Quantitative characterization of tissue states using multiomics and ecological spatial analysis

Not checked yet

• Native figure graphics and journal layout, Supplementary Figures 1–21, Reporting Summary and Peer Review File remain unread; caption reading is separate.
• This is computational reanalysis of previously generated tonsil, mouse-spleen, colorectal-cancer and liver datasets, with no new biological experiments. Cells, spatial patches and subsamples are dependent; exact cross-modal patient overlap and original study ethics were not independently audited.
• The liver comparison uses one healthy tissue and one HCC tissue, each divided into ten regions. Reported tests and confidence intervals over regions do not establish between-patient disease effects; region overlap is unestablished.
• Colorectal-cancer analysis reuses 35 tissues: 17 CLR and 18 DII. Fivefold internal Cox validation gives C-indices 0.734±0.106, 0.717±0.088 and combined 0.763±0.085. The ± quantities are not labelled confidence intervals; external clinical validation, a locked threshold and out-of-fold plotted survival groups are not established.
• The mouse main figure reports three healthy and six diseased tissues; Extended Data Figure 3 instead reports 36 diseased. Introduction reference 33 says human intestine, but its citation and data availability identify colorectal cancer; intestine atlas reference 9 is separate. These discrepancies remain unresolved.
• MaxFuse transfers inferred RNA profiles rather than directly measuring each imaged cell transcriptome. Spatial co-occurrence and ligand–receptor coexpression do not demonstrate active signalling, cell contact or causation. Of 62 original ligand–receptor predictions, 44 persist and 18 do not; 68 new predictions lack direct experimental validation.
• Several whole-tissue comparisons are null despite hotspot differences. Finer computational partitions are not adjudicated biological accuracy. CellCharter null outputs were omitted from one displayed comparison and low-score matches filtered; preserve these alongside the no-original-data-exclusion statement.
• Nolan is credited with conceptualization, original drafting, editing and supervision, jointly supervising with Tibshirani. Ding, Tang and Zhu developed, implemented and analysed the algorithms. Authorship does not establish personal execution of every method.
• Nolan is explicitly named for NIH 3U54HG010426 and FDA Medical Countermeasures Initiative contracts HHSF223201610018C and 75F40120C00176, Parker Institute support, Cancer Research UK C27165/A29073 and the Harris professorship. These are published support acknowledgements; original award scope, classified work or a UAP government approach was not verified.
• Shalek is named for NIH 1P01AI177687-01 and Gates Foundation INV-055706; Ding for Stanford fellowships. These awards are not automatically Nolan funding. Article views expressly do not represent funding institutions.
• Nolan declares unrelated Pfizer, Vaxart, Celgene and Juno grants, Ionpath cofounder status and Akoya cofounder, patent US9909167 and advisory roles. Article-time declarations are not current registry audits, proof of bias or secret programmes.
• MESA/MaxFuse code, Zenodo 10.5281/zenodo.14504057, raw datasets and cited original bodies were not retrieved or executed in this lane. Earlier independently accepted source scopes elsewhere remain intact; 87 reference entries do not establish 87 original-body reviews.
• Full publisher correction/version history, the illustrated article and whole Nolan/project corpus remain incomplete. Source/platform/receipt dates are separate.