LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration
Citation integrity
Citation Integrity Audit Report
LATTICE: Graph Self-Supervised Learning for Multimodal Spatial Omics Integration
Checklist Categories in Play
The manuscript contains:
- Reference resolvability triggers: Multiple in-text citations to prior work (refs [1]–[20])
- Claim–citation support triggers: Specific factual claims attributed to references (e.g., "GraphST [3] and related models use graph neural networks")
- Self-citation triggers: References [4] (SARSIM) and implicit internal tools (ReCAST) authored by the same group
- Quotation/number fidelity triggers: None detected (no direct quotations from sources)
Findings by Category
1. Reference Resolvability
Status: PRESENT with UNVERIFIABLE elements
All 20 references are listed in the reference section with author names and publication venues. However, several entries lack complete identifiers:
| Ref | Title/Venue | DOI/PMID | Status | Severity |
|---|---|---|---|---|
| [1] | Ståhl et al., Science 2016 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [2] | Kaya-Okur et al., Nat. Commun. 2019 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [3] | Long et al., Nat. Commun. 2023 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [4] | Dwarampudi et al., bioRxiv 2026 | Not provided | UNVERIFIABLE – future date, preprint | HARD |
| [5] | Dong & Zhang, Nat. Commun. 2022 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [6] | Hu et al., Nat. Methods 2021 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [7] | Yang et al., Nat. Commun. 2025 | Not provided | UNVERIFIABLE – future date | HARD |
| [8] | Zhang et al., Cell Genomics 2026 | Not provided | UNVERIFIABLE – future date | HARD |
| [9] | Zhu & Ma, Nat. Biotechnol. 2024 | Not provided | UNVERIFIABLE – future date | HARD |
| [10]–[13] | Software/methods papers | arXiv IDs or URLs provided for [11], [13] | Resolvable | SOFT |
| [14], [15], [17] | 10x Genomics URLs | URLs provided; accessed 2026-05-06 | UNVERIFIABLE – future access date | HARD |
| [16] | Strehl & Ghosh, JMLR 2002 | Not provided | Resolvable (established venue, specific year) | SOFT |
| [18] | Rousseeuw, J. Comput. Appl. Math. 1987 | Not provided | Resolvable (classic reference, specific year) | SOFT |
| [19] | Traag et al., Sci. Rep. 2019 | Not provided | Resolvable (high-impact journal, specific year) | SOFT |
| [20] | McInnes et al., arXiv 2018 | arXiv ID provided | Resolvable | SOFT |
Key Issues:
-
References [4], [7], [8], [9] are dated 2025–2026, which is in the future relative to a 2024 preprint. This is a red flag for either:
- Manuscript date error (should be 2026 or later)
- Reference date error (should be 2024 or earlier)
- Speculative/in-preparation citations
-
References [14], [15], [17] cite 10x Genomics URLs with an access date of 2026-05-06, which is in the future. This is internally inconsistent with a preprint dated 2024.
-
Reference [4] (SARSIM) is a bioRxiv preprint by the same authors. It is load-bearing (cited in Section 1, Figure 1 caption, Section 2, Section 4.1, and Appendix F). The future date makes it unverifiable.
2. Claim–Citation Support
Status: MOSTLY PRESENT with CRITICAL UNVERIFIABLE CLAIM
Load-bearing claims checked:
Claim 1: "GraphST [3] and related models use graph neural networks and self-supervision on expression and spatial proximity."
- Citation: [3] Long et al., Nat. Commun. 2023
- Verification: Title states "Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with graphst" — plausibly supports claim of graph-based spatial method
- Status: PLAUSIBLE (not directly verified but title is consistent)
- Severity: SOFT
Claim 2: "SARSIM [4] is a framework for spatially anchored regulatory inference that integrates paired Visium and single-cell multiome data, learns a soft cell-to-spot mapping, and projects accessibility and motif activity into tissue space."
- Citation: [4] Dwarampudi et al., bioRxiv 2026
- Verification: Cannot verify — reference is dated 2026 (future date) and is a preprint by the same authors
- Status: UNVERIFIABLE
- Severity: HARD (this is a central upstream dependency for the entire LATTICE pipeline)
Claim 3: "Spatial transcriptomics [1] enable genome-wide gene expression across sections"
- Citation: [1] Ståhl et al., Science 2016
- Verification: Title "Visualization and analysis of gene expression in tissue sections by spatial transcriptomics" directly supports this claim
- Status: PLAUSIBLE
- Severity: SOFT
Claim 4: "Spatial CUT&Tag [2] add section-level chromatin and histone-modification context"
- Citation: [2] Kaya-Okur et al., Nat. Commun. 2019, "CUT&tag for efficient epigenomic profiling of small samples and single cells"
- Verification: Title supports claim about epigenomic profiling; plausibly includes histone modifications
- Status: PLAUSIBLE
- Severity: SOFT
Claim 5: "Noise-contrastive estimation [12]"
- Citation: [12] Gutmann & Hyvärinen, AISTATS 2010
- Verification: Title "Noise-contrastive estimation: A new estimation principle for unnormalized statistical models" directly supports the method name
- Status: CONFIRMED
- Severity: SOFT
3. Self-Citation and Citation Inflation
Status: PRESENT
-
Reference [4] (SARSIM): Authored by Dwarampudi et al., same first author as the LATTICE manuscript. Cited 6+ times as a load-bearing upstream dependency. This is appropriate (not inflated) because SARSIM is genuinely a prerequisite for the multimodal feature construction (Section 4.1, Appendix F).
-
ReCAST pipeline: Described in Appendix F as "an internal engineering pipeline" by the same authors. Not formally cited as a reference but mentioned throughout. This is appropriate transparency (not self-citation inflation) because it is disclosed as internal work.
-
No evidence of non-germane self-padding detected.
Severity: SOFT (appropriate self-citation for genuine dependencies)
4. Retracted or Predatory Sources
Status: NONE DETECTED
All cited venues are established:
- Science, Nature Communications, Nature Methods, Nature Biotechnology (high-impact)
- Cell Genomics, Scientific Reports (peer-reviewed)
- JMLR, AISTATS (established conferences)
- arXiv, bioRxiv (preprint servers, not predatory)
No retracted papers identified.
5. Quotation and Number Fidelity
Status: NOT APPLICABLE
The manuscript does not include direct quotations from cited sources or specific numerical claims attributed to references (e.g., "Smith et al. found X% improvement"). Claims are paraphrased rather than quoted.
Summary of Critical Issues
| Issue | Ref(s) | Severity | Status | Action Required |
|---|---|---|---|---|
| Future publication dates (2025–2026) on preprint | [4], [7], [8], [9] | HARD | UNVERIFIABLE | Authors must clarify: are these in-preparation works, or is the manuscript date incorrect? |
| Future access dates on URLs | [14], [15], [17] | HARD | UNVERIFIABLE | Authors must verify and correct access dates. |
| SARSIM (ref [4]) is load-bearing but unverifiable | [4] | HARD | UNVERIFIABLE | Central to multimodal feature construction (Section 4.1, Figure 1). Cannot verify claims about SARSIM's functionality without access to the actual preprint. |
| No DOIs provided for any reference | All | SOFT | PRESENT | Recommended: add DOIs or PubMed IDs where available for reproducibility. |
Questions for Authors
-
Reference [4] (SARSIM): The citation is dated bioRxiv 2026, which is in the future. Is this:
- A manuscript date error (should the LATTICE preprint be dated 2026)?
- A reference date error (should SARSIM be dated 2024)?
- An in-preparation work that should be marked as such?
-
References [7], [8], [9]: These are dated 2025–2026. Are these published, in press, or in preparation? If in preparation, please mark them explicitly.
-
References [14], [15], [17]: The access date is listed as 2026-05-06, which is in the future. Please provide the actual access date when the URLs were retrieved.
-
Verification of SARSIM functionality: The manuscript claims SARSIM "learns a soft cell-to-spot mapping, and projects accessibility and motif activity into tissue space." Can you provide a preprint link or DOI so reviewers can verify these claims?
Conclusion
Overall Status: The manuscript's reference list is structurally complete but contains critical date inconsistencies that make several load-bearing references unverifiable. The most serious issue is Reference [4] (SARSIM), which is central to the entire LATTICE pipeline but cannot be verified due to its future publication date.
Recommendation: Authors must resolve the date inconsistencies and provide verifiable access to SARSIM before the manuscript can be fully audited for citation integrity.