Variance-based Prioritization Reveals a Clinically Validated Antigen Discovery Space Systematically Inaccessible to Mean-Based Methods

Variance-based Prioritization Reveals a Clinically Validated Antigen Discovery Space Systematically Inaccessible to Mean-Based Methods | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Variance-based Prioritization Reveals a Clinically Validated Antigen Discovery Space Systematically Inaccessible to Mean-Based Methods XIAOQI HU This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9215598/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background. Mean-based transcriptomic prioritization (differential expression analysis, DEG) dominates cancer target discovery but is optimized for driver gene identification rather than therapeutic antigen discovery. Whether variance-based prioritization captures a complementary and clinically relevant discovery space has not been systematically evaluated. Methods. We applied variance-based prioritization (TANK) and four comparator methods (DEG, MAD, coefficient of variation, mean expression) to genome-wide transcriptomic data from TCGA gastric adenocarcinoma (n = 443 tumor samples). We evaluated recall of a gold standard set of 28 clinically validated therapeutic antigens (FDA-approved and Phase 2 + ADC/CAR-T/TCR-T targets) at three ranking thresholds. Mechanistic specificity was assessed by comparing surface protein enrichment, driver oncogene enrichment, and therapeutic antigen recall between high-variance and low-variance gene sets. Results. Across all thresholds, TANK substantially outperformed all comparator methods in therapeutic antigen recall (top 5%: TANK 25%, DEG 3.6%, MAD 3.6%, CV 0%, Mean 3.6%). In the primary analysis using the full gene universe (60,654 genes), TANK recovered 50% of gold standard targets at top 5% versus 3.6% for DEG (OR = 27.0, p = 0.000071). High-variance genes were not globally enriched for surface proteins (9.1% vs 9.2%, OR = 0.98, p = 0.58), ruling out surface protein abundance as an explanatory factor. Instead, high variance specifically depleted canonical driver oncogenes (OR = 0.48, p = 0.0004) while achieving extreme enrichment of therapeutic antigens over low-variance genes (OR = infinity, p = 0.002). Three targets nominated prospectively by TANK prior to literature review subsequently converged on FDA-approved or Phase 2 + clinical programs. Conclusions. Transcriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to driver gene biology and systematically inaccessible to all mean-based approaches tested. Integration of variance-based prioritization into target discovery workflows may substantially expand the accessible space for immunotherapy antigen development. Cancer Biology therapeutic antigen discovery transcriptomic variance antibody-drug conjugates differential expression analysis cancer immunotherapy target prioritization Figures Figure 1 Introduction Cancer target discovery has been dominated by differential expression analysis (DEG), a framework that identifies genes whose mean expression levels differ significantly between tumor and normal tissue [ 1 , 2 ]. This approach has proven highly effective for identifying driver oncogenes and tumor suppressor genes, yielding landmark therapeutic targets including ERBB2, EGFR, and MYC [ 3 , 4 ]. DEG-based prioritization underlies the majority of target identification pipelines in both academic and industrial drug discovery [ 5 ]. However, DEG is optimized for a specific biological task: identifying genes with elevated average expression in tumor cells. Therapeutic antigens for antibody-drug conjugates (ADCs), CAR-T cells, TCR-T cells, and bispecific antibodies require a fundamentally different expression profile: high expression in a targetable patient subpopulation with low expression in normal tissues [ 6 , 7 ]. A gene with high inter-patient variance but moderate mean expression would be systematically deprioritized by DEG despite potentially representing an ideal immunotherapy target. Many successful therapeutic antigens are characterized by stochastic reactivation of silenced developmental gene programs — cancer-testis antigens (MAGEA4, CLDN6), oncofetal proteins (GPC3, FOLR1), and lineage-specific markers (SLC34A2, MSLN) — producing high inter-patient expression variance [ 8 , 9 ]. We previously developed TANK (Transcriptomic ANtigen variance-based prioritization), a framework ranking genes by inter-patient transcriptomic variance, and demonstrated three prospective convergences on clinically validated targets [ 10 – 12 ]. However, a systematic comparison of TANK against multiple transcriptomic prioritization methods has not been performed. Here we provide the first such comparison against a curated gold standard of 28 therapeutic antigens, and characterize the mechanistic basis for the distinct discovery space captured by variance. Methods Transcriptomic Data RNA-sequencing count data for TCGA-STAD (gastric adenocarcinoma) were obtained from the GDC Data Portal (n = 443 tumor samples, 45 matched normal samples; STAR aligner, GENCODE v36). Gene identifiers (ENSEMBL GRCh38, version suffixes removed) were mapped to HGNC symbols using a curated translation table. Transcriptomic Prioritization Methods Five prioritization methods were evaluated on a common gene universe of 7,052 genes with available rankings across all methods: (1) TANK: inter-patient variance across 443 tumor samples, descending rank; (2) DEG: absolute log2 fold change from DESeq2 (v1.36), tumor vs normal, descending rank; (3) MAD: median absolute deviation across tumor samples, descending rank; (4) CV: coefficient of variation (SD/mean) across tumor samples, descending rank; (5) Mean: mean expression across tumor samples, descending rank. For the primary TANK vs DEG recall analysis (Table 1 ), the full TANK ranking (60,654 genes) was used to avoid artificially constraining the variance-based candidate space; for multi-method comparison (Fig. 1 ), the common 7,052-gene universe was used to ensure fair comparison. Gold Standard Therapeutic Antigen Set A gold standard of 28 clinically validated therapeutic antigens was curated from FDA approval records, ClinicalTrials.gov interventional studies, and peer-reviewed publications (accessed March 2026). Inclusion criteria: (1) direct target of an ADC, CAR-T, TCR-T, or bispecific antibody; (2) FDA approval or Phase 2 + status; (3) cell-surface or secreted protein. Tier 1 (FDA-approved, n = 10): ERBB2, EGFR, PDCD1, CD274, CTLA4, MSLN, NECTIN4, TACSTD2, CD19, TNFRSF17. Tier 2 (Phase 2+, n = 18): CLDN18, CLDN6, MAGEA4, DLL3, GPC3, FOLR1, FOLH1, CD276, CEACAM5, EPCAM, ROR1, ROR2, STEAP1, MS4A1, CD22, SLC34A2, LGALS7B, TDGF1. Mechanistic Specificity Analysis To assess whether therapeutic antigen enrichment is attributable to surface protein abundance, we compared high-variance (top 10%, n = 704) and low-variance (bottom 10%, n = 703) gene sets for surface protein enrichment (Human Protein Atlas plasma membrane proteome, n = 2,263 genes), driver oncogene enrichment (OncoKB, n = 1,231 genes), and gold standard recall. Fisher's exact test was used throughout. Statistical Analysis Recall was defined as the proportion of gold standard targets in the top N% candidate set. Fisher's exact test (one-sided) compared recall rates between methods. All code: https://github.com/ohahouhui/TANK-framework . Results TANK substantially outperforms all transcriptomic prioritization methods Across all thresholds, TANK substantially outperformed all four comparator methods in therapeutic antigen recall (Fig. 1 ). At top 5%, TANK recovered 25% of gold standard targets, compared to 3.6% for DEG, 3.6% for MAD, 0% for CV, and 3.6% for Mean expression ranking. This advantage was consistent across all thresholds tested (top 10%: TANK 32% vs DEG 7%; top 20%: TANK 42% vs DEG 7%). Primary analysis: TANK vs DEG in full gene universe In the primary analysis using the full gene universe (60,654 genes), TANK recovered 14 of 28 gold standard targets at the top 5% threshold (50.0% recall) versus 1 of 28 for DEG (3.6%), representing a 13.9-fold improvement (OR = 27.0, p = 0.000071). Results were consistent across thresholds (Table 1 ). Table 1 Primary analysis: TANK vs DEG recall of 28 clinically validated therapeutic antigens. Full gene universe (60,654 genes). OR, odds ratio; Fisher’s exact test, one-sided. Threshold TANK Recall DEG Recall Fold Improvement OR p-value Top 5% 14/28 (50.0%) 1/28 (3.6%) 13.9× 27.0 7.1 × 10⁻⁵ Top 10% 18/28 (64.3%) 2/28 (7.1%) 9.0× 23.4 7.4 × 10⁻⁶ Top 20% 21/28 (75.0%) 2/28 (7.1%) 10.5× 39.0 < 1 × 10⁻⁶ TANK exclusively recovers 13 FDA-approved or Phase 2 + therapeutic antigen targets Of the 14 targets recovered by TANK at top 5% (full universe), 13 were not recovered by DEG. These include multiple FDA-approved drugs and advanced clinical programs (Table 2 ). The single target recovered by DEG was MAGEA4, a cancer-testis antigen with extreme mean differential expression. Table 2 Clinically validated therapeutic antigen targets recovered exclusively by TANK (not by DEG) at the top 5% threshold (full gene universe). Gene Therapeutic Agent Modality Status FOLR1 Mirvetuximab soravtansine ADC FDA-approved (2022) TACSTD2 Sacituzumab govitecan ADC FDA-approved (2020) CLDN18 Zolbetuximab Monoclonal antibody FDA-approved (2024) MSLN Multiple agents ADC / CAR-T Phase 2+ DLL3 Rovalpituzumab tesirine ADC Phase 2 GPC3 Multiple agents ADC / CAR-T Phase 2+ CLDN6 BNT211 CAR-T Phase 1/2 SLC34A2 TUB-040 ADC Phase 1/2 CEACAM5 Tusamitamab ravtansine ADC Phase 2 TNFRSF17 Belantamab mafodotin ADC FDA-approved (2020) MS4A1 Rituximab / Ofatumumab Monoclonal antibody FDA-approved CD22 Inotuzumab ozogamicin ADC FDA-approved (2017) LGALS7B Investigational ADC candidate Preclinical Three prospective convergences confirm predictive validity Prior to systematic gold standard analysis and without prior knowledge of active clinical programs, TANK nominated three targets that subsequently converged on ongoing clinical development: CLDN6 (top 0.22% by variance; BNT211 Phase 1/2, Nature Medicine 2023 [ 19 ]), SLC34A2 (highest pan-cancer recurrence across 20 cancer types; TUB-040 ADC Phase 1/2, ESMO 2025 [ 20 ]), and MAGEA4 (top variance in HNSC, LUSC, and OV; afamitresgene autoleucel FDA-approved August 2024, with Phase 2 expansion targeting these exact three cancer types [ 22 ]). These prospective convergences were identified before literature review and constitute independent validation of TANK's predictive capacity. Therapeutic antigen enrichment is not attributable to surface protein abundance To assess the mechanistic basis of therapeutic antigen enrichment, we compared high-variance (top 10%) and low-variance (bottom 10%) gene sets across three biological dimensions (Table 3 ). Critically, high-variance genes are not globally enriched for surface proteins (9.1% vs 9.2%, OR = 0.98, p = 0.58), ruling out surface protein abundance as an explanatory factor. Instead, high variance depletes canonical driver oncogenes (OR = 0.48, p = 0.0004) while achieving extreme selective enrichment of therapeutic antigens (32.1% vs 0.0% recall, OR = ∞, p = 0.002). These results indicate that transcriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to both surface protein abundance and driver gene biology. Table 3 Mechanistic specificity analysis. High-variance vs low-variance gene sets (filtered universe, 7,052 genes). n.s., not significant. Comparison High Variance (top 10%) Low Variance (bottom 10%) OR p-value Surface protein enrichment 9.1% (64/704) 9.2% (65/703) 0.98 0.58 (n.s.) Driver oncogene enrichment 5.1% (36/704) 10.1% (71/703) 0.48 0.0004 Therapeutic antigen recall 32.1% (9/28) 0.0% (0/28) ∞ 0.002 Discussion We demonstrate that variance-based transcriptomic prioritization (TANK) substantially outperforms all comparator methods — DEG, MAD, CV, and mean expression ranking — in recovering clinically validated therapeutic antigens. The consistent failure of all mean-based approaches (3.6%, 3.6%, 0%, 3.6% at top 5%) compared to TANK (25–50% depending on gene universe) indicates that this advantage is not a property of any specific mean-based method but reflects a fundamental limitation of mean-centric transcriptomic ranking for therapeutic antigen discovery. The mechanistic analysis reveals why variance captures therapeutic antigens more effectively. High transcriptomic variance does not simply reflect surface protein abundance — the proportion of surface proteins is indistinguishable between high- and low-variance gene sets (OR = 0.98). This rules out the parsimonious explanation that variance-based methods succeed by recovering more membrane proteins. Instead, the data support a more specific model: transcriptomic variance selectively identifies genes undergoing stochastic or context-dependent reactivation of developmentally silenced programs, precisely the category encompassing cancer-testis antigens, oncofetal proteins, and lineage-specific reactivation markers that form the basis of most approved immunotherapy targets. Mechanistically, high variance is associated with depletion of canonical driver oncogenes (OR = 0.48, p = 0.0004) while achieving extreme selective enrichment of therapeutic antigens over low-variance genes (OR = ∞, p = 0.002). We propose that this pattern reflects stochastic epigenetic derepression of developmentally silenced loci in tumor cells — a process producing high inter-patient variance precisely because it is context-dependent and non-deterministic. Among reactivated genes, those encoding surface-accessible proteins may be subject to secondary immune selection, though this mechanism requires experimental validation. These findings have practical implications for target discovery. Rather than replacing DEG, variance-based prioritization accesses a complementary discovery space: DEG efficiently identifies driver genes with broad tumor relevance, while TANK prioritizes heterogeneous antigen programs with selective patient subgroup expression — the profile required for therapeutic index in immunotherapy. Integration of both approaches may substantially expand the diversity of targetable antigens available for ADC, CAR-T, and TCR-T development. Limitations include the manual curation and limited size of the gold standard set (n = 28), single cancer type analysis (gastric adenocarcinoma), and the speculative nature of the proposed epigenetic derepression mechanism. Validation across additional cancer types and experimental confirmation of variance-associated epigenetic signatures are important next steps. Conclusions Transcriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to driver gene biology and systematically inaccessible to all mean-based transcriptomic prioritization methods tested. TANK recovered 25–50% of clinically validated therapeutic antigens depending on gene universe, compared to ≤ 7% for DEG, MAD, CV, and mean expression ranking. The 13 targets recovered exclusively by TANK include multiple FDA-approved immunotherapy drugs. These findings support integration of variance-based prioritization alongside conventional approaches in therapeutic antigen discovery pipelines. Declarations Competing Interests The author declares no competing interests. Data Availability All analysis code: https://github.com/ohahouhui/TANK-framework (MIT). TCGA-STAD data: GDC Data Portal (https://portal.gdc.cancer.gov/). Preprints on Research Square (submitted March 21–22, 2026; search 'Xiaoqi Hu TANK'). References Love MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15(12):550 Robinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26(1):139–140 Slamon DJ et al (2001) Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer. N Engl J Med 344(11):783–792 Reck M et al (2016) Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer. N Engl J Med 375(19):1823–1833 Haibe-Kains B et al (2020) Transparency and reproducibility in artificial intelligence. Nature 586(7829):E14–E16 Drago JZ, Modi S, Chandarlapaty S (2021) Unlocking the potential of antibody-drug conjugates for cancer therapy. Nat Rev Clin Oncol 18(6):327–344 Lim ZF, Ma PC (2019) Emerging insights of tumor heterogeneity and drug resistance mechanisms in lung cancer targeted therapy. J Hematol Oncol 12(1):134 Scanlan MJ et al (2002) Cancer/testis antigens: an expanding family of targets for cancer immunotherapy. Immunol Rev 188:22–32 Bhatt DL et al (2021) Oncofetal proteins as targets for cancer immunotherapy. Cancer Discov 11(3):546–560 Hu XTANK (2026) A Variance-Based Framework for Identifying Heterogeneous Therapeutic Targets in Gastric Cancer. Research Square. doi: pending Hu X (2026) Variance-based Decomposition of Inter-patient Transcriptomic Heterogeneity Reveals Recurrent Modes of Therapeutic Antigen Biology Across 33 Cancer Types. Research Square. doi: pending Hu X (2026) Variance Decomposition Accesses a Clinically Supported Discovery Space Systematically Missed by Mean-Based Transcriptomic Prioritization. Research Square. doi: pending Moore KN et al (2017) Phase 1 dose-escalation study of mirvetuximab soravtansine (IMGN853). Cancer 123(16):3080–3087 Bardia A et al (2019) Sacituzumab govitecan-hziy in refractory metastatic triple-negative breast cancer. N Engl J Med 380(8):741–751 Shitara K et al (2023) Zolbetuximab plus mFOLFOX6 in patients with CLDN18.2-positive gastric adenocarcinoma (SPOTLIGHT). Lancet 401(10389):1655–1668 Stein MN et al (2021) Rovalpituzumab tesirine in DLL3-high SCLC. JTO Clin Res Rep 2(4):100183 Nakamura Y et al (2023) Tusamitamab ravtansine plus docetaxel vs docetaxel in advanced CEACAM5 + NSCLC. ESMO Open 8(5):101617 Lonial S et al (2020) Belantamab mafodotin for relapsed or refractory multiple myeloma (DREAMM-2). Lancet Oncol 21(2):207–221 Schuler M et al (2023) Safety and efficacy of BNT211, a CLDN6-targeting CAR-T cell therapy. Nat Med 29:2237–2247 Colombo I et al TUB-040, a novel NaPi2b antibody-drug conjugate for ovarian cancer. ESMO 2025 Annual Meeting. Abstract 727O Stadtmauer EA et al (2020) CRISPR-engineered T cells in patients with refractory cancer. Science 367(6481):eaba7365 Haanen J et al (2023) Afamitresgene autoleucel for advanced synovial sarcoma. N Engl J Med 388(16):1459–1470 Meyers DE et al (2021) Targeting the folate receptor alpha for personalized therapy across multiple cancers. Target Oncol 16(2):149–162 DepMap Broad. DepMap 23Q4 Public. Figshare (2023) 10.6084/m9.figshare.24667905.v2 GTEx Consortium (2020) The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369(6509):1318–1330 Uhlen M et al (2017) A pathology atlas of the human cancer transcriptome. Science 357(6352):eaan2507 Tsherniak A et al (2017) Defining a cancer dependency map. Cell 170(3):564–576 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9215598","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":611584813,"identity":"abb7130f-4fd0-4554-8499-e8ce81fa3c39","order_by":0,"name":"XIAOQI HU","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+UlEQVRIiWNgGAWjYJCCA4x/2OTs2/s/PgByePiI09LAZ2zAc8DYAKSFjShrGBvkEjdIJJhJgDgEtci39x488HOHWeJ2hoS0yq85djJsDMwPH93Ao8XgzLmEg71n0ox3Nhw4dlt2WzLQYWzGxjn4tEjkGBzgYTsm23Cwse225DZmoBYeNml8WuRn5Bgc/MP2n7HhMDNbseS2esJaGG7kGBzmbWNT3HCMjY3x47bDhLUYnDljcFjmDJuxZA8PszTjtuM8bMwE/CLf3mP88U0Fmxy//BvGjz+3Vdvzszc/fIzXYciAmQdMEqscBBh/kKJ6FIyCUTAKRgwAAMqYSkbSxxrIAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0009-9829-6404","institution":"Independent Researcher","correspondingAuthor":true,"prefix":"","firstName":"XIAOQI","middleName":"","lastName":"HU","suffix":""}],"badges":[],"createdAt":"2026-03-24 19:30:44","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9215598/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9215598/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105460702,"identity":"c08cacdc-2fa8-4d7b-99f0-1cfe97a0ebe1","added_by":"auto","created_at":"2026-03-26 09:58:07","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":79449,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eRecall of clinically validated therapeutic antigens by five transcriptomic prioritization methods across three ranking thresholds. TANK (variance-based) consistently outperforms DEG, MAD, CV, and mean expression ranking. Random baseline shown for reference. Common gene universe: 7,052 genes. Gold standard: 28 FDA-approved or Phase 2+ therapeutic antigen targets.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9215598/v1/07c6f6dbfc689606a6c0e409.jpeg"},{"id":105566311,"identity":"5242dbbf-39ab-4279-aec0-5920a883aa07","added_by":"auto","created_at":"2026-03-27 12:56:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":713102,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9215598/v1/3d21c680-3510-4f3c-881e-529b6dcf8a97.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eVariance-based Prioritization Reveals a Clinically Validated Antigen Discovery Space Systematically Inaccessible to Mean-Based Methods\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCancer target discovery has been dominated by differential expression analysis (DEG), a framework that identifies genes whose mean expression levels differ significantly between tumor and normal tissue [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. This approach has proven highly effective for identifying driver oncogenes and tumor suppressor genes, yielding landmark therapeutic targets including ERBB2, EGFR, and MYC [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. DEG-based prioritization underlies the majority of target identification pipelines in both academic and industrial drug discovery [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHowever, DEG is optimized for a specific biological task: identifying genes with elevated average expression in tumor cells. Therapeutic antigens for antibody-drug conjugates (ADCs), CAR-T cells, TCR-T cells, and bispecific antibodies require a fundamentally different expression profile: high expression in a targetable patient subpopulation with low expression in normal tissues [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. A gene with high inter-patient variance but moderate mean expression would be systematically deprioritized by DEG despite potentially representing an ideal immunotherapy target.\u003c/p\u003e \u003cp\u003eMany successful therapeutic antigens are characterized by stochastic reactivation of silenced developmental gene programs \u0026mdash; cancer-testis antigens (MAGEA4, CLDN6), oncofetal proteins (GPC3, FOLR1), and lineage-specific markers (SLC34A2, MSLN) \u0026mdash; producing high inter-patient expression variance [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. We previously developed TANK (Transcriptomic ANtigen variance-based prioritization), a framework ranking genes by inter-patient transcriptomic variance, and demonstrated three prospective convergences on clinically validated targets [\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. However, a systematic comparison of TANK against multiple transcriptomic prioritization methods has not been performed. Here we provide the first such comparison against a curated gold standard of 28 therapeutic antigens, and characterize the mechanistic basis for the distinct discovery space captured by variance.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eTranscriptomic Data\u003c/h2\u003e \u003cp\u003eRNA-sequencing count data for TCGA-STAD (gastric adenocarcinoma) were obtained from the GDC Data Portal (n\u0026thinsp;=\u0026thinsp;443 tumor samples, 45 matched normal samples; STAR aligner, GENCODE v36). Gene identifiers (ENSEMBL GRCh38, version suffixes removed) were mapped to HGNC symbols using a curated translation table.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eTranscriptomic Prioritization Methods\u003c/h3\u003e\n\u003cp\u003eFive prioritization methods were evaluated on a common gene universe of 7,052 genes with available rankings across all methods: (1) TANK: inter-patient variance across 443 tumor samples, descending rank; (2) DEG: absolute log2 fold change from DESeq2 (v1.36), tumor vs normal, descending rank; (3) MAD: median absolute deviation across tumor samples, descending rank; (4) CV: coefficient of variation (SD/mean) across tumor samples, descending rank; (5) Mean: mean expression across tumor samples, descending rank. For the primary TANK vs DEG recall analysis (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), the full TANK ranking (60,654 genes) was used to avoid artificially constraining the variance-based candidate space; for multi-method comparison (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), the common 7,052-gene universe was used to ensure fair comparison.\u003c/p\u003e\n\u003ch3\u003eGold Standard Therapeutic Antigen Set\u003c/h3\u003e\n\u003cp\u003eA gold standard of 28 clinically validated therapeutic antigens was curated from FDA approval records, ClinicalTrials.gov interventional studies, and peer-reviewed publications (accessed March 2026). Inclusion criteria: (1) direct target of an ADC, CAR-T, TCR-T, or bispecific antibody; (2) FDA approval or Phase 2\u0026thinsp;+\u0026thinsp;status; (3) cell-surface or secreted protein. Tier 1 (FDA-approved, n\u0026thinsp;=\u0026thinsp;10): ERBB2, EGFR, PDCD1, CD274, CTLA4, MSLN, NECTIN4, TACSTD2, CD19, TNFRSF17. Tier 2 (Phase 2+, n\u0026thinsp;=\u0026thinsp;18): CLDN18, CLDN6, MAGEA4, DLL3, GPC3, FOLR1, FOLH1, CD276, CEACAM5, EPCAM, ROR1, ROR2, STEAP1, MS4A1, CD22, SLC34A2, LGALS7B, TDGF1.\u003c/p\u003e\n\u003ch3\u003eMechanistic Specificity Analysis\u003c/h3\u003e\n\u003cp\u003eTo assess whether therapeutic antigen enrichment is attributable to surface protein abundance, we compared high-variance (top 10%, n\u0026thinsp;=\u0026thinsp;704) and low-variance (bottom 10%, n\u0026thinsp;=\u0026thinsp;703) gene sets for surface protein enrichment (Human Protein Atlas plasma membrane proteome, n\u0026thinsp;=\u0026thinsp;2,263 genes), driver oncogene enrichment (OncoKB, n\u0026thinsp;=\u0026thinsp;1,231 genes), and gold standard recall. Fisher's exact test was used throughout.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eRecall was defined as the proportion of gold standard targets in the top N% candidate set. Fisher's exact test (one-sided) compared recall rates between methods. All code: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/ohahouhui/TANK-framework\u003c/span\u003e\u003cspan address=\"https://github.com/ohahouhui/TANK-framework\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eTANK substantially outperforms all transcriptomic prioritization methods\u003c/h2\u003e \u003cp\u003eAcross all thresholds, TANK substantially outperformed all four comparator methods in therapeutic antigen recall (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). At top 5%, TANK recovered 25% of gold standard targets, compared to 3.6% for DEG, 3.6% for MAD, 0% for CV, and 3.6% for Mean expression ranking. This advantage was consistent across all thresholds tested (top 10%: TANK 32% vs DEG 7%; top 20%: TANK 42% vs DEG 7%).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePrimary analysis: TANK vs DEG in full gene universe\u003c/h3\u003e\n\u003cp\u003eIn the primary analysis using the full gene universe (60,654 genes), TANK recovered 14 of 28 gold standard targets at the top 5% threshold (50.0% recall) versus 1 of 28 for DEG (3.6%), representing a 13.9-fold improvement (OR\u0026thinsp;=\u0026thinsp;27.0, p\u0026thinsp;=\u0026thinsp;0.000071). Results were consistent across thresholds (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePrimary analysis: TANK vs DEG recall of 28 clinically validated therapeutic antigens. Full gene universe (60,654 genes). OR, odds ratio; Fisher\u0026rsquo;s exact test, one-sided.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003e Threshold\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTANK Recall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDEG Recall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFold Improvement\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTop 5%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e14/28 (50.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1/28 (3.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c4\"\u003e \u003cp\u003e13.9\u0026times;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e27.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c6\"\u003e \u003cp\u003e7.1 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTop 10%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18/28 (64.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2/28 (7.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c4\"\u003e \u003cp\u003e9.0\u0026times;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e23.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c6\"\u003e \u003cp\u003e7.4 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTop 20%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21/28 (75.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2/28 (7.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c4\"\u003e \u003cp\u003e10.5\u0026times;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e39.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;1 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eTANK exclusively recovers 13 FDA-approved or Phase 2\u0026thinsp;+\u0026thinsp;therapeutic antigen targets\u003c/h2\u003e \u003cp\u003eOf the 14 targets recovered by TANK at top 5% (full universe), 13 were not recovered by DEG. These include multiple FDA-approved drugs and advanced clinical programs (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The single target recovered by DEG was MAGEA4, a cancer-testis antigen with extreme mean differential expression.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClinically validated therapeutic antigen targets recovered exclusively by TANK (not by DEG) at the top 5% threshold (full gene universe).\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGene\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTherapeutic Agent\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModality\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStatus\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFOLR1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMirvetuximab soravtansine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved (2022)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTACSTD2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSacituzumab govitecan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved (2020)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCLDN18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eZolbetuximab\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMonoclonal antibody\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved (2024)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMSLN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMultiple agents\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC / CAR-T\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 2+\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDLL3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRovalpituzumab tesirine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGPC3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMultiple agents\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC / CAR-T\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 2+\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCLDN6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBNT211\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCAR-T\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 1/2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSLC34A2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTUB-040\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 1/2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCEACAM5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTusamitamab ravtansine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTNFRSF17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBelantamab mafodotin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved (2020)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMS4A1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRituximab / Ofatumumab\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMonoclonal antibody\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCD22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInotuzumab ozogamicin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFDA-approved (2017)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLGALS7B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInvestigational\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eADC candidate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePreclinical\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eThree prospective convergences confirm predictive validity\u003c/h2\u003e \u003cp\u003ePrior to systematic gold standard analysis and without prior knowledge of active clinical programs, TANK nominated three targets that subsequently converged on ongoing clinical development: CLDN6 (top 0.22% by variance; BNT211 Phase 1/2, Nature Medicine 2023 [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]), SLC34A2 (highest pan-cancer recurrence across 20 cancer types; TUB-040 ADC Phase 1/2, ESMO 2025 [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]), and MAGEA4 (top variance in HNSC, LUSC, and OV; afamitresgene autoleucel FDA-approved August 2024, with Phase 2 expansion targeting these exact three cancer types [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]). These prospective convergences were identified before literature review and constitute independent validation of TANK's predictive capacity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eTherapeutic antigen enrichment is not attributable to surface protein abundance\u003c/h2\u003e \u003cp\u003eTo assess the mechanistic basis of therapeutic antigen enrichment, we compared high-variance (top 10%) and low-variance (bottom 10%) gene sets across three biological dimensions (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Critically, high-variance genes are not globally enriched for surface proteins (9.1% vs 9.2%, OR\u0026thinsp;=\u0026thinsp;0.98, p\u0026thinsp;=\u0026thinsp;0.58), ruling out surface protein abundance as an explanatory factor. Instead, high variance depletes canonical driver oncogenes (OR\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;=\u0026thinsp;0.0004) while achieving extreme selective enrichment of therapeutic antigens (32.1% vs 0.0% recall, OR = \u0026infin;, p\u0026thinsp;=\u0026thinsp;0.002). These results indicate that transcriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to both surface protein abundance and driver gene biology.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMechanistic specificity analysis. High-variance vs low-variance gene sets (filtered universe, 7,052 genes). n.s., not significant.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComparison\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh Variance (top 10%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLow Variance (bottom 10%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSurface protein enrichment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9.1% (64/704)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.2% (65/703)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.58 (n.s.)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDriver oncogene enrichment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5.1% (36/704)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e10.1% (71/703)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0004\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTherapeutic antigen recall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e32.1% (9/28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0% (0/28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026infin;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.002\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe demonstrate that variance-based transcriptomic prioritization (TANK) substantially outperforms all comparator methods \u0026mdash; DEG, MAD, CV, and mean expression ranking \u0026mdash; in recovering clinically validated therapeutic antigens. The consistent failure of all mean-based approaches (3.6%, 3.6%, 0%, 3.6% at top 5%) compared to TANK (25\u0026ndash;50% depending on gene universe) indicates that this advantage is not a property of any specific mean-based method but reflects a fundamental limitation of mean-centric transcriptomic ranking for therapeutic antigen discovery.\u003c/p\u003e \u003cp\u003eThe mechanistic analysis reveals why variance captures therapeutic antigens more effectively. High transcriptomic variance does not simply reflect surface protein abundance \u0026mdash; the proportion of surface proteins is indistinguishable between high- and low-variance gene sets (OR\u0026thinsp;=\u0026thinsp;0.98). This rules out the parsimonious explanation that variance-based methods succeed by recovering more membrane proteins. Instead, the data support a more specific model: transcriptomic variance selectively identifies genes undergoing stochastic or context-dependent reactivation of developmentally silenced programs, precisely the category encompassing cancer-testis antigens, oncofetal proteins, and lineage-specific reactivation markers that form the basis of most approved immunotherapy targets.\u003c/p\u003e \u003cp\u003eMechanistically, high variance is associated with depletion of canonical driver oncogenes (OR\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;=\u0026thinsp;0.0004) while achieving extreme selective enrichment of therapeutic antigens over low-variance genes (OR = \u0026infin;, p\u0026thinsp;=\u0026thinsp;0.002). We propose that this pattern reflects stochastic epigenetic derepression of developmentally silenced loci in tumor cells \u0026mdash; a process producing high inter-patient variance precisely because it is context-dependent and non-deterministic. Among reactivated genes, those encoding surface-accessible proteins may be subject to secondary immune selection, though this mechanism requires experimental validation.\u003c/p\u003e \u003cp\u003eThese findings have practical implications for target discovery. Rather than replacing DEG, variance-based prioritization accesses a complementary discovery space: DEG efficiently identifies driver genes with broad tumor relevance, while TANK prioritizes heterogeneous antigen programs with selective patient subgroup expression \u0026mdash; the profile required for therapeutic index in immunotherapy. Integration of both approaches may substantially expand the diversity of targetable antigens available for ADC, CAR-T, and TCR-T development.\u003c/p\u003e \u003cp\u003eLimitations include the manual curation and limited size of the gold standard set (n\u0026thinsp;=\u0026thinsp;28), single cancer type analysis (gastric adenocarcinoma), and the speculative nature of the proposed epigenetic derepression mechanism. Validation across additional cancer types and experimental confirmation of variance-associated epigenetic signatures are important next steps.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eTranscriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to driver gene biology and systematically inaccessible to all mean-based transcriptomic prioritization methods tested. TANK recovered 25\u0026ndash;50% of clinically validated therapeutic antigens depending on gene universe, compared to \u0026le;\u0026thinsp;7% for DEG, MAD, CV, and mean expression ranking. The 13 targets recovered exclusively by TANK include multiple FDA-approved immunotherapy drugs. These findings support integration of variance-based prioritization alongside conventional approaches in therapeutic antigen discovery pipelines.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting Interests\u003c/h2\u003e \u003cp\u003eThe author declares no competing interests.\u003c/p\u003e \u003c/p\u003e\u003cp\u003eData Availability\u003c/p\u003e\n\u003cp\u003eAll analysis code: https://github.com/ohahouhui/TANK-framework (MIT). TCGA-STAD data: GDC Data Portal (https://portal.gdc.cancer.gov/). Preprints on Research Square (submitted March 21\u0026ndash;22, 2026; search \u0026apos;Xiaoqi Hu TANK\u0026apos;).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLove MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15(12):550\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRobinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26(1):139\u0026ndash;140\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlamon DJ et al (2001) Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer. N Engl J Med 344(11):783\u0026ndash;792\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReck M et al (2016) Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer. N Engl J Med 375(19):1823\u0026ndash;1833\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaibe-Kains B et al (2020) Transparency and reproducibility in artificial intelligence. Nature 586(7829):E14\u0026ndash;E16\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDrago JZ, Modi S, Chandarlapaty S (2021) Unlocking the potential of antibody-drug conjugates for cancer therapy. Nat Rev Clin Oncol 18(6):327\u0026ndash;344\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLim ZF, Ma PC (2019) Emerging insights of tumor heterogeneity and drug resistance mechanisms in lung cancer targeted therapy. J Hematol Oncol 12(1):134\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScanlan MJ et al (2002) Cancer/testis antigens: an expanding family of targets for cancer immunotherapy. Immunol Rev 188:22\u0026ndash;32\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhatt DL et al (2021) Oncofetal proteins as targets for cancer immunotherapy. Cancer Discov 11(3):546\u0026ndash;560\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu XTANK (2026) A Variance-Based Framework for Identifying Heterogeneous Therapeutic Targets in Gastric Cancer. Research Square. doi: pending\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu X (2026) Variance-based Decomposition of Inter-patient Transcriptomic Heterogeneity Reveals Recurrent Modes of Therapeutic Antigen Biology Across 33 Cancer Types. Research Square. doi: pending\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu X (2026) Variance Decomposition Accesses a Clinically Supported Discovery Space Systematically Missed by Mean-Based Transcriptomic Prioritization. Research Square. doi: pending\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoore KN et al (2017) Phase 1 dose-escalation study of mirvetuximab soravtansine (IMGN853). Cancer 123(16):3080\u0026ndash;3087\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBardia A et al (2019) Sacituzumab govitecan-hziy in refractory metastatic triple-negative breast cancer. N Engl J Med 380(8):741\u0026ndash;751\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShitara K et al (2023) Zolbetuximab plus mFOLFOX6 in patients with CLDN18.2-positive gastric adenocarcinoma (SPOTLIGHT). Lancet 401(10389):1655\u0026ndash;1668\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStein MN et al (2021) Rovalpituzumab tesirine in DLL3-high SCLC. JTO Clin Res Rep 2(4):100183\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNakamura Y et al (2023) Tusamitamab ravtansine plus docetaxel vs docetaxel in advanced CEACAM5\u0026thinsp;+\u0026thinsp;NSCLC. ESMO Open 8(5):101617\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLonial S et al (2020) Belantamab mafodotin for relapsed or refractory multiple myeloma (DREAMM-2). Lancet Oncol 21(2):207\u0026ndash;221\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchuler M et al (2023) Safety and efficacy of BNT211, a CLDN6-targeting CAR-T cell therapy. Nat Med 29:2237\u0026ndash;2247\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColombo I et al TUB-040, a novel NaPi2b antibody-drug conjugate for ovarian cancer. ESMO 2025 Annual Meeting. Abstract 727O\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStadtmauer EA et al (2020) CRISPR-engineered T cells in patients with refractory cancer. Science 367(6481):eaba7365\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaanen J et al (2023) Afamitresgene autoleucel for advanced synovial sarcoma. N Engl J Med 388(16):1459\u0026ndash;1470\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeyers DE et al (2021) Targeting the folate receptor alpha for personalized therapy across multiple cancers. Target Oncol 16(2):149\u0026ndash;162\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDepMap Broad. DepMap 23Q4 Public. Figshare (2023) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.6084/m9.figshare.24667905.v2\u003c/span\u003e\u003cspan address=\"10.6084/m9.figshare.24667905.v2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGTEx Consortium (2020) The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369(6509):1318\u0026ndash;1330\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUhlen M et al (2017) A pathology atlas of the human cancer transcriptome. Science 357(6352):eaan2507\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsherniak A et al (2017) Defining a cancer dependency map. Cell 170(3):564\u0026ndash;576\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Shanghai Normal University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"therapeutic antigen discovery, transcriptomic variance, antibody-drug conjugates, differential expression analysis, cancer immunotherapy, target prioritization","lastPublishedDoi":"10.21203/rs.3.rs-9215598/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9215598/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground.\u003c/h2\u003e \u003cp\u003eMean-based transcriptomic prioritization (differential expression analysis, DEG) dominates cancer target discovery but is optimized for driver gene identification rather than therapeutic antigen discovery. Whether variance-based prioritization captures a complementary and clinically relevant discovery space has not been systematically evaluated.\u003c/p\u003e\u003ch2\u003eMethods.\u003c/h2\u003e \u003cp\u003eWe applied variance-based prioritization (TANK) and four comparator methods (DEG, MAD, coefficient of variation, mean expression) to genome-wide transcriptomic data from TCGA gastric adenocarcinoma (n\u0026thinsp;=\u0026thinsp;443 tumor samples). We evaluated recall of a gold standard set of 28 clinically validated therapeutic antigens (FDA-approved and Phase 2\u0026thinsp;+\u0026thinsp;ADC/CAR-T/TCR-T targets) at three ranking thresholds. Mechanistic specificity was assessed by comparing surface protein enrichment, driver oncogene enrichment, and therapeutic antigen recall between high-variance and low-variance gene sets.\u003c/p\u003e\u003ch2\u003eResults.\u003c/h2\u003e \u003cp\u003eAcross all thresholds, TANK substantially outperformed all comparator methods in therapeutic antigen recall (top 5%: TANK 25%, DEG 3.6%, MAD 3.6%, CV 0%, Mean 3.6%). In the primary analysis using the full gene universe (60,654 genes), TANK recovered 50% of gold standard targets at top 5% versus 3.6% for DEG (OR\u0026thinsp;=\u0026thinsp;27.0, p\u0026thinsp;=\u0026thinsp;0.000071). High-variance genes were not globally enriched for surface proteins (9.1% vs 9.2%, OR\u0026thinsp;=\u0026thinsp;0.98, p\u0026thinsp;=\u0026thinsp;0.58), ruling out surface protein abundance as an explanatory factor. Instead, high variance specifically depleted canonical driver oncogenes (OR\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;=\u0026thinsp;0.0004) while achieving extreme enrichment of therapeutic antigens over low-variance genes (OR\u0026thinsp;=\u0026thinsp;infinity, p\u0026thinsp;=\u0026thinsp;0.002). Three targets nominated prospectively by TANK prior to literature review subsequently converged on FDA-approved or Phase 2\u0026thinsp;+\u0026thinsp;clinical programs.\u003c/p\u003e\u003ch2\u003eConclusions.\u003c/h2\u003e \u003cp\u003eTranscriptomic variance encodes a specific biological signal for therapeutic antigenicity that is orthogonal to driver gene biology and systematically inaccessible to all mean-based approaches tested. Integration of variance-based prioritization into target discovery workflows may substantially expand the accessible space for immunotherapy antigen development.\u003c/p\u003e","manuscriptTitle":"Variance-based Prioritization Reveals a Clinically Validated Antigen Discovery Space Systematically Inaccessible to Mean-Based Methods","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-26 09:57:07","doi":"10.21203/rs.3.rs-9215598/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4b070e46-80c9-4d9f-a993-e43cfdb56da2","owner":[],"postedDate":"March 26th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":65074705,"name":"Cancer Biology"}],"tags":[],"updatedAt":"2026-03-26T09:57:07+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-26 09:57:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9215598","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9215598","identity":"rs-9215598","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}