Pith. sign in

REVIEW 4 major objections 5 minor 49 references

Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that automated agentic analysis of transcriptomic signatures across 1,384 disease–condition pairs recovers known disease relationships and uncovers previously undocumented molecular links, with pathway-level convergence sug

desk verdict A large-scale, useful resource undermined by a mathematically impossible pathway filter and circular pathway logic—needs major repair before the discovery claims can be trusted. read the letter →

arxiv 2508.04742 v1 pith:UBHQXKHS submitted 2025-08-06 q-bio.GN cs.LG

Ke Chen , Haohan Wang This is my paper
classification q-bio.GNcs.LG
keywords diseasesimilaritynetworktranscriptomicsignatureagenticAIpathwayenrichmenttaxonomycomorbiditydrugrepurposingGenoTEXbenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper aims to establish that disease relationships can be discovered at scale from transcriptomic signatures alone, without clinical labels. It applies the agentic AI system GenoMAS to 1,384 disease–condition pairs from GenoTEX, derives significant gene lists for each pair, and builds gene- and pathway-level similarity networks. When compared with ICD-10-CM categories, the networks recover established comorbidities and reveal previously unreported cross-category links, interpreted through shared pathways. The larger point is that automated transcriptomic analysis can scale to thousands of cohorts and yield mechanistic, actionable disease hypotheses.

What carries the argument

The machinery is a two-level transcriptomic similarity pipeline. First, GenoMAS produces a Lasso-based significant gene list ($p<0.05$) for each of the 1,384 disease–condition pairs; gene-level edges are called by a bidirectional hypergeometric test with Benjamini–Hochberg FDR control. Second, for the 1,293 significant combinations, shared genes are mapped to pathways through g:Profiler using six databases (GO:BP, Reactome, KEGG, transcription-factor targets, miRNA targets, and Human Phenotype Ontology), and a pathway is called shared only when the same gene is enriched in both pairs. The strength of functional convergence is summed as $\sum_{i=1}^{n}[\log(1-p_{1i})+\log(1-p_{2i})]$ over sha

What would settle it

Recompute the top 1,293 gene-level edges from the same cohorts using a conventional differential-expression pipeline with uniform per-cohort normalization; if a substantial fraction of the highest-weight edges, especially the novel cross-category links, falls below significance, the network's hidden-relationship claims would be falsified. A narrower check: for the ASD–osteoporosis and ASD–T1D edges, test whether the shared genes are differentially expressed in ASD brain tissue rather than only in the peripheral cohorts used in GenoTEX.

Watch

Extended reading notes

Core claim

The central claim is that the transcriptomic signatures produced by a fully automated agentic AI system carry enough consistent biological signal to reconstruct a useful disease-relationship map. Gene-level overlap tests across roughly a million pairwise combinations yield about 65,000 significant links, reduced to 1,293 after removal of redundant generalized condition entries; pathway-level enrichment with six databases supports 1,060 of these combinations. The resulting network clusters broadly by ICD-10-CM chapter, yet contains validated cross-category edges such as ankylosing spondylitis–osteoporosis and previously unrecorded similarities such as Alzheimer's disease–ocular melanoma. The

Load-bearing premise

The load-bearing premise is that the automatically generated Lasso-selected gene lists faithfully represent each disease–condition pair's transcriptomic state across heterogeneous cohorts; if those lists are noisy or systematically biased, every gene overlap, pathway score, and network edge built on them is uninterpretable.

Editorial extensions

If this is right

  • The 1,384-condition network recovers ICD-10-CM categories without using clinical labels, so transcriptomic signatures can serve as an independent check on disease taxonomy.
  • Condition-specific edges show that background states such as obesity and hypertension alter molecular similarity: 11 of the top 20 links involve hypertension, framing hypertension as a systemic amplifier of disease vulnerability.
  • Pathway-supported edges (1,060 of 1,293) provide functional hypotheses for comorbidities, including a lipid-inflammation-bone axis linking ankylosing spondylitis and osteoporosis.
  • Rare-disease neighborhoods—ASD is the example—connect molecularly to better-characterized conditions, yielding concrete repurposing candidates such as lipid-modulating agents or metabolic and immune modulators.
  • The complete network is released publicly, allowing researchers to generate new disease-pair hypotheses directly.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • We infer that the pathway-level 'same gene in same pathway in both pairs' rule is conservative, so the 1,060 supported combinations likely undercount functional convergence; a gene-set-level enrichment overlap test would probably surface more edges.
  • We infer that the novel cross-category edges, e.g., Alzheimer's disease–ocular melanoma, are testable hypotheses: one could check whether their shared metabolic-axis genes are co-regulated in relevant tissues or arise from the shared neural-crest origin the paper suggests.
  • We infer that the drug-repurposing suggestions are not direct evidence; a shared pathway only constrains the mechanism, so prospective studies or model systems would be needed before statins or PPAR agonists are considered for ASD.
  • We infer that reproducibility across cohorts and tissues is the main unresolved risk: re-analysis with a different signature method could remove many high-weight edges, which would require revising the hidden-relationship claims.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes an agentic-AI pipeline (GenoMAS) over the GenoTEX benchmark to obtain transcriptomic signatures for 1,384 disease-condition pairs, computes statistically significant gene overlaps, and builds a gene-level disease similarity network. It then adds a pathway-level layer: for each gene-sharing disease combination, it identifies pathways enriched for shared genes and computes a cumulative 'similarity score', filtering to 1,060 pathway-supported combinations. The paper claims this network recovers known comorbidities, finds undocumented cross-category links, shows obesity/hypertension modulate similarity, and suggests ASD drug-repurposing candidates. All results are claimed to be publicly available on GitHub.

Significance. If the findings were fully supported, the scale and the explicit public release would make this a useful hypothesis-generation resource. The attempt to validate against ICD-10 categories and the use of six pathway databases are strengths. However, as written the pathway-level central claim is unsupported by the stated mathematics: the similarity score cannot be positive under the printed formula, and the pathway-sharing definition is circular with respect to the gene overlap. The gene-level network might survive a re-analysis, but the mechanistic 'functional convergence' layer and all interpretations built on it require substantial rework before the paper's main claims can be evaluated.

major comments (4)
  1. [Section 3.3, similarity-score equation] As printed, Similarity Score = sum_i [log(1-p1_i) + log(1-p2_i)]. For any p-value in (0,1), log(1-p) is negative, so every term and every sum is <= 0. The text states the score 'becomes more positive when both p1 and p2 are small', but log(1-p) approaches 0 from below as p -> 0; it never becomes positive. The Filtering step then retains only combinations where 'at least one shared pathway had a positive similarity score'; under the stated formula this set is empty. Yet Section 2.2 reports 1,060 pathway-supported combinations and Figure 3 plus all downstream ASD/hypertension analyses are built on them. This is an internal contradiction, not a minor typo. If the code implements a different quantity (e.g., -log p), the manuscript must state this and the results must be recomputed with the correct formula. As written, all pathway-level results are uninterpretable.
  2. [Section 3.3, 'Identification of Shared Pathways'] A pathway is declared shared only if the same gene is significantly enriched for the same pathway in both pairs, and the analysis 'focused only on the genes that were shared between the two pairs'. Therefore the pathway-level overlap is a deterministic function of the gene-level overlap. It cannot serve as the independent 'functional convergence' evidence claimed in Sections 2.2 and 2.4, nor can it validate gene-level links. For example, Table 2's shared pathways are selected because shared genes happen to be enriched for those pathways; this recapitulates the gene overlap by construction. The authors should either (i) perform pathway enrichment on each disease-condition pair's full significant gene list and test overlap of enriched pathways against a permutation/random-set null, or (ii) substantially weaken the claim that pathway analysis provides independent mechanistic support.
  3. [Sections 3.1-3.2] All downstream analyses inherit the quality of the GenoMAS Lasso p<0.05 gene lists. The paper provides no evidence that these lists are valid, comparable, or reproducible across the 911 cohorts, which differ in tissue, platform, and normalization. If these gene lists are noisy or systematically biased, every hypergeometric overlap, FDR value, network edge, and pathway score built on them is uninterpretable. Please benchmark a random subset of GenoMAS outputs against a standard differential-expression pipeline (e.g., limma/voom or DESeq2 on the same normalized data) or validate a handful of predicted links on independent disease-vs-control datasets. This is a load-bearing validation step, not an optional robustness check.
  4. [Sections 2.2-2.4, Figure 5] The claims of previously undocumented cross-category links, obesity/hypertension modulation of similarity, and ASD drug repurposing are all derived from the 1,060 pathway-supported combinations and Figure 3. Because those combinations are selected by a formula that cannot yield positive scores as printed, and because the pathway-sharing definition is circular with the gene overlap, these specific findings must be regenerated after correcting the methodology. In particular, the ASD subgraph (Figure 5) and the hypertension examples (Section 2.3) should be recomputed with an independent pathway-overlap null, and the number of significant links should be recalculated. Without this re-analysis, the 'therapeutic repurposing' and 'hidden disease relationships' claims are not supported.
minor comments (5)
  1. [Section 3.3] The similarity-score equation is rendered unclearly; define p1_i and p2_i explicitly and number the equation. The phrasing 'joint probability that both enrichments are non-random' is statistically incorrect: (1-p1)(1-p2) is not a probability of non-random enrichment.
  2. [Section 3.2] The edge weight is described as 'log10(FDR)', but a meaningful positive similarity weight would be -log10(FDR) or similar. Please clarify the sign convention and its relationship to edge thickness.
  3. [Throughout] Several typos and rendering artifacts should be corrected: 'biomarker discover' in the Introduction, 'pairwaise' in Section 3.2, 'theagenetic AI' in the Figure 1 caption, and the garbled inline symbols in Sections 1-2. The reference list for GenoMAS [26], GenoTEX [27], and g:Profiler [48] is also missing from the provided text.
  4. [Section 2.1 / 3.2] The GPT-4o assignment of ICD-10-CM categories is used both for validation and for node coloring, but no validation of the assignment is provided. Please release the prompt, the assignments, and an estimate of agreement with human expert coding, since misclassification could bias the heatmap and the network interpretation.
  5. [Data/code availability] The GitHub link is stated, but the manuscript does not provide a versioned commit, the exact code for the similarity formula, or the intermediate gene lists and pathway tables. To make the results checkable, please supply these artifacts or a clear pointer to them.

Circularity Check

2 steps flagged · score 6.0 of 10

Pathway-level results reduce to gene-overlap by construction, and the stated similarity-score formula makes the 'positive similarity score' filter impossible; the paper's pathway-based claims are not supported as independent.

  1. self definitional [Section 3.3, 'Identification of Shared Pathways'; used in Sections 2.2-2.4]
    "For each disease–condition pairwise combination (hereafter, “combination”), we focused only on the genes that were shared between the two pairs. For each such shared gene, we retrieved its pathway annotations in both pairs. A pathway was considered “shared” if the same pathway was significantly enriched for the same gene in both pairs."

    The pathway-level edge set is constructed only from genes already shared between two disease–condition pairs, and only from the 1,293 combinations that survived the gene-level FDR filter (Section 3.2). Hence any 'shared pathway' implies a shared gene, so the pathway network is necessarily a subgraph of the gene-overlap network. The paper presents pathway-level findings as functional evidence beyond gene overlap, but those findings are determined by the same gene overlap and add no independent confirmatory information. The pathway analysis is a re-annotation of the gene-level result, not an independent derivation.

  2. other [Section 3.3, 'Similarity Scoring' and 'Filtering']
    "Similarity Score = Σ_{i=1}^{n} [log(1−p1_i) + log(1−p2_i)] ... the score becomes more positive when both p1 and p2 are small ... We retained only combinations where at least one shared pathway had a positive similarity score, indicating non-random co-enrichment. This yielded 1,060 pathway-supported combinations out of the original 1,293 gene-sharing ones."

    For any p_i in (0,1), log(1−p_i) is always negative, so every term and every finite sum in the formula is negative. The Filtering criterion requiring 'at least one shared pathway had a positive similarity score' can never be satisfied under the stated formula, yet the paper reports 1,060 pathway-supported combinations and builds Figure 3 and all downstream interpretations on them. This is an internal contradiction: either the formula is misprinted and the code implements a different quantity (e.g., −log p), or the reported pathway-level results are not derivable from the stated method. Either way, the pathway layer cannot independently confirm the gene-level discoveries, and the drug-repurposing/mechanism claims based on it are unsupported.

full rationale

The derivation chain is: GenoTEX inputs → GenoMAS gene lists → shared-gene overlap → hypergeometric FDR → 1,293 links → pathway annotations of shared genes → 'shared pathway' → similarity score → filter → 1,060 links → network → biological claims. The gene-level portion is self-contained: it computes overlap statistics from transcriptomic inputs and does not rely on its own conclusions. However, the pathway-level portion is not independent: 'shared pathway' is defined through shared genes, so the pathway network is a constructed subset of the gene-overlap network. Moreover, the stated score formula cannot produce positive values, making the filter that yields the 1,060 combinations vacuous. These are concrete, quotable reductions, not mere external-validation concerns. The self-citations to GenoTEX/GenoMAS are not themselves circular here because the paper applies a prior pipeline rather than invoking a uniqueness theorem; lack of external validation of GenoMAS outputs is a correctness risk, not circularity. Overall, one central layer of the paper reduces by construction and is internally inconsistent, but the gene-level network retains independent content; hence score 6.

Assumptions & free parameters 5 free parameters · 7 assumptions · 0 invented entities

The central claim rests on the validity of the GenoTEX cohort definitions and the GenoMAS agentic AI gene lists, plus several hand-chosen thresholds. The pathway layer is not independent because shared pathways are defined from shared genes. No external validation or permutation baseline is provided.

free parameters (5)
  • gene significance threshold = p < 0.05
    Genes from Lasso regression with p < 0.05 are retained as the disease signature; this cutoff is chosen by hand and directly determines all overlap counts (Section 3.2).
  • FDR threshold for gene overlap = 0.05
    Only pairwise gene overlaps with FDR < 0.05 are retained; the number of edges (about 65,000 before filtering) depends on this cutoff (Section 3.2).
  • redundancy removal rule = remove 'all' entries
    Links involving unconditioned disease entries labeled 'all' are removed to prevent redundancy; this rule is a modeling choice and reduces the graph from about 65,000 to 1,293 edges (Section 3.2).
  • pathway similarity filter = at least one shared pathway with 'positive' score
    Combinations are kept only if a shared pathway has a positive similarity score; as written the formula yields only negative values for p in (0,1), making the filter ill-defined (Section 3.3).
  • GPT-4o ICD assignment = prompt not provided
    Each disease is assigned an ICD-10-CM category by GPT-4o; the prompt and validation are unspecified, yet category labels are used to color the network, define within and cross-category comparisons, and interpret clustering (Section 3.2).
assumptions (7)
  • domain assumption GenoTEX cohorts and normalized gene features are biologically valid and comparable across cohorts.
    Stated in Section 3.1; the entire pipeline relies on this without external validation.
  • domain assumption GenoMAS agentic AI correctly performs preprocessing, normalization, and Lasso inference for all 1,384 pairs.
    Invoked in Section 3.1; no benchmark or manual QC is reported for the outputs used here.
  • ad hoc to paper Lasso regression with p < 0.05 identifies the relevant disease-associated gene set.
    No justification or comparison to other differential expression methods is given (Section 3.2).
  • standard math Hypergeometric overlap test with an 18,000-gene universe and BH correction is the correct null model for gene overlap.
    Section 3.2 uses it; standard test, but the null model assumes all genes exchangeable and ignores gene set size and pathway structure.
  • ad hoc to paper A pathway is shared iff the same gene is enriched in both disease lists; this captures pathway-level convergence.
    Section 3.3 defines shared pathways this way; it couples pathway evidence to gene overlap, creating circularity.
  • domain assumption g:Profiler's higher-level pathway selection across six databases is an unbiased functional representation.
    Section 3.3; results depend on database choice and term abstraction.
  • domain assumption ICD-10-CM categories assigned by GPT-4o are accurate.
    Section 3.2 uses them for validation and coloring; no accuracy check is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI." pith.science (2026). https://pith.science/paper/UBHQXKHS

@misc{pith2026250804742,
  author       = {Pith},
  title        = {Pith review of: Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBHQXKHS}},
  note         = {Machine review of arXiv:2508.04742}
}
read the original abstract

Modern disease classification often overlooks molecular commonalities hidden beneath divergent clinical presentations. This study introduces a transcriptomics-driven framework for discovering disease relationships by analyzing over 1300 disease-condition pairs using GenoMAS, a fully automated agentic AI system. Beyond identifying robust gene-level overlaps, we develop a novel pathway-based similarity framework that integrates multi-database enrichment analysis to quantify functional convergence across diseases. The resulting disease similarity network reveals both known comorbidities and previously undocumented cross-category links. By examining shared biological pathways, we explore potential molecular mechanisms underlying these connections-offering functional hypotheses that go beyond symptom-based taxonomies. We further show how background conditions such as obesity and hypertension modulate transcriptomic similarity, and identify therapeutic repurposing opportunities for rare diseases like autism spectrum disorder based on their molecular proximity to better-characterized conditions. In addition, this work demonstrates how biologically grounded agentic AI can scale transcriptomic analysis while enabling mechanistic interpretation across complex disease landscapes. All results are publicly accessible at github.com/KeeeeChen/Pathway_Similarity_Network.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 47 canonical work pages

  1. [1]

    for Medicare & Medicaid Services (U.S.) and N

    C. for Medicare & Medicaid Services (U.S.) and N. C. for Health Statistics (U.S.), ICD-10-CM Official Guidelines for Coding and Reporting FY 2024 -- UPDATED October 1, 2023 (October 1, 2023 - September 30, 2024) , tech. rep., Centers for Medicare & Medicaid Services (CMS) (July 2023), Published July 24, 2023

    work page 2024

  2. [2]

    A. S. Fauci, E. Braunwald, D. L. Kasper et al. , Harrison's Principles of Internal Medicine (McGraw-Hill, 2008)

    work page 2008

  3. [3]

    K. I. Goh, M. E. Cusick, D. Valle et al. , The human disease network, Proceedings of the National Academy of Sciences 104 , 8685 (2007)

    work page 2007

  4. [4]

    L. P. Santamar \'i a, E. P. G. del Valle, M. Zanin et al. , Classifying diseases by using biological features to identify potential nosological models, Scientific Reports 11 , p. 21096 (2021)

    work page 2021

  5. [5]

    Zanin, J

    M. Zanin, J. M. Tu \ n as and E. Menasalvas, Understanding diseases as increased heterogeneity: A complex network computational framework, Journal of The Royal Society Interface 15 , p. 20180405 (2018)

    work page 2018

  6. [6]

    N. M. Ferraro et al. , Transcriptomic signatures across human tissues identify functional rare genetic variation, Science 369 , p. eaaz5900 (2020)

    work page 2020

  7. [7]

    F. Q. Wang et al. , Unraveling transcriptomic signatures and dysregulated pathways in systemic lupus erythematosus across disease states, Arthritis Research & Therapy 26 , p. 99 (2024)

    work page 2024

  8. [8]

    Shao et al

    L. Shao et al. , T-cell transcriptomic signatures in adults with primary untreated immune thrombocytopenia segregate by age, revealing distinct druggable pathways, Blood 144 , p. 124 (2024)

    work page 2024

Show all 49 references
  1. [9]

    Poon et al

    Z. Poon et al. , Transcriptomic signature and functional abnormalities of bone marrow mesenchymal stromal cells mediate disease progression of myelodysplastic/myeloproliferative neoplasms, Blood 142 , p. 5618 (2023)

    2023

  2. [10]

    Wang et al

    Q. Wang et al. , Deep learning-based brain transcriptomic signatures associated with the neuropathological and clinical severity of alzheimer’s disease, Brain Communications 4 , p. fcab293 (2022)

    2022

  3. [11]

    S. C. Mendelsohn et al. , Transcriptomic signatures of progression to tuberculosis disease among close contacts in brazil, The Journal of Infectious Diseases , p. jiae237 (2024)

    2024

  4. [12]

    Chaudhuri et al

    S. Chaudhuri et al. , Cell-specific transcriptional signatures of vascular cells in alzheimer’s disease: Perspectives, pathways, and therapeutic directions, Molecular Neurodegeneration 20 , p. 12 (2025)

    2025

  5. [13]

    T. M. Pelaia, M. Shojaei and A. S. McLean, The role of transcriptomics in redefining critical illness, in Annual Update in Intensive Care and Emergency Medicine 2023\/ , 2023 pp. 3--14

    2023

  6. [14]

    D. B. Antcliffe et al. , Transcriptomic signatures in sepsis and a differential response to steroids: From the vanish randomized trial, American Journal of Respiratory and Critical Care Medicine 199 , 980 (2019)

    2019

  7. [15]

    Bigot et al

    J. Bigot et al. , Transcriptomic signature of the cd, American Journal of Transplantation 16 , 3430 (2016)

    2016

  8. [16]

    R. Q. Figueiredo et al. , Towards a global investigation of transcriptomic signatures through co-expression networks and pathway knowledge for the identification of disease mechanisms, Nucleic Acids Research 49 , 7939 (2021)

    2021

  9. [17]

    Jha et al

    A. Jha et al. , Identifying common transcriptome signatures of cancer by interpreting deep learning models, Genome Biology 23 , p. 117 (2022)

    2022

  10. [18]

    J. A. Harrill et al. , Signature analysis of high-throughput transcriptomics screening data for mechanistic inference and chemical grouping, Toxicological Sciences 202 , 103 (2024)

    2024

  11. [19]

    Kim et al

    M. Kim et al. , Refining diagnosis of renal cell carcinoma subtypes through single-cell resolution transcriptomic signatures, Cancer Research 84 , 3500 (2024)

    2024

  12. [20]

    Namba, M

    S. Namba, M. Iwata and Y. Yamanishi, From drug repositioning to target repositioning: Prediction of therapeutic targets using genetically perturbed transcriptomic signatures, Bioinformatics 38 , i68 (2022)

    2022

  13. [21]

    Zhai et al

    Z. Zhai et al. , Disignatlas: An atlas of human and mouse disease signatures based on bulk and single-cell transcriptomics, Nucleic Acids Research 52 , D1236 (2024)

    2024

  14. [22]

    Lessard et al

    S. Lessard et al. , Leveraging large-scale multi-omics evidences to identify therapeutic targets from genome-wide association studies, BMC Genomics 25 , p. 1111 (2024)

    2024

  15. [23]

    Wang and C

    F. Wang and C. A. Barrero, Multi-omics analysis identified drug repurposing targets for chronic obstructive pulmonary disease, International Journal of Molecular Sciences 25 , p. 11106 (2024)

    2024

  16. [24]

    rep., Anaconda, Inc

    Anaconda, The State of Data Science 2020: Moving from Hype Toward Maturity , tech. rep., Anaconda, Inc. (2020)

    2020

  17. [25]

    BPC, Navigating the Intersection of Biostatistics, Bioinformatics, and Machine Learning , tech

    R. BPC, Navigating the Intersection of Biostatistics, Bioinformatics, and Machine Learning , tech. rep. (2023), Publication details not fully specified

    2023

  18. [26]

    H. Liu, Y. Li and H. Wang, Genomas: A multi-agent framework for scientific discovery via code-driven gene expression analysis, arXiv preprint arXiv:2507.21035 (2025)

    2025 arXiv

  19. [27]

    H. Liu, S. Chen, Y. Zhang et al. , Genotex: An llm agent benchmark for automated gene expression data analysis, arXiv preprint arXiv:2406.15341 (2024)

    2024 arXiv

  20. [28]

    Sharif, A

    K. Sharif, A. M. Tsur, N. Ben-Shabat et al. , The risk of osteoporosis in patients with ankylosing spondylitis—a large retrospective matched cohort study, Medicina Cl \'i nica 160 , 373 (2023)

    2023

  21. [29]

    J. Mei, H. Hu, H. Ding et al. , Investigating the causal relationship between ankylosing spondylitis and osteoporosis in the european population: A bidirectional mendelian randomization study, Frontiers in Immunology 14 , p. 1163258 (2023)

    2023

  22. [30]

    Zhang, J

    D. Zhang, J. Liu, B. Gao et al. , Immune mechanism of low bone mineral density caused by ankylosing spondylitis based on bioinformatics and machine learning, Frontiers in Genetics 13 , p. 1054035 (2022)

    2022

  23. [31]

    J. K. Olynyk and G. A. Ramm, Risk of liver cancer in hfe-hemochromatosis (2021)

    2021

  24. [32]

    Jayachandran, R

    A. Jayachandran, R. Shrestha, K. R. Bridle et al. , Association between hereditary hemochromatosis and hepatocellular carcinoma: A comprehensive review, Hepatoma Research 6 , N/A (2020)

    2020

  25. [33]

    Tian, Y.-Z

    M. Tian, Y.-Z. Cui, G.-H. Song et al. , Proteomic analysis identifies mmp-9, dj-1 and a1bg as overexpressed proteins in pancreatic juice from pancreatic ductal adenocarcinoma patients, BMC Cancer 8 , p. 241 (2008)

    2008

  26. [34]

    Piyaphanee, Q

    N. Piyaphanee, Q. Ma, O. Kremen et al. , Discovery and initial validation of 1-b glycoprotein fragmentation as a differential urinary biomarker in pediatric steroid-resistant nephrotic syndrome, Proteomics Clinical Applications (2011)

    2011

  27. [35]

    Fujii, S

    C. Fujii, S. Harumiya, Y. Sato, M. Kawakubo, H. Matoba and J. Nakayama, 1,4-linked n-acetylglucosamine suppresses gastric cancer development by inhibiting mucin-1-mediated signaling, Cancer Science 113 , 3852 (2022)

    2022

  28. [36]

    C. Sun, C. Cao, T. Zhao et al. , A2m inhibits inflammatory mediators of chondrocytes by blocking il‐1 /nf‐ b pathway, Journal of Orthopaedic Research 41 , 241 (2023)

    2023

  29. [37]

    Okuno, M

    E. Okuno, M. Tsujimoto, M. Nakamura and R. Kido, 2-aminoadipate-2-oxoglutarate aminotransferase isoenzymes in human liver: A plausible physiological role in lysine and tryptophan metabolism, Enzyme and Protein 47 , 136 (1993)

    1993

  30. [38]

    Hallen, J

    A. Hallen, J. F. Jamie and A. J. L. Cooper, Lysine metabolism in mammalian brain: An update on the importance of recent discoveries, Amino Acids 45 , 1249 (2013)

    2013

  31. [39]

    M. M. Essa, S. Subash, N. Braidy et al. , Role of nad ^ + , oxidative stress, and tryptophan metabolism in autism spectrum disorders, International Journal of Tryptophan Research 6 , p. IJTR.S11355 (2013)

    2013

  32. [40]

    L. Yang, Z. Chu, M. Liu et al. , Amino acid metabolism in immune cells: Essential regulators of the effector functions, and promising opportunities to enhance cancer immunotherapy, Journal of Hematology & Oncology 16 , p. 59 (2023)

    2023

  33. [41]

    Michaudel, C

    C. Michaudel, C. Danne, A. Agus et al. , Rewiring the altered tryptophan metabolism as a novel therapeutic strategy in inflammatory bowel diseases, Gut 72 , 1296 (2023)

    2023

  34. [42]

    E. O. Olufunmilayo, M. B. Gerke-Duncan and R. M. D. Holsinger, Oxidative stress and antioxidants in neurodegenerative disorders, Antioxidants 12 , p. 517 (2023)

    2023

  35. [43]

    Lumaquin-Yin, E

    D. Lumaquin-Yin, E. Montal, E. Johns, A. Baggiolini, T. H. Huang, Y. Ma, C. LaPlante, S. Suresh, L. Studer and R. M. White, Lipid droplets are a metabolic vulnerability in melanoma, Nature Communications 14 , p. 3192 (2023)

    2023

  36. [44]

    Castro-P \'e rez, M

    E. Castro-P \'e rez, M. Singh, S. Sadangi et al. , Connecting the dots: Melanoma cell of origin, tumor cell plasticity, trans-differentiation, and drug resistance, Pigment Cell & Melanoma Research 36 , 330 (2023)

    2023

  37. [45]

    Gaudet, N

    P. Gaudet, N. S kunca, J. C. Hu et al. , Primer on the gene ontology, Methods in Molecular Biology 1446 , 25 (2017)

    2017

  38. [46]

    Kanehisa, The kegg database, in In Silico Simulation of Biological Processes: Novartis Foundation Symposium 247\/ , (John Wiley & Sons, Ltd, Chichester, UK, 2002)

    M. Kanehisa, The kegg database, in In Silico Simulation of Biological Processes: Novartis Foundation Symposium 247\/ , (John Wiley & Sons, Ltd, Chichester, UK, 2002)

    2002

  39. [47]

    K \"o hler, L

    S. K \"o hler, L. Carmody, N. Vasilevsky et al. , Expansion of the human phenotype ontology (hpo) knowledge base and resources, Nucleic Acids Research 47 , D1018 (2019)

    2019

  40. [48]

    Raudvere, L

    U. Raudvere, L. Kolberg, I. Kuzmin et al. , g:profiler: A web server for functional enrichment analysis and conversions of gene lists (2019 update), Nucleic Acids Research 47 , W191 (2019)

    2019

  41. [49]

    write newline

    " write newline "" before.all 'output.state := FUNCTION blank.sep after.quote 'output.state := FUNCTION fin.entry output.state after.quoted.block = 'skip 'add.period if write newline FUNCTION new.block output.state before.all = 'skip output.state after.quote = after.quoted.blo...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.