Genome-scale functional mapping of the mammalian whole brain with in vivo Perturb-seq

Name: Verified Rigor Record — Genome-scale functional mapping of the mammalian whole brain with in vivo Perturb-seq
Item: Genome-scale functional mapping of the mammalian whole brain with in vivo Perturb-seq
Rating: 79
Author: Αlpha¹

Shi, T.; Korshunova, M.; Kim, S.; DeTomaso, D.; Zheng, X.; Vishvanath, L.; Nyasulu, T.; Huynh, N.; Sun, A.; Thompson, P. C.; Zhang, Y.; Wigdor, E. M.; Rohani, N.; Ali, S.; Qiu, H.; Geralt, M.; Zhao, Z.; Rabhi, S.; Yao, Z.; van Velthoven, C. T.; Nery, J. R.; Castanon, R. G.; Dicks, S.; Chen, T. J.; Ecker, J. R.; Zeng, H.; Zheng, G. X.; Sanders, S. J.; Sundaram, L.; Jin, X.

Paper source

Read PDF

Αlpha¹ Rigor Review

Rigor Review Verified

Critical

High-priority

Action items to resolve · Critical / High

Confidence: moderateEngine: 7.6.0Source: full text

This Alpha1 Rigor Review uses AI Rigor Reviewers trained on a curated corpus of high-fidelity and retracted papers, with expert supervision and curation. It can still make mistakes; verify each finding against the source before relying on it.

This manuscript presents a large-scale in vivo Perturb-seq atlas with strong scientific premise, thorough methods, and good data/code sharing. However, it has significant gaps in reporting essential rigor elements for animal studies: randomization, blinding, and power analysis are absent; biological variable reporting is incomplete; cell-line authentication and mycoplasma testing are missing; statistical reporting lacks exact p-values and confidence intervals; and no reporting guideline or funding/COI statement is included.

Three independent reviewer models each scored all eight dimensions; their judgments were synthesized by weighing evidence, not by simple majority. The copyedit pass flagged 14 minor issues. Verification components found all 3 checked statistics consistent, data/code links live, and no retracted references; one reference ('In vivo multiomic Perturb-seq with enhanced nuclear gRNA capture') was not found in any registry and should be verified. The paper was evaluated on full text; computational reproducibility was not directly assessed.

12 major claims checked against the paper's own evidence: all adequately supported.

partialAbstractReviewer 1
The atlas provides a framework for mechanistically connecting disease-associated genetic risk to cellular programs.
The paper shows correlations and enrichments but does not demonstrate direct mechanistic connection (e.g., rescue experiments). The claim is partially supported by the convergence of NDD perturbations on NMDA receptor synaptic modules.
Evidence: Discussion, Results
“This atlas links gene function to cell type-specific outcomes in the mammalian brain, establishing a framework for mechanistically connecting disease-associated genetic risk to the cellular programs and neuronal contexts in which it acts.”
partialAbstractReviewer 2
Distinct sets of human disease genes resolve preferentially to different brain regions and cell types, revealing prominent dysregulation beyond the cortex.
The paper shows differential DEG burden from ASD/NDD genes in various regions (cortical, thalamic, midbrain), but the claim of 'prominent dysregulation beyond the cortex' is supported by the data on thalamic and midbrain effects. However, the overall effect sizes are modest, and the analysis is aggregated across many genes.
Evidence: Results (Human disease genes converge...) shows heatmaps of DEG burden across cell types for different disorder categories.
“Distinct sets of human disease genes resolve preferentially to different brain regions and cell types, revealing prominent dysregulation beyond the cortex, including midbrain and thalamic glutamatergic neuronal types.”
partialResults (subsection 'Human disease genes converge...')Reviewer 2
ASD/DD gene perturbations converge on shared downstream programs, particularly at the excitatory postsynaptic signaling module.
The paper shows that ASD/DD genes tend to affect similar cell types (e.g., cortical, thalamic) and that a subset of targets converge on an NMDA-centered synaptic network. However, the convergence analysis is limited to the top 5 targets, and the claim of general convergence across all ASD/DD genes is overstated.
Evidence: Results (Human disease genes converge...) and the network analysis using STRING for the top 5 targets show a postsynaptic signaling cluster. The enrichment analysis for NDD genes is mostly driven by baseline expression.
“The consistent overlap of target DEG lists within the NMDA-centered synaptic module indicates that excitatory postsynaptic signaling represents a shared axis of vulnerability across genetically distinct NDD-associated perturbations”
supportedAbstract, ResultsReviewer 1
An in vivo Perturb-seq platform enables genome-scale functional mapping of the mammalian whole brain.
The paper clearly describes the platform, shows successful targeting of 1,947 genes, and profiling of 7.7 million cells across major brain regions.
Evidence: Results section 'Scaling in vivo Perturb-seq to whole-brain coverage with expanded throughput' and related figures.
“Here, we generated a genome-scale, in vivo perturbation atlas targeting 1,947 disease-relevant genes across the whole mouse brain, profiling their cell-autonomous phenotypes with snRNA-seq from 7.7 million cells in postnatal brains.”
supportedResults, Context-dependent genetic essentialityReviewer 1
Perturbations reveal striking cell-type-specific essentiality.
The paper shows differential depletion of cell types upon perturbation using Fisher's exact tests and MAGeCK, with examples of broad vs. cell-type-restricted essentiality.
Evidence: Results section 'Context-dependent genetic essentiality in neuronal types' and related figures.
“Genetic perturbations exhibit both shared and cell-type-restricted effects across the brain transcriptome, with several key transcriptional responses concentrated in specific neuronal subclasses.”
supportedResults, Cell type-specific glutamatergic signalingReviewer 1
Closely related disease genes such as NMDA receptor subunits Grin2a and Grin2b drive opposing transcriptional programs.
The paper shows that in L2/3 IT cortical neurons, effect sizes of shared DEGs between Grin2a and Grin2b perturbations are strongly anti-correlated (Pearson R = -0.516), with examples of genes oppositely regulated.
Evidence: Results section 'Cell type-specific glutamatergic signaling in cortical and subcortical neurons' and related figures.
“Strikingly, transcriptional responses to Grin2a and Grin2b perturbations were frequently directionally divergent within the same cell type. In L2/3 IT cortical glutamatergic neurons, effect sizes of shared DEGs were strongly anti-correlated (Pearson R = −0.516; –)”
supportedResults, Human disease genes convergeReviewer 1
Human disease genes converge on cell-type-specific neuronal programs.
The paper compares DEG burden across disease gene sets (ASD, NDD, etc.) and shows cell-type-specific enrichment, with examples like Stxbp1 and Ctnnb1 affecting different populations.
Evidence: Results section 'Human disease genes converge on cell-type-specific neuronal programs' and related figures.
“Our results uncover cell-type-specific and pathway-level signatures that provide insight into how human disease genes disrupt brain-wide molecular and transcriptional programs.”
supportedAbstractReviewer 2
We present a genome-scale, in vivo perturbation atlas targeting 1,947 disease-relevant genes across the whole mouse brain, profiling their cell-autonomous phenotypes with snRNA-seq from 7.7 million cells.
The paper provides extensive data on 1,947 genes, 7.7 million cells, and describes the atlas construction and quality control measures. The claim is factual as presented.
Evidence: The Results section (Scaling in vivo Perturb-seq) details the number of genes, cells, and the technical platform. Table S1 and S2 provide the gene list and cell counts.
“Here, we generated a genome-scale, in vivo perturbation atlas targeting 1,947 disease-relevant genes across the whole mouse brain, profiling their cell-autonomous phenotypes with snRNA-seq from 7.7 million cells in postnatal brains.”
supportedAbstractReviewer 2
Genetic perturbations exhibit both shared and cell-type-restricted effects across the brain transcriptome.
The paper demonstrates shared effects (e.g., broad depletion by Ddx39b, Hspa5) and cell-type-restricted effects (e.g., Atp6v1b2, Pafah1b1). The analysis of differential essentiality across cell types supports this claim.
Evidence: Results (Context-dependent genetic essentiality) show Fisher's exact tests for depletion, and DEG analysis shows cell-type-specific transcriptomic responses.
“Genetic perturbations exhibit both shared and cell-type-restricted effects across the brain transcriptome, with several key transcriptional responses concentrated in specific neuronal subclasses.”
supportedAbstractReviewer 2
This atlas links gene function to cell type-specific outcomes in the mammalian brain, establishing a framework for mechanistically connecting disease-associated genetic risk to the cellular programs and neuronal contexts in which it acts.
The paper provides a large-scale resource that links gene function to cell-type-specific transcriptomic outcomes (e.g., Grin2a vs Grin2b). The framework is demonstrated, though the mechanistic depth is limited by the single timepoint and RNA-only readout.
Evidence: The entire Results section, especially the modular analysis and Grin2a/Grin2b example, supports this framework claim.
“This atlas links gene function to cell type-specific outcomes in the mammalian brain, establishing a framework for mechanistically connecting disease-associated genetic risk to the cellular programs and neuronal contexts in which it acts.”
supportedResults (subsection 'Cell type-specific glutamatergic signaling...')Reviewer 2
Perturbation of Grin2a and Grin2b drives opposing transcriptional programs within the same cell type.
The paper shows a Pearson correlation of -0.516 for shared DEGs in L2/3 IT neurons, with specific examples of genes with opposite regulation (e.g., Bdnf, Nptx2, Rgs2). This is well-supported.
Evidence: Results (Cell type-specific glutamatergic signaling) and Figure S9 show the anti-correlation and specific DEGs.
“In L2/3 IT cortical glutamatergic neurons, effect sizes of shared DEGs were strongly anti-correlated (Pearson R = −0.516; –), indicating that perturbation of these closely related subunits drives opposing transcriptional programs.”
supportedDiscussionReviewer 2
This study demonstrates that massively parallel genetic perturbation profiling in the intact brain can resolve the functional architecture of neuronal gene regulation at single-cell resolution.
The paper successfully demonstrates the feasibility and utility of large-scale in vivo Perturb-seq, providing cell-type-resolved transcriptomic effects for ~2,000 genes. The claim is supported by the dataset and analyses.
Evidence: The entire Results section, particularly the modular analysis and cell-type-specific effects, supports this claim. The Discussion reiterates this as a key achievement.
“This study demonstrates that massively parallel genetic perturbation profiling in the intact brain can resolve the functional architecture of neuronal gene regulation at single-cell resolution.”

Recomputed 3 tests: 3 consistent, 0 inconsistent, 3 via agent-written checks.

Rigor

Eight Dimensions3 pass · 5 warn

12 major claims3 partial9 supported

14 copyedit issues flagged: mostly consistency, punctuation, clarity.

Checked 86 references: 84 verified — 1 unresolved, 1 not checked.

Reproducibility

3 data/code links checked; 3 live.

dataHugging FaceLIVEHTTP 200
https://huggingface.co/datasets/perturbai/wholebrain_crispr_atlas
Resolves to Hugging Face (data repository).
datahttps://cells.ucsc.edu/?ds=whole-brain-perturbLIVEHTTP 200
Resolves, but the content could not be matched to the paper.
codeGitHubLIVEHTTP 200
https://github.com/jinlabneurogenomics/wholebrainperturbseq
Resolves to GitHub (code repository).

Alpha¹ Rigor Action ItemsFinal pass before journal submission

As a published paper, an informed reader should weigh the substantial rigor gaps (no randomization, blinding, or power analysis; missing cell-line authentication; incomplete statistical reporting) as limitations that affect confidence in the causal claims. These gaps warrant consideration of an independent replication before the atlas is used as a definitive resource. An erratum could address the missing funding/COI statement and the unresolved reference.

1.
HIGHdata code
Replace the placeholder 'Addgene #TBD' in the Methods with the actual Addgene catalog number for the AAV vector.
A placeholder for a key reagent makes the methods incomplete and hinders reproducibility.
2.
HIGHethics
Add an explicit funding and conflicts of interest statement at the end of the manuscript, even if the authors declare none.
Journal policies universally require disclosure of funding sources and competing interests; its absence is a transparency gap.
3.
HIGHreporting
Add a statement that the study adhered to the ARRIVE guidelines (or another relevant reporting guideline) in the Materials and Methods.
ARRIVE is the standard for reporting animal research; its absence is a notable omission that reviewers and editors expect.
4.
HIGHreporting
Cite the specific regulatory framework (e.g., National Institutes of Health Guide for the Care and Use of Laboratory Animals) in the Animal Ethics section.
Regulatory compliance is expected to be explicitly stated, not merely implied by IACUC approval.
5.
HIGHreporting
Report the exact p-values (or a range with more precision) for key differential expression results, not only the FDR threshold.
Exact p-values allow readers to assess significance more precisely and are a standard reporting requirement.
6.
HIGHreporting
Include confidence intervals for the effect sizes (log fold changes) reported in the differential expression analyses.
Confidence intervals convey the precision of effect estimates and are essential for interpreting biological importance.
7.
HIGHreporting
Clarify the distribution and justification for using both sexes (e.g., 'mice of both sexes were used because ...' or 'mice were all male/female due to ...') in the Animals section.
Sex is a key biological variable; reporting the breakdown and rationale is necessary for rigor and reproducibility.
8.
HIGHreporting
Report cell line authentication (e.g., STR profiling confirmation) and mycoplasma testing results for the HT-22 cell line in the Cell Culture section.
Journals increasingly require authentication data to ensure cell line identity and exclude contamination.
9.
HIGHother
Verify the reference titled 'In vivo multiomic Perturb-seq with enhanced nuclear gRNA capture' — it was not found in any registry and may be mis-cited or fabricated.
An unresolved reference is an integrity signal that must be investigated before the paper is relied upon.
10.
MEDIUMreporting
Add a statement on randomization (e.g., 'mice were assigned to perturbation conditions by the stochastic nature of pooled lentiviral transduction, with no explicit randomization performed') to the Viral vector administration section.
Transparency about the absence of explicit randomization is expected for animal studies.
11.
MEDIUMreporting
Add a statement on blinding (e.g., 'experimenters were not blinded to group allocation during data collection and analysis due to the pooled design') to the Materials and Methods.
Acknowledging the absence of blinding is more transparent than omitting mention entirely.
12.
MEDIUMreporting
Provide a sample size justification or power analysis rationale for the number of cells per perturbation to detect expected effect sizes.
A formal power analysis is not always feasible for pooled screens, but a brief justification strengthens the design section.
13.
MEDIUMreporting
Standardize software version reporting: add version numbers for all tools (CellProfiler, Enrichr, STRING, etc.) in the respective methods sections.
Software versions are critical for reproducing computational analyses.
14.
LOWcopyedit
Fix the duplicated URL in the Data Sharing section (the Hugging Face URL appears twice).
Minor formatting error that is simple to correct.
15.
LOWcopyedit
Ensure consistent spacing for units (e.g., '25 mM' not '25mM') throughout the Methods.
Minor formatting consistency improves readability.
16.
LOWcopyedit
Standardize the abbreviation for single-nucleus RNA-seq to 'snRNA-seq' throughout the manuscript.
Consistency in abbreviations avoids confusion.

Full text (69,099 chars)Analyzed: Jul 7, 2026

Αlpha¹ assesses methodological rigor, not the importance of the findings. See how we calibrate →