BGPT: Paper Review: FixNCut: single-cell genomics through reversible tissue fixation and dissociation

Explore by Goal

Quick Explanation Copied

FixNCut in one view

FixNCut uses reversible DSP (Lomant’s reagent) to fix tissue before dissociation, then proceeds with standard scRNA-seq workflows, aiming to preserve RNA integrity, library complexity, and cellular composition while reducing dissociation/sampling artifacts across PBMCs and solid tissues (mouse lung/colon; human IBD colon biopsies), with protocol deposition and raw data in GEO/Zenodo. Evidence: mapping/exonic fractions, pseudo-bulk correlations (e.g., PBMC R²≈0.99), and signature-score reductions for stress/dissociation artifacts are reported across conditions.

Long Explanation

FixNCut: single-cell genomics through reversible tissue fixation and dissociation

Critical, evidence-based visual paper review (last-mile skeptical assessment)

DOI: 10.1186/s13059-024-03219-5 • Received/Accepted: Received 12 Apr 2023; Accepted 18 Mar 2024 • Data: GEO GSE229944, Zenodo, Protocols.io, GitHub.

What FixNCut claims (and what to verify)

Key idea: apply DSP (Lomant’s reagent) fixation before mechanical/enzymatic dissociation to decouple sampling logistics from downstream processing while preserving RNA and composition.
Measured readouts: mapping/exonic fractions, gene/UMI vs sequencing depth (library complexity), pseudo-bulk correlations (mean expression concordance), compositional changes (scCODA), and reductions in published dissociation/sampling stress signatures.
Compatibility: scRNA-seq 3′ workflows, antibody/LMO labeling for hashing/FACS-style workflows, and spatial proteomics imaging workflow compatibility are reported as feasibility demonstrations.

Visual evidence dashboard (from paper-reported metrics)

Note: This plot mixes different magnitudes (genes vs cells). It is used only to convey scale, not direct performance comparison. All values originate from paper-reported counts/summary metrics.

Concordance checks (pseudo-bulk / mean expression)

These are paper-reported summary concordance metrics and do not by themselves guarantee that all biologically relevant signals are preserved (e.g., cell-type composition can shift even if mean concordance is high).

Library complexity preservation (genes/UMIs vs sequencing depth)

For strict quantification (exact slopes/CI), you should compute directly from GEO counts. FixNCut reports that gene/UMI complexity vs sequencing depth remains comparable for PBMC and mouse lung, while mouse colon shows improved mapping/exonic fraction and complexity-related metrics after DSP fixation.

Mechanistic rationale (DSP reversibility) — what is plausible

Why a reversible crosslinker might help

The paper describes DSP as a homo-bifunctional NHS-ester crosslinker with amine-reactive groups and a central disulfide that enables reversibility via reducing agents present in typical reverse transcription buffers.
DSP is presented as membrane-permeable and used previously to preserve single cells in suspension for RNA sequencing, but the paper states its use in tissues prior to dissociation is novel for their context.

Artifact reduction: stress signatures as external validity tests

This figure intentionally avoids invented magnitudes and only encodes directionality explicitly stated in the provided paper text (e.g., warm-collagenase signature lower in fixed vs fresh for mouse lung; cryo-only stress higher vs fixed/cryo; fixed vs fresh showing no significant differences in human colon signatures).

Critical appraisal (skeptical review)

Strengths (what supports the claims)

Multi-tissue validation: PBMCs (suspension) and solid tissues (mouse lung/colon; human IBD colon) are included, and FixNCut reports both short-term (fresh vs fixed) and long-term (fixed/cryo) comparisons.
Multiple orthogonal metrics: mapping/exonic fractions, library complexity vs depth, pseudo-bulk correlation, cell-composition modeling (scCODA), and dissociation/sampling signature scores are used together—this reduces the chance that a single proxy metric drives conclusions.
Data/code availability: FASTQs and count matrices are submitted to GEO under GSE229944, processed objects at Zenodo, DSP protocol at Protocols.io, and code on GitHub. This directly supports independent reanalysis and falsification attempts.

Blind spots / uncertainties (what could still be misleading)

Cell-type proportion shifts: The paper reports that proportions can change in some tissues (e.g., specific immune subsets in PBMCs; differential representation of epithelial/stromal and TA-like populations in mouse colon; immune composition shifts in lung). Even with high pseudo-bulk concordance, such shifts can still affect biological interpretation of differential abundance.
Stress signature incompleteness: Stress artifacts are approximated via published gene signatures and signature scores, which can miss unmodeled artifacts and can be sensitive to how integration/clustering is done. This is not a failure of the method per se, but a limitation of using signatures as the detection mechanism.
Cryopreservation interactions: Fixed+cryo is tested, and cryo-only shows higher stress signatures, but the paper also reports reduced UMIs in fixed/cryo vs cryo-only (while genes can remain similar). This could reflect differences in RNA capture sensitivity or cell quality effects rather than only transcriptomic preservation.
Labeling readouts may not be transcriptome-neutral: Flow cytometry/LMO experiments indicate that fixation/cryostorage can alter MFI and membrane integrity markers (e.g., increased DAPI/Annexin V/PI after cold storage). This suggests FixNCut may be compatible with antibody/LMO labeling, but quantitation of surface proteins (or viability-dependent gating) may require optimization per antigen panel and storage duration.

Reproducibility notes (what you can already test)

Because raw FASTQs/count matrices and analysis code are available, an immediate reproducibility test is to recompute: (i) gene/UMI vs sequencing depth curves; (ii) pseudo-bulk correlation; (iii) scCODA composition estimates; (iv) UCell signature score changes; and (v) DEGs under the paper’s thresholds (e.g., Log2FC>|1|, FDR<0.05, present in ≥10% of cells).

Practical takeaways (for experimental design)

If your study requires decentralized sampling or delayed processing, FixNCut’s conceptual value is that fixation occurs before dissociation, aiming to reduce time-of-handling transcriptomic artifacts.
However, treat cell proportion and surface label quantitation as potentially protocol-dependent (fixation and cold storage can alter MFI/viability markers), and validate your panel and gating strategy.

What would most likely disprove the core advantage?

If reanalysis on held-out tissues/disease contexts shows that fixation fails to preserve (i) RNA integrity proxies, (ii) library complexity relationships, or (iii) reductions in dissociation/sampling artifact signatures, compared with fresh processing.
If compositional shifts (cell-state or abundance) remain uncontrolled and correlate with fixation in a way that cannot be attributed to known biology or integration artifacts.

Note on citation style: Every paper-specific factual claim above is tied to the FixNCut publication text and/or its specified methods/results; general scRNA-seq context claims are intentionally avoided unless explicitly grounded in citations.

Author reviews (click for targeted critiques)

If any author name differs in BGPT’s author-review index, click the closest match available.

Feedback:

Updated: March 25, 2026