run rate reported)
Reproduced
91.65% uniquely mapped (24800721/27058906 reads), ~95% overall; bit-for-bit identical to an independent earlier run (deterministic)
within tolerance
C1d
Reported
human lncRNA count 10716 transcripts (authors' deposit)
Reproduced
11645 transcripts -- FULL lncEvo assembly module re-run from scratch on the paper's 3 human runs (genuine «our HPC» compute, ~26 CPU-h); +8.7% vs deposit, explained by Ensembl reference version drift
within tolerance
C1a
Reported
dog lncRNAs ~5000 (Fig6)
Reproduced
5268 transcripts (deposit re-audit)
within tolerance
C1c
Reported
human/chimp/horse intermediate (Fig6)
Reproduced
human 10716, chimp 11893, horse 7065 (all between dog 5268 and mouse ~25000)
partial
C1b
Reported
mouse lncRNAs ~25000 (Fig6)
Reproduced
mouse ABSENT from public deposit - uncheckable
partial
C2a
Reported
human->mouse positionally conserved 550
Reproduced
mouse absent from deposit - not verifiable
partial
C2b
Reported
mouse->human positionally conserved 569
Reproduced
mouse absent from deposit - not verifiable
partial
C3a
Reported
human->chimp conserved, majority high identity (Fig7)
Reproduced
5808 rows / 5025 genes (deposit), highest of 3 pairs
partial
C3b
Reported
human->horse conserved ~50% (Fig7)
Reproduced
1202 rows / 1000 genes (deposit)
partial
C3c
Reported
human->dog conserved ~50% (Fig7)
Reproduced
853 rows / 716 genes (deposit)
partial

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 61/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟡1. Data identity
🟡2. Endpoint comparability
🟡3. Location of the main deviation
🔴4. Cause of the deviation
🟡5. Derivability / plausibility
🟢6. Severity of the deviation
🟡7. Core claim
🟡8. Severity of the miss (overall human judgment)
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: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +6

Where data exists, lncEvo reproduces well: the STAR alignment of the paper's own SRR4421334 reproduces cleanly (91.65% uniquely mapped), the dog lncRNA count (5268 tx ≈ Fig6 '~5000') is within tolerance, and the human/chimp/horse counts and human→chimp/horse/dog conservation structure are consistent with the deposited authors' files. The decisive limitation is data availability on the authors' side: the deposit ships no mouse files, so the headline human-mouse reversal numbers (550/569) and the ~25000 mouse count in Fig6 are not derivable and cannot be verified. This is a completeness/derivability gap, not a demonstrated computational error or fabrication—hence partial rather than red on the core verdict, with q5 yellow and q4 red (authors'/deposit incompleteness).

🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at support@doesitreproduce.com.

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

519.5 k
tokens (I/O) · 58.7 M incl. cache
268 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.