--- title: Euretos Target Assessement description: An overview of the Euretos Target Assessment workflow for a particular gene - disease combination. --- [![Euretos Logo](https://www.euretos.com/hubfs/Euretos%20Logo%202020%20-%20Strapline%20-%20White%20small.png)](https://www.euretos.com?hsLang=en) [Disease Atlas](https://www.euretos.com/disease-atlas?hsLang=en) [Get Access](https://www.euretos.com/pricing?hsLang=en) [Blog](https://www.euretos.com/blog?hsLang=en) [![Login](https://hubspot-no-cache-eu1-prod.s3.amazonaws.com/cta/default/7609166/interactive-167498164057.png)](https://cta-eu1.hubspot.com/web-interactives/public/v1/track/redirect?encryptedPayload=AVxigLJf9pG9laoInA%2Bz8gKxkaZMXkYxVyrWH1l5uNHVK%2Fo5oOtg%2B143DZtLL6U8nfdppuWGIHX22fcdi2LKOcCWNpHF7l4L1Cy6ljyre9W1LuhvqK7Le4IObRfcDMu1ZeUcGj9AS2K2VBbUvXF7LJcfZDfdIINbh9uVaq1L56f4cZiLgGpS4XVfBg%3D%3D&webInteractiveContentId=167498164057&portalId=7609166) [![Contact Us](https://hubspot-no-cache-eu1-prod.s3.amazonaws.com/cta/default/7609166/interactive-167498163994.png)](https://cta-eu1.hubspot.com/web-interactives/public/v1/track/redirect?encryptedPayload=AVxigLJf9pG9laoInA%2Bz8gKxkaZMXkYxVyrWH1l5uNHVK%2Fo5oOtg%2B143DZtLL6U8nfdppuWGIHX22fcdi2LKOcCWNpHF7l4L1Cy6ljyre9W1LuhvqK7Le4IObRfcDMu1ZeUcGj9AS2K2VBbUvXF7LJcfZDfdIINbh9uVaq1L56f4cZiLgGpS4XVfBg%3D%3D&webInteractiveContentId=167498163994&portalId=7609166) Panels [1 · Genetic Evidence](https://www.euretos.com/target-assessment-example#genetic-evidence)[2 · Developability](https://www.euretos.com/target-assessment-example#developability)[3 · Safety](https://www.euretos.com/target-assessment-example#safety)[4 · Competition](https://www.euretos.com/target-assessment-example#competition)[5 · Score Overview](https://www.euretos.com/target-assessment-example#score-decomposition)[6 · Cell-Type Expression](https://www.euretos.com/target-assessment-example#transcriptomics)[7 · Differential Expression](https://www.euretos.com/target-assessment-example#differential-expression)[8 · Mechanism](https://www.euretos.com/target-assessment-example#mechanistic-chain)[9 · Protein Networks](https://www.euretos.com/target-assessment-example#protein-networks)[10 · Phenotypes](https://www.euretos.com/target-assessment-example#phenotypes)[11 · Literature](https://www.euretos.com/target-assessment-example#literature)[12 · Clinical Trials](https://www.euretos.com/target-assessment-example#clinical-trials)[13 · Protein Structure](https://www.euretos.com/target-assessment-example#protein-structure)[14 · Perturbation](https://www.euretos.com/target-assessment-example#perturbation)[15 · Experimental Models](https://www.euretos.com/target-assessment-example#cell-lines) # Target assessment · Disease Atlas by Euretos – TNFSF15 in ulcerative colitis This page reproduces a target assessment from the Disease Atlas as the platform generates it for one target-disease pair. A score summary and key findings open the page. The panels below bring together genetic evidence, developability, safety, competition, expression, mechanism, protein networks, phenotypes, literature, clinical trials, protein structure, perturbation and experimental models. Each panel has a “How to read this” button that explains what the panel answers, how to read it, and where its limits are. This is a static copy of the page. In the platform, findings link through to the underlying evidence, and tabs and tooltips can be opened. Those features are not available here. Role of TNFSF15 (homo sapiens) in ulcerative colitis ENSG00000181634O95150HGNC:11931 How to read this page EXAMPLE REPORTTarget–disease assessment: causal biology, translational evidence, tractability and safety in one view.Higher-confidence diseaseData snapshot 11 June 2026 Biology 100 Translational 100 Tractability 77 Safety 63 Priority 90 Confidence High Biology drivers:Genetic1.00Network0.25Expression0.17Transfer0.48 Biology Strong causal evidence. Strongest support from genetic association. Translational High development plausibility, supported by multiple genetic risk loci. Safety No safety concerns identified. Tractability Best modality: antibody. Accessible extracellular domain. Recommended posture Pursue with conditions Synthesis of all evidence axes ModalityAntibody CompetitionModerate ValidationPhase 3 Opportunity note5 antibodies in UC trials, 3 in Phase 3, none approved EvidenceHigh DirectionInhibit ### Key findings - **Strong causal genetic support** Human genetics places TNFSF15 among the strongest causal candidates for ulcerative colitis: a biology score of 99.8 out of 100, driven almost entirely by genetic evidence. Six independently fine-mapped credible sets converge on the gene. - **Direction of effect matches the clinic** In digestive-tract tissues (stomach, oesophageal mucosa, pancreas), genetically higher TNFSF15 expression raises ulcerative colitis risk. That points to blocking the target, which is exactly what every clinical programme against it does. - **Clinically validated, not yet approved** Five antibodies against TNFSF15 (the protein TL1A) are in ulcerative colitis trials, three of them in Phase 3: tulisokibart, duvakitug and afimkibart. No drug against the target is approved yet. - **Clean safety profile** Low predicted risk in all six organs assessed. Expression is tissue-restricted, and the gene tolerates loss-of-function variation in the human population. ### 1Genetic Evidence How to read this Biology score 99.8 Genetic driver 99.6% Recommended action Inhibit · supported in disease-relevant tissues Genetic instruments in the digestive tract (stomach, oesophageal mucosa, pancreas) show that higher TNFSF15 expression raises ulcerative colitis risk. Instruments in blood and skin point the other way, so the call rests on the tissues closest to the disease. Direction by tissue Digestive tract: risk-raising · Blood, skin: risk-lowering Confidence Higher-confidence disease · evidence confidence high Mendelian randomisation: effect of genetically predicted TNFSF15 expression on ulcerative colitis, by tissue Whole\_Blood: Steiger-correct — the gene-to-disease causal direction holds for this instrument.n=10, F=84Whole\_BloodStomach: Steiger-correct — the gene-to-disease causal direction holds for this instrument.n=10, F=40StomachEsophagus\_Mucosa: Steiger-correct — the gene-to-disease causal direction holds for this instrument.n=10, F=36Esophagus\_MucosaSkin\_Sun\_Exposed\_Lower\_leg: Steiger-correct — the gene-to-disease causal direction holds for this instrument.n=4, F=18Skin\_Sun\_Exposed\_Lower\_legPancreas: Steiger-correct — the gene-to-disease causal direction holds for this instrument.n=3, F=19Pancreas-0.40-0.200.000.200.400.600.80β (standardised effect size) Steiger-correct Reverse causation candidate ·Sorted by rank · point size reflects number of genetic variants · bars = 95% confidence interval Strong causal genetic support **TNFSF15** ranks in the top 0.2% of genes for **ulcerative colitis** on causal biology (99.8 / 100). The case rests on human genetics: genome-wide association signals, fine-mapping and Mendelian randomisation all point to this gene. Supporting evidence Fine-mapped credible set converges on this gene 6 variants reach a locus-to-gene score of 0.5 or better for **TNFSF15**, and 6 SuSiE / 1 PICS fine-mapped credible sets sit behind them. Convergent fine-mapping makes the gene assignment at this locus much harder to explain by linkage disequilibrium alone. Risk variants act on expression All 7 recorded variants for **TNFSF15** are regulatory: none changes the protein sequence. Risk is driven by how much TNFSF15 is made, which fits a blocking antibody approach. ### 2Developability How to read this Modality fit Format scores on a 0–100 scale Best modalityAntibody Antibody score90 / 100 · HIGH Small moleculeNo known binders Developability riskLOW · 0.20 | Format | Fit | Why | | --- | --- | --- | | **Blocking antibody** | HIGH | Accessible extracellular domain and selective epitope space; every TL1A drug in Phase 2 or 3 is a monoclonal antibody | | Small molecule | LOW | No ChEMBL ligands; the target acts through a protein–protein interface with its receptor | | ADC, T-cell engager | NOT THE GOAL | TL1A is a signalling ligand to neutralise, not a cell population to remove | What drives the antibody score Component scores, 0 = unfavourable, 1 = ideal Surface accessibility1.0Selectivity (2 paralogs)1.0Extracellular domain size1.0Approved-drug precedent0.0No TL1A drug is approved yet, so the precedent component is 0. Extracellular fraction0.78 Membrane proteinYes Close paralogs2 Selectivity riskLOW Antibody is the right format 78% of the **TNFSF15** protein lies outside the cell, it has only two close relatives and no known small-molecule binders. A selective blocking antibody is the natural format, and every TL1A drug in Phase 2 or 3 is one. DATA SOURCE Modality scoring Surface access, selectivity, domain size and precedent from the platform modality model; ligand counts from ChEMBL. ### 3Safety How to read this Organ risk matrix Rows: organs · Columns: evidence types | | Predicted | Post-market reports | Adverse-event data | Clinical | | --- | --- | --- | --- | --- | | Liver | LOW | — | — | — | | Heart | LOW | — | — | — | | Kidney | LOW | — | — | — | | Lung | LOW | — | — | — | | CNS | LOW | — | — | — | | Immune | LOW | — | — | — | Overall riskLOW · 0.15Tissue specificity0.85Cell-type specificity0.94pLI0.00LOEUF0.98 Clean safety profile Predicted risk is **low** in all six organs assessed (liver, heart, kidney, lung, central nervous system, immune). TNFSF15 is expressed in a narrow set of tissues and cell types, and the gene tolerates loss-of-function variation in the human population (LOEUF 0.98). No TNFSF15 drug is approved yet, so organ risks are model predictions rather than post-market reports. ### 4Competition How to read this Antibodies in UC trials5 In Phase 33 Approved0 Sponsors5 Anti-TL1A programmes in ulcerative colitis Highest phase reached in UC, with key UC trials (enrolment) | Drug | Sponsor | UC phase | Key UC trials | Other indications in trials | | --- | --- | --- | --- | --- | | **Tulisokibart** MK-7240, ex PRA023 | Merck (via Prometheus) | **Phase 3** | NCT06052059 ATLAS-UC (1,020) NCT06651281 UC + Crohn's (1,380) | Crohn's (Ph 3), SSc-ILD, HS, RA, axSpA, PsA (Ph 2) | | **Duvakitug** TEV-48574 | Sanofi and Teva | **Phase 3** | NCT07184996 (980) NCT07185009 (671) | Crohn's (Ph 3); asthma (Ph 2, terminated) | | **Afimkibart** RVT-3101, RO7790121 | Roche | **Phase 3** | NCT06589986 (400) NCT06588855 (350) NCT07158242 | Crohn's (Ph 3), atopic dermatitis, RA (Ph 2), MASH (Ph 1) | | Anti-TL1A antibody | Biocad | Phase 2 | NCT07080034 (198) | Crohn's (Ph 2) | | BB-TL1A | Vial | Phase 1 (terminated) | NCT07029971 | – | Where the genetics points next Disease Atlas causal-biology percentile for TNFSF15 versus registered TL1A trials | Disease | Biology percentile | TL1A trials registered | | --- | --- | --- | | Diffuse cutaneous systemic sclerosis | 100.0 | Tulisokibart Ph 2 (lung involvement) | | Eosinophilic oesophagitis | 99.7 | **None registered** | | Ulcerative colitis | 99.8 | Three programmes in Phase 3 | | Systemic lupus erythematosus | 99.3 | **None registered** | | Psoriasis | 98.0 | **None registered** | | Crohn's disease | 72.1 | Three programmes in Phase 3 | Crowded at the front, open behind Three sponsors are in Phase 3 in **ulcerative colitis**, so a new entrant needs a clear point of difference. The genetics points to indications no one is testing: eosinophilic oesophagitis and lupus both rank in the top 1% of genes for **TNFSF15**, with no TL1A trials registered. DATA SOURCE ClinicalTrials.gov (AACT registry) Interventional trials naming an anti-TL1A antibody, April 2026 extract. Biology percentiles from the 11 June 2026 Disease Atlas snapshot. ### 5Score Overview How to read this Five-axis score profile Biology · Translational · Tractability · Safety · Priority — 0–100 Biology: 99.8 — Strong causal evidence. Strongest support from genetic association.Translational: 100.0 — High development plausibility. Key factors: strong genetic support (multiple GWAS loci).Tractability: 77.0 — Best modality: mab.Safety: 62.5 — No safety concerns identified.Priority: 90.4Biology100Translational100Tractability77Safety63Priority90 Evidence confidence: **high** · higher-confidence disease Biology drivers:Genetic1.00Network0.25Expression0.17Transfer0.48 Pursue with conditions Causal genetics, antibody tractability, a clean safety profile and a coherent cell-type mechanism support the case. The open question is differentiation: three antibodies against the target are already in Phase 3 in ulcerative colitis. ### 6Cell-Type Expression How to read this Who makes TL1A: TNFSF15 expression by cell type Mean TPM in single-cell reference data · selected gut and immune cell types Macrophage · immune176Colonocyte · rectum57Enteroendocrine cell · rectum55Classical monocyte · blood46M cell of gut · rectum42M2 macrophage · immune41Goblet cell · gut39Colon macrophage · colon27Median across 269 measured cell types: 7.9 TPM Who responds: expression of its receptor DR3 (TNFRSF25) Mean TPM in single-cell reference data Group 3 innate lymphoid cell · immune294Innate lymphoid cell · immune268Regulatory T cell · immune156Effector memory CD4 T cell · immune148T-helper 17 cell · rectum143Group 2 innate lymphoid cell · immune114Tuft cell of colon · colon84Group 1 innate lymphoid cell · immune70Median across 621 measured cell types: 10.3 TPM Tissues with TNFSF15 (HPA)17 Tissues with high expression0 Highest tissue level8.9 nTPM Tissue specificity0.85 Made by myeloid and gut-lining cells, sensed by ILC3 and Th17 cells **TNFSF15** is highest in macrophages (176 TPM, 22× the cell-type median) and in colon epithelium. Its receptor DR3 is highest in group 3 innate lymphoid cells, regulatory T cells and rectal Th17 cells. That is a ligand–receptor pair spanning the two cell compartments that drive gut inflammation. At tissue level the gene is restricted: detected in 17 tissues and high in none. DATA SOURCE Single-cell reference atlas and Human Protein Atlas Cell-type expression from the platform single-cell reference (CELLxGENE); tissue expression from HPA normal tissue. ### 7Differential Expression How to read this TNFSF15 in inflamed bowel tissue log2 fold change, disease vs normal, bulk tissue studies (Expression Atlas) Active UC, colon · E-MEXP-2083, p = 0.004+2.5UC, ascending colon · E-MTAB-5464, p = 0.046+1.2Active Crohn's, colon · E-MEXP-2083, p = 0.019+1.6log2 fold change of 2.5 is about 5.7 times higher in active UC Cell types that switch TNFSF15 on in Crohn's ileum Single-cell differential expression, significant cell types (q \< 0.05). No UC single-cell dataset in this release. | Cell type | Tissue | log2 FC | q value | Cells expressing, disease vs normal | | --- | --- | --- | --- | --- | | Enterocyte | ileal mucosa | +0.89 | 0.004 | 61% vs 0% | | Macrophage | ileum | +0.40 | 0.043 | 32% vs 0% | | B cell | ileum | +0.65 | 0.026 | 43% vs 0% | | Goblet cell | ileum | +0.62 | 0.046 | 43% vs 0% | | CD4 T cell | ileum | +0.56 | 0.011 | 32% vs 0% | | Transit amplifying cell | ileum | −0.36 | 0.046 | 0% vs 32% | Switched on in inflamed colon **TNFSF15** is raised in two independent ulcerative colitis biopsy studies, up to about 5.7-fold in active disease. In Crohn's ileum it is switched on in enterocytes, macrophages and lymphocytes that do not express it in healthy tissue. Higher expression in disease fits the genetic result: in gut tissues, genetically higher TNFSF15 raises risk. DATA SOURCE Expression Atlas and single-cell DE Bulk studies E-MEXP-2083 and E-MTAB-5464; single-cell differential expression from the platform scRNA disease atlas. ### 8Mechanism How to read this From risk gene to disease: the TL1A–DR3 axis Built from platform evidence: cell-type expression, protein interactions (BioGRID, STRING), Reactome and GO pathways, Mendelian randomisation PRODUCING CELLSLIGANDRECEPTORSIGNALLINGRESPONDING CELLSMyeloid cellsMacrophage 176 TPMClassical monocyte 46Colon macrophage 27Gut epitheliumColonocyte 57 TPMM cell of gut 42Up 5.7× in active UCTNFSF15 (TL1A)TNF superfamily ligandGWAS + MR in UCBiology 99.8DR3 (TNFRSF25)Signalling receptorBioGRID + STRING 0.999DcR3 (TNFRSF6B)Decoy receptorSoaks up TL1ANF-κB activationTNFs bind their receptors(Reactome)Partners RELA, RELInnate lymphoid cellsILC3 294 TPM DR3ILC2 114 · ILC1 70All ILCs 268T cellsTreg 156 TPM DR3Th17 (rectum) 143Effector memory CD4 148Ulcerative colitisGut-tissue genetic instruments: higher TNFSF15 raises risk · IBD risk genes NOD2, LACC1 and TNFRSF14 sit one step away in the interaction network Interaction partners from curated databases Direct TNFSF15 partners, combined score | Partner | Role | Score | Source | | --- | --- | --- | --- | | TNFRSF25 (DR3) | Signalling receptor | 0.999 | BioGRID, STRING | | TNFRSF6B (DcR3) | Decoy receptor | 0.999 | BioGRID, STRING | | NOD2 | IBD risk gene | 0.770 | STRING | | TNFRSF14 (HVEM) | IBD risk gene, TNF receptor family | 0.758 | STRING | | LACC1 | IBD risk gene | 0.728 | STRING | A coherent ligand–receptor axis The data line up into one mechanism: macrophages and gut epithelium make **TNFSF15**, it binds DR3 on innate lymphoid cells and T cells, and DR3 signals through NF-κB. A decoy receptor, DcR3, competes for the ligand. Three other IBD risk genes sit next to TNFSF15 in the interaction network, which places it within the genetic architecture of inflammatory bowel disease. DATA SOURCE Platform evidence layers Single-cell expression, BioGRID and STRING interactions, Reactome and GO pathway membership, Mendelian randomisation. ### 9Protein Networks How to read this Protein interaction network (platform knowledge graph, two-step neighbourhood) 1 hop2 hopsTNFSF15IRF8MAZRELAPOLR2FCEBPATNFRSF6BTNFRSF25RELIRF1STAT1STAT2HLFNR1H4FOXI1FOXL1IRF7PRKCBCXCR2IL23A Selected geneDirect interactor & disease seedDisease seedApproved drug targetClinical drug targetReal 1-hop interactor Network proximity score0.25 Direct interactors16 Disease anchor genes200 Anchors 1 step away1 Anchors 2 steps away42 Shortest path to an anchor1 step Median path to an anchor2 Close to the ulcerative colitis gene set **TNFSF15** sits one step from the genes that anchor **ulcerative colitis** in the platform network, with 42 anchor genes within two steps. Its 16 direct partners include its signalling receptor **TNFRSF25** (DR3) and its decoy receptor **TNFRSF6B** (DcR3), alongside NF-κB and interferon-regulatory transcription factors. DATA SOURCE Platform knowledge graph Direct and two-step interaction partners, disease anchor genes and drug-target anchors, snapshot 11 June 2026. ### 10Phenotypes How to read this Closest phenotype match: primary sclerosing cholangitis TNFSF15 phenotype profile versus disease profiles, weighted by how specific each sign is Shared signs18 Specificity-weighted score203.7 Most specific shared signIC 15.4 | Shared sign | Specificity (IC) | Frequency in PSC | | --- | --- | --- | | Abnormal biliary tract morphology | 15.4 | Very frequent (80–99%) | | Coeliac disease | 13.9 | Occasional (5–29%) | | Liver carcinoma | 13.5 | Occasional (5–29%) | | Hepatitis | 12.9 | Occasional (5–29%) | | Portal hypertension | 12.2 | Frequent (30–79%) | Most common signs across TNFSF15-linked diseases Number of linked diseases showing each sign (677 diseases in total) Hepatomegaly211Splenomegaly190Fatigue101Jaundice83Cirrhosis73Recurrent infection73Autoimmunity69Ascites61 The phenotype profile points to the gut–liver axis Ulcerative colitis has no curated phenotype profile in HPO, so a direct match is not computed. The diseases linked to **TNFSF15** share a liver and immune signature, and the single closest match is primary sclerosing cholangitis: the bile-duct disease that occurs alongside inflammatory bowel disease, usually ulcerative colitis, in most patients. DATA SOURCE Human Phenotype Ontology Curated HPO annotations; signs weighted by information content so that specific signs count more than generic ones. ### 11Literature How to read this TNFSF15 × ulcerative colitis co-mention trend (2003 – 2026) PubMed papers that mention both TNFSF15 and ulcerative colitis 05102003200520102015202020252026Landmark (1,488 citations) Total90 20265 Reviews25 Most-cited papers | Year | Journal | Cited by | Title | | --- | --- | --- | --- | | 2015 | Nature genetics | 1,488 | Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations. | | 2011 | Nature genetics | 716 | Meta-analysis identifies 29 additional ulcerative colitis risk loci, increasing the number of confirmed associations to 47. | | 2016 | Nature genetics | 412 | Analysis of five chronic inflammatory diseases identifies 27 new associations and highlights disease-specific patterns at shared loci. | | 2009 | Gut | 322 | Crohn's disease: Th1, Th17 or both? The change of a paradigm: new immunological and genetic insights implicate Th17 cells in the pathogenesis of Crohn's disease. | | 2014 | World journal of gastroenterology | 270 | Immunopathology of inflammatory bowel disease. | | 2014 | The Journal of experimental medicine | 249 | CX₃CR1⁺ mononuclear phagocytes support colitis-associated innate lymphoid cell production of IL-22. | | 2005 | Human molecular genetics | 173 | Single nucleotide polymorphisms in TNFSF15 confer susceptibility to Crohn's disease. | | 2015 | Nature medicine | 149 | Meta-analysis of shared genetic architecture across ten pediatric autoimmune diseases. | | 2010 | Gastroenterology | 144 | Imbalance of NKp44(+)NKp46(-) and NKp44(-)NKp46(+) natural killer cells in the intestinal mucosa of patients with Crohn's disease. | | 2003 | Journal of immunology (Baltimore, Md. : 1950) | 133 | Expression, localization, and functional activity of TL1A, a novel Th1-polarizing cytokine in inflammatory bowel disease. | | 2010 | Human molecular genetics | 100 | Comparative genetic analysis of inflammatory bowel disease and type 1 diabetes implicates multiple loci with opposite effects. | | 2017 | Gastroenterology | 96 | Genome-Wide Association Study Identifies African-Specific Susceptibility Loci in African Americans With Inflammatory Bowel Disease. | | 2010 | Inflammatory bowel diseases | 71 | Genetic predictors of medically refractory ulcerative colitis. | | 2004 | Clinical immunology (Orlando, Fla.) | 68 | Potential role for TL1A, the new TNF-family member and potent costimulator of IFN-gamma, in mucosal inflammation. | | 2013 | Gut | 68 | Mucosal CXCR4+ IgG plasma cells contribute to the pathogenesis of human ulcerative colitis through FcγR-mediated CD14 macrophage activation. | | 2023 | Digestion | 60 | New and Emerging Treatments for Inflammatory Bowel Disease. | | 2012 | World journal of gastroenterology | 59 | Role of ATG16L, NOD2 and IL23R in Crohn's disease pathogenesis. | | 2012 | Cytokine | 59 | New insights into the dichotomous role of innate cytokines in gut homeostasis and inflammation. | | 2008 | Current gastroenterology reports | 54 | Recent advances in IBD pathogenesis: genetics and immunobiology. | | 2019 | Cells | 53 | Genetic Studies of Inflammatory Bowel Disease-Focusing on Asian Patients. | Top-5 papers carry 56% of total citations. Anchored in the large IBD genetics studies 90 papers mention **TNFSF15** together with **ulcerative colitis**. The most cited are the genome-wide association studies of 2011 to 2016 that placed TNFSF15 among the shared risk genes for inflammatory bowel disease. Papers from 2003 and 2004 first described TL1A, the TNFSF15 protein, in gut inflammation. DATA SOURCE PubMed co-mention and citation index A paper counts only when both the gene and the disease are linked to it. ### 12Clinical Trials How to read this Trial timeline: TNFSF15 and closely related targets in ulcerative colitis Source: ClinicalTrials.gov (AACT registry), snapshot 10 June 2026 TNFSF15 — Sanofi Duvakitug TNFSF14 — Avalo Therapeutics, Inc. MDGN-002 TNFSF15 — Merck Sharp & Dohme LLC Tulisokibart TNFSF15 — Merck Sharp & Dohme LLC IV Tulisokibart TNFSF15 — Teva Branded Pharmaceutical Products R&D LLC TEV-48574 Dose Regimen A P3P3P1×P3P1P3P22017202020242026 TNFSF15 trials9 Related-target trials1 Active (any phase)5 Completed4 Discontinued1 Highest phasePhase 3 Related-target highest phase Phase 1 DATA SOURCE ClinicalTrials.gov via the AACT registry Trials linked to the gene and disease through drug-target mappings, refreshed weekly. Clinically validated, not yet approved Five antibodies against **TNFSF15** (TL1A) are in **ulcerative colitis** trials. Three are in Phase 3: tulisokibart (Merck), duvakitug (Sanofi and Teva) and afimkibart (Roche). On 22 June 2026, after this data snapshot, Merck reported that tulisokibart met its Phase 3 primary endpoint in ulcerative colitis. ### 13Protein Structure How to read this Domain architecture · TNFSF15 (251 aa) Tumour necrosis factor domain Length251 aa PDB2RJL AlphaFoldAF-O95150-F1 Protein 2D structure Annotated sequence features and experimental proteomic data in the context of protein topology — image generated by Protter (Omasits et al., Bioinformatics 2014). generated using ProtterExtracellularCytoplasmicMet1Met1Ala2Ala2Glu3Glu3Asp4Asp4Leu5Leu5Gly6Gly6Leu7Leu7Ser8Ser8Phe9Phe9Gly10Gly1010Glu11Glu11Thr12Thr12Ala13Ala13Ser14Ser14Val15Val15Glu16Glu16Met17Met17Leu18Leu18Pro19Pro19Glu20Glu2020His21His21Gly22Gly22Ser23Ser23Cys24Cys24Arg25Arg25Pro26Pro26Lys27Lys27Ala28Ala28Arg29Arg29Ser30Ser3030Ser31Ser31Ser32Ser32Ala33Ala33Arg34Arg34Trp35Trp35Ala36: Transmembrane domainAla36: Transmembrane domainLeu37: Transmembrane domainLeu37: Transmembrane domainThr38: Transmembrane domainThr38: Transmembrane domainCys39: Transmembrane domainCys39: Transmembrane domainCys40: Transmembrane domainCys40: Transmembrane domain40Leu41: Transmembrane domainLeu41: Transmembrane domainVal42: Transmembrane domainVal42: Transmembrane domainLeu43: Transmembrane domainLeu43: Transmembrane domainLeu44: Transmembrane domainLeu44: Transmembrane domainPro45: Transmembrane domainPro45: Transmembrane domainPhe46: Transmembrane domainPhe46: Transmembrane domainLeu47: Transmembrane domainLeu47: Transmembrane domainAla48: Transmembrane domainAla48: Transmembrane domainGly49: Transmembrane domainGly49: Transmembrane domainLeu50: Transmembrane domainLeu50: Transmembrane domain50Thr51: Transmembrane domainThr51: Transmembrane domainThr52: Transmembrane domainThr52: Transmembrane domainTyr53: Transmembrane domainTyr53: Transmembrane domainLeu54: Transmembrane domainLeu54: Transmembrane domainLeu55: Transmembrane domainLeu55: Transmembrane domainVal56: Transmembrane domainVal56: Transmembrane domainSer57Ser57Gln58Gln58Leu59Leu59Arg60Arg6060Ala61Ala61Gln62Gln62Gly63Gly63Glu64Glu64Ala65Ala65Cys66Cys66Val67Val67Gln68Gln68Phe69Phe69Gln70Gln7070Ala71Ala71Leu72Leu72Lys73Lys73Gly74Gly74Gln75Gln75Glu76Glu76Phe77Phe77Ala78Ala78Pro79Pro79Ser80Ser8080His81His81Gln82Gln82Gln83Gln83Val84Val84Tyr85Tyr85Ala86Ala86Pro87Pro87Leu88Leu88Arg89Arg89Ala90Ala9090Asp91Asp91Gly92Gly92Asp93Asp93Lys94Lys94Pro95Pro95Arg96Arg96Ala97Ala97His98His98Leu99Leu99Thr100Thr100100Val101Val101Val102Val102Arg103Arg103Gln104Gln104Thr105Thr105Pro106Pro106Thr107Thr107Gln108Gln108His109His109Phe110: variantsPhe110: variants110Lys111Lys111Asn112Asn112Gln113Gln113Phe114Phe114Pro115Pro115Ala116Ala116Leu117Leu117His118His118Trp119Trp119Glu120Glu120120His121His121Glu122Glu122Leu123Leu123Gly124Gly124Leu125Leu125Ala126Ala126Phe127Phe127Thr128Thr128Lys129Lys129Asn130Asn130130Arg131Arg131Met132Met132Asn133: PTMsAsn133: PTMsTyr134Tyr134Thr135Thr135Asn136Asn136Lys137Lys137Phe138Phe138Leu139Leu139Leu140Leu140140Ile141Ile141Pro142Pro142Glu143Glu143Ser144Ser144Gly145Gly145Asp146Asp146Tyr147Tyr147Phe148Phe148Ile149Ile149Tyr150Tyr150150Ser151Ser151Gln152Gln152Val153Val153Thr154Thr154Phe155Phe155Arg156Arg156Gly157Gly157Met158Met158Thr159Thr159Ser160Ser160160Glu161Glu161Cys162: disulfide bondsCys162: disulfide bondsSer163Ser163Glu164Glu164Ile165Ile165Arg166Arg166Gln167Gln167Ala168Ala168Gly169Gly169Arg170Arg170170Pro171Pro171Asn172Asn172Lys173Lys173Pro174Pro174Asp175Asp175Ser176Ser176Ile177Ile177Thr178Thr178Val179Val179Val180Val180180Ile181Ile181Thr182Thr182Lys183Lys183Val184Val184Thr185Thr185Asp186Asp186Ser187Ser187Tyr188Tyr188Pro189Pro189Glu190Glu190190Pro191Pro191Thr192Thr192Gln193Gln193Leu194Leu194Leu195Leu195Met196Met196Gly197Gly197Thr198Thr198Lys199Lys199Ser200Ser200200Val201Val201Cys202: disulfide bondsCys202: disulfide bondsGlu203Glu203Val204Val204Gly205Gly205Ser206Ser206Asn207Asn207Trp208Trp208Phe209Phe209Gln210Gln210210Pro211Pro211Ile212Ile212Tyr213Tyr213Leu214Leu214Gly215Gly215Ala216Ala216Met217Met217Phe218Phe218Ser219Ser219Leu220Leu220220Gln221Gln221Glu222Glu222Gly223Gly223Asp224Asp224Lys225Lys225Leu226Leu226Met227Met227Val228Val228Asn229: PTMsAsn229: PTMsVal230Val230230Ser231Ser231Asp232Asp232Ile233Ile233Ser234Ser234Leu235Leu235Val236Val236Asp237Asp237Tyr238Tyr238Thr239Thr239Lys240Lys240240Glu241Glu241Asp242Asp242Lys243Lys243Thr244Thr244Phe245Phe245Phe246Phe246Gly247Gly247Ala248Ala248Phe249Phe249Leu250Leu250250Leu251Leu251251Binding SiteActive siteTransmembrane domainPTMsvariantsdisulfide bondssignal peptideN-term: UniProtTMRs: UniProtTNF15\_HUMANMAEDLGLSFGETASVEMLPEHGSCRPKARSSSARWALTCCLVLLPFLAGLTTYLLVSQLRAQGEACVQFQALKGQEFAPSHQQVYAPLRADGDKPRAHLTVVRQTPTQHFKNQFPALHWEHELGLAFTKNRMNYTNKFLLIPESGDYFIYSQVTFRGMTSECSEIRQAGRPNKPDSITVVITKVTDSYPEPTQLLMGTKSVCEVGSNWFQPIYLGAMFSLQEGDKLMVNVSDISLVDYTKEDKTFFGAFLL Protein 3D structure Experimental structure from the Protein Data Bank, where one is available (Berman et al., Nucleic Acids Res. 2000). Crystal structure available: PDB 2RJL (X-ray, 2.0 Å) Antibody-tractable: accessible extracellular surface 78% of the **TNFSF15** protein lies outside the cell, giving antibodies a large, accessible surface. A high-resolution crystal structure (PDB 2RJL) supports epitope work. ### 14Perturbation How to read this TNFSF15 knockdown versus ulcerative colitis expression signatures 23 disease signatures (21 UC, 1 combined IBD, 1 colitis) Favours inhibitionFavours activation-0.2-0.10+0.1+0.2Concordance score per disease signature · coloured if p \< 0.05 · large dot = combined IBD signature Signatures tested23 Favour inhibition (p \< 0.05)3 Favour activation (p \< 0.05)3 Strongest signal−0.20 · IBD, p = 0.002 Significant after correction0 Knockdown data do not settle direction here The strongest single results favour inhibition, led by the combined IBD signature (−0.20, p = 0.002), but individual UC datasets split both ways and none survive multiple-testing correction. That is expected for **TNFSF15**: it is a ligand passed between immune cells, and knocking it down in a single cell line is a weak model of that. Direction on this page comes from the genetics and the disease expression data. DATA SOURCE Perturbation concordance Gene knockdown signatures compared with disease expression signatures; p values per signature, with false-discovery correction across all genes. ### 15Experimental Models How to read this Recommended experimental system Matched to the cell types that make and sense TL1A | Role | Model | Evidence | | --- | --- | --- | | **TL1A source** | Primary classical monocytes, monocyte-derived macrophages | TNFSF15 46 and 176 TPM | | **Responders** | ILC3, Th17 and regulatory T cells from blood or gut biopsies | DR3 294, 143 and 156 TPM | | **Barrier** | Colon organoids (colonocytes, M cells) | TNFSF15 57 and 42 TPM; up in active UC | | **Co-culture readout** | Myeloid–lymphoid co-culture with TL1A blockade | Tests the full ligand–receptor axis | Cell lines expressing TNFSF15 Cancer cell line expression (CCLE), TPM · gut and lymphoid lineages, 984 lines screened | Cell line | Lineage | TNFSF15 TPM | | --- | --- | --- | | SNU-C2A | Colon adenocarcinoma | 97.8 | | SNU-C5 | Colon adenocarcinoma | 44.6 | | DB | Diffuse large B-cell lymphoma | 13.1 | | HD-MY-Z | Hodgkin lymphoma | 11.8 | | TO 175.T | Hodgkin lymphoma | 10.7 | | KARPAS-422 | Diffuse large B-cell lymphoma | 10.3 | Cancer cell dependencyNone (mean effect 0.03) Loss-of-function constraintpLI 0.00 Use primary immune co-cultures, not tumour lines **TNFSF15** works between cells: myeloid and epithelial cells make it, ILC3 and T cells respond to it. The most informative models pair those cell types, with colon organoids for the barrier side. Tumour lines are useful only as a TL1A source (SNU-C2A expresses 97.8 TPM), and because the gene is not a survival dependency, assays should read out signalling and cytokine release rather than cell death. DATA SOURCE Platform cell-type and cell-line data Single-cell reference expression; CCLE expression and DepMap dependency for 984 candidate lines. From a Disease Atlas by Euretos target–disease report. Scores from the 11 June 2026 data snapshot; trial data from ClinicalTrials.gov. ## How to read this page What this page answers In one screen: is this gene a good drug target for this disease, why, and what should you look at next? How to read it - **Title and IDs**: the gene and disease, with the gene's database identifiers (Ensembl, UniProt, HGNC). The banner below gives the disease confidence level and the date of the data snapshot. - **Score strip**: six values, each 0–100 except confidence. - **Biology**: how strongly the evidence says this gene causes the disease, as a rank among all genes for the disease (100 = top). - **Translational**: how plausible it is to turn that biology into a medicine for this disease. - **Tractability**: how feasible it is to make a drug against the target. - **Safety**: higher = safer. - **Priority**: the combined score used to rank targets. - **Confidence**: High, Medium or Low, reflecting how much data exists for this disease. - **Biology drivers (0–1)**: how strongly the biology score is supported by genetics, the protein network, cell-type expression, and evidence transferred from similar genes. Each driver is scored on its own, so the four values do not add up to 1. - **Score explanations**: one line per axis saying what drives it. - **Recommended posture**: the overall recommendation, here "Pursue with conditions". - **Context pills**: best modality, competition level, validation (the most advanced clinical stage reached), an opportunity note summarising the programmes in this disease, evidence level, and recommended direction (Inhibit or Activate). - **Key findings**: a short set of statements, each a headline plus one or two sentences, covering the axes that carry the case. Green = supports the target, amber = a caution or open question. Interpreting the result - Start with Biology and its drivers: a high score driven by genetics is the strongest starting point. For TNFSF15 the genetic driver is 1.00. - Priority multiplies the axes, so one weak axis pulls it down. Look for the lowest axis to find the main risk. - "Pursue with conditions" means the evidence supports the target and names what to resolve; the condition appears in the key findings (here, three antibodies already in Phase 3). - Direction tells you what a drug should do. Check it against the Genetic Evidence panel and the drugs already in trials. - Each key finding draws on one panel below; open that panel before relying on it. Methods Each axis is computed separately from its own evidence and then combined multiplicatively into priority. Key findings are generated by fixed rules from the panel data, not written by a language model, so the same data always gives the same text. Data sources - All panels below; each panel's guide lists its sources. The snapshot date in the banner applies to the scores. Definitions | Term | Meaning | | --- | --- | | Percentile | Rank among all genes scored for this disease, 0–100 | | Modality | The type of drug: antibody, small molecule, degrader, oligonucleotide | | Programme | One drug against this target, counted once however many trials it has | | Direction | Whether a drug should lower (Inhibit) or raise (Activate) the target's activity | | Posture | The page's overall recommendation | Limits - Scores rank options against each other; they are not probabilities of clinical success. - Clinical news after the snapshot date is not in the scores; it may appear in the Clinical Trials panel text. - The page assesses the target in one disease; the same gene can score very differently in another. ## How to read: Genetic Evidence What this panel answers Does human genetics say this gene causes the disease, and should a drug raise or lower its activity? How to read it - **Biology score (0–100)**: where the gene ranks among all genes scored for this disease. 99.8 means it ranks above 99.8% of them. - **Genetic driver**: how much of the biology score comes from genetic evidence rather than network or expression evidence. - **Recommended action**: Inhibit or Activate, taken from genetic instruments, with a note on whether the disease-relevant tissues support it. **Direction by tissue** summarises which tissues point to higher and which to lower risk. - **Forest plot**: one row per tissue. The dot is the estimated effect of genetically higher expression on disease risk; the bar is its 95% confidence interval. Right of zero = higher expression raises risk; left of zero = lowers it. Point size reflects the number of genetic variants used. - **Row labels** show the number of genetic variants used (n) and instrument strength (F). Colour shows the directionality check: Steiger-correct, or a reverse-causation candidate. - **Supporting evidence** lists fine-mapping results and the type of risk variants (coding or regulatory). Interpreting the result - Strong: high biology score, dots in disease-relevant tissues on one side of zero, intervals not crossing zero, F above 10. - A split between tissues is common and informative: trust the tissues closest to the disease. For TNFSF15 in ulcerative colitis, digestive-tract tissues (stomach, oesophageal mucosa, pancreas) are risk-raising while blood and skin point the other way. - Regulatory-only risk variants mean the disease is driven by how much protein is made, which suits blocking or lowering approaches. All seven recorded TNFSF15 variants are regulatory. - A reverse-causation flag on a row means the disease may change expression rather than the reverse; give that row less weight. Methods We use Mendelian randomisation: common variants that change a gene's expression in a tissue act as a natural experiment. Comparing their effect on expression with their effect on disease risk estimates the causal effect of the gene. Each tissue is analysed separately, and a directionality test checks that the variants act on the gene first. Data sources - Genome-wide association studies for the disease (public summary statistics, including FinnGen) - GTEx expression-QTLs per tissue - Open Targets Genetics fine-mapping and locus-to-gene scores Definitions | Term | Meaning | | --- | --- | | Mendelian randomisation (MR) | Using inherited variants as a natural experiment to estimate cause and effect | | eQTL | A genetic variant that changes how much of a gene is expressed | | β (effect size) | Change in disease risk per unit of genetically higher expression | | F-statistic | Strength of the genetic instrument; above 10 is considered reliable | | Steiger test | Checks that variants affect the gene before the disease, not the reverse | | Credible set | The smallest set of variants likely to contain the true causal variant | | Locus-to-gene score | Probability (0–1) that a risk locus acts through this gene | Limits - MR estimates the effect of lifelong, modest changes in expression; a drug gives a larger, shorter change. - Expression in a reference tissue is not always expression in the diseased cell type. - No genetic signal does not mean no role: the gene may lack strong common variants. ## How to read: Developability What this panel answers Which kind of drug can reach this target, and how hard will it be to make one? How to read it - **Top chips**: best modality, antibody score (0–100 with its rating), small-molecule status and developability risk. - **Modality fit table**: each relevant format rated HIGH, MEDIUM or LOW, with the main reason. Formats that do not suit the target's biology are marked "not the goal". - **Component bars (0–1)**: what drives the antibody score: surface accessibility, selectivity, extracellular domain size and approved-drug precedent. - **Stat chips** under the bars: extracellular fraction, membrane protein, number of close paralogs, selectivity risk. - For enzymes and kinases the panel shows chemistry instead: known ligands, best potency and selectivity across related proteins. Interpreting the result - Antibody HIGH needs an accessible extracellular region; TNFSF15 has 78% of its protein outside the cell. - Small-molecule fit rises with known ligands and potent ones (pChEMBL 7 or more, about 100 nM). TNFSF15 has none. - Many close paralogs raise the risk of off-target binding; a few, as here (two, selectivity risk low), are manageable. - A precedent component of 0 only means no drug against this target is approved yet; it does not mean the target is undruggable. Methods Each format has its own score built from protein features (location, topology, domain size, pockets), chemistry (known ligands and potency) and precedent (approved or clinical drugs on the target). Scores are reported on one 0–100 scale and combined into the overall tractability score in the header. Data sources - UniProt (location, topology, sequence) - ChEMBL (ligands, potency, target class) - Structural data from PDB and AlphaFold - Drug and trial records for precedent Definitions | Term | Meaning | | --- | --- | | Modality | The type of drug: antibody, small molecule, degrader, oligonucleotide, cell therapy | | Extracellular fraction | Share of the protein sequence outside the cell, reachable by antibodies | | Paralog | A related gene in the same family; risk of cross-reactivity | | pChEMBL | Potency on a log scale; 6 = 1 µM, 7 = 100 nM, 8 = 10 nM | | ADC / T-cell engager | Antibody formats that kill or remove target-bearing cells | Limits - Scores describe feasibility, not efficacy. - Soluble or shed forms of membrane proteins are not always annotated. - New chemistry (covalent binders, glues) can open targets that score LOW today. ## How to read: Safety What this panel answers What could go wrong if this target is drugged, and in which organs? How to read it - **Organ risk matrix**: six organs (liver, heart, kidney, lung, CNS, immune) by four evidence types. Each cell is LOW, MEDIUM or HIGH; a dash means no evidence of that type. - **Evidence types, left to right**: predicted (model), post-market reports, adverse-event data from drugs on the target, clinical observations. - **Overall risk**: combined rating with its score (0–1, higher = more risk). - **Tissue and cell-type specificity**: how narrowly the gene is expressed (0–1, higher = more restricted). - **pLI and LOEUF**: how well people tolerate losing one copy of the gene. Interpreting the result - Read the organ pattern, not just the overall rating. A single HIGH organ that matches the mechanism (immune for an immune target) is expected and manageable. - Restricted expression (specificity above about 0.8) lowers the chance of effects in unrelated tissues. TNFSF15 scores 0.85 for tissues and 0.94 for cell types. - LOEUF above about 0.6 and pLI near 0 mean loss of function is tolerated in the population; LOEUF below 0.35 or pLI near 1 means it is not. - Columns to the right carry more weight: post-market and clinical evidence beat predictions. With no approved TNFSF15 drug, only the predicted column is filled here. Methods Three independent axes are combined. Per-organ adverse-event prediction integrates pharmacovigilance, toxicology screens and mouse knockout phenotypes. Genetic essentiality uses population constraint and cancer cell-line dependency. Expression context uses tissue and cell-type specificity. For targets without approved drugs, only the predicted column can be filled. Data sources - FDA Adverse Event Reporting System (FAERS) - Toxicology databases including DILIrank - Mouse knockout phenotypes (IMPC, MGI) - gnomAD constraint (pLI, LOEUF), DepMap - GTEx and single-cell expression Definitions | Term | Meaning | | --- | --- | | pLI | Probability (0–1) that a gene is intolerant to losing one copy | | LOEUF | Observed vs expected loss-of-function variants; lower = less tolerant | | Tissue specificity (Tau) | 0 = expressed everywhere, 1 = expressed in one tissue only | | Post-market reports | Side effects reported for approved drugs acting on this target | Limits - Predictions for first-in-class targets have no clinical calibration yet. - On-target risk only: off-target effects depend on the specific molecule. - Tolerated loss of one copy does not guarantee safe full inhibition. ## How to read: Competition What this panel answers Who else is developing drugs against this target in this disease, how far they are, and where the genetics points that nobody is testing? How to read it - **Stat chips**: number of antibodies in trials for this disease, how many reached Phase 3, how many are approved, number of sponsors. - **Programme table**: one row per drug (not per trial): drug and code names, sponsor, highest phase in this disease, key trials with enrolment, and other indications the same drug is in trials for. - **"Where the genetics points next"**: other diseases where this target scores high on biology, next to the trials registered for them. Rows marked "None registered" are untested opportunities. Interpreting the result - Several Phase 3 programmes mean the target is validated in the clinic and the field is crowded; a new entrant needs a clear difference (format, dosing, patient group, indication). - No programmes plus strong biology suggests white space, but check why: the target may have failed before or be hard to drug. - High-scoring diseases without trials are candidate expansion indications. For TNFSF15, eosinophilic oesophagitis (99.7) and lupus (99.3) have no TL1A trials. - A drug's trials in other diseases show where its sponsor sees the biology going. Methods Trials are linked to the target through the drugs they test, then grouped per drug. A drug counts once per disease, at its highest phase. The genetics column comes from the platform's biology scores for the same target across all diseases. Data sources - ClinicalTrials.gov via the AACT database, refreshed weekly - Drug–target mappings from ChEMBL and curated antibody catalogues - Disease Atlas biology scores Definitions | Term | Meaning | | --- | --- | | Programme | One drug against one target, counted once regardless of its number of trials | | Highest phase | Most advanced clinical phase reached in this disease | | Biology percentile | The target's rank among all genes for that disease (0–100) | Limits - Only registered trials are visible; early discovery and many non-Western programmes are missing. - Trial registration lags announcements by weeks to months. - Programme counts here and trial counts in the Clinical Trials panel measure different things. ## How to read: Score Overview What this panel answers How does this target–disease pair score on each axis, and what is the overall recommendation? How to read it - **Radar chart**: five axes on a 0–100 scale. A full, round shape means no weak axis. - **Biology**: strength of causal evidence that the gene drives the disease, as a percentile within the disease. - **Translational**: how plausible it is to turn that biology into a medicine for this disease. - **Tractability**: how feasible it is to make a drug against the target (from the Developability panel). - **Safety**: higher = safer (from the Safety panel). - **Priority**: the combined score used to rank targets. - **Evidence confidence**: High, Medium or Low, with the disease confidence level. - **Biology drivers (0–1)**: how strongly each evidence type supports the biology score: genetic, network, expression, and transfer from similar genes. - **Posture card**: the overall recommendation (here "Pursue with conditions"), with the case for it and the open question. Interpreting the result - Priority multiplies the axes rather than averaging them, so one weak axis pulls the total down. A pair cannot score high on priority with poor biology. - A genetic driver close to 1 means the case rests on human genetics, the strongest evidence type for clinical success. TNFSF15 scores 1.00. - "Pursue with conditions" means the evidence supports the target but names an open question, for TNFSF15 the crowded Phase 3 field. - Confidence (High, Medium, Low) reflects how much data exists for the disease; long-tail diseases get lower confidence. Methods Each axis is computed from its own evidence. Biology combines causal genetics, protein-network position and cell-type expression, with a transfer component from similar genes when direct evidence is missing. Literature is used only to choose candidate genes; it is not a scoring signal. The axes are combined multiplicatively into priority. Data sources - All panels on this page; see each panel's guide. Definitions | Term | Meaning | | --- | --- | | Percentile | Rank among all genes scored for this disease, 0–100 | | Transfer | Evidence borrowed from genes with similar profiles, used when direct data is thin | | Posture | The page's overall recommendation | Limits - Scores rank options; they are not probabilities of clinical success. - Scores from different snapshots are not directly comparable. ## How to read: Cell-Type Expression What this panel answers In which cells is the gene switched on, and are those the cells that matter for the disease? How to read it - **Bars**: mean expression (TPM) of the gene in each cell type, from single-cell reference data. Muted text after the name shows the tissue of origin. - **Note under each chart**: the median across all measured cell types, to show how far above background the top cell types are. - **For ligands** (signalling molecules such as cytokines) the panel shows two charts: the cells that make the ligand, and the cells that carry its receptor and respond. For TNFSF15, macrophages make it (176 TPM) and ILC3, regulatory T and Th17 cells carry its receptor DR3. - **For cancer** the panel compares tumour with normal tissue and shows which part of the tumour (cancer cells, stroma, immune cells) expresses the gene. - **Stat chips**: number of tissues where the gene is detected, tissues with high expression, highest tissue level, tissue specificity. Interpreting the result - Good: the top cell types belong to the disease organ or its immune infiltrate, well above the median. TNFSF15 in macrophages is 22 times the cell-type median. - A ligand–receptor pair spanning two cell compartments (myeloid producers, lymphoid responders) points to a cell-to-cell mechanism that a blocking antibody can interrupt. - Expression restricted to few tissues supports safety; broad expression is a flag to check in the Safety panel. - High TPM in a small fraction of cells can still matter if those cells are the disease drivers. Methods Single-cell RNA data are grouped by cell type and averaged. Cell types are filtered to the organs the disease affects plus immune cells that infiltrate them. Tissue-level values come from bulk RNA data. Data sources - CELLxGENE single-cell reference atlas - Human Protein Atlas (normal tissues) - GTEx (tissue RNA) Definitions | Term | Meaning | | --- | --- | | TPM | Transcripts per million; a normalised measure of RNA level | | nTPM | TPM normalised across Human Protein Atlas samples | | Median across cell types | The middle expression value over all measured cell types; a reference for background level | | Ligand / receptor | Signalling molecule / the protein on the responding cell that it binds | Limits - RNA is not protein; secreted proteins act far from the cells that make them. - Reference data come mostly from healthy tissue; disease changes are in the Differential Expression panel. - Rare or activated cell states may be missing from the reference. ## How to read: Differential Expression What this panel answers Is the gene switched up or down in diseased tissue compared with healthy tissue, and in which cells? How to read it - **Bars**: one per study, showing the log2 fold change (disease vs normal). Right = higher in disease. The study ID and p-value sit next to the label. - **Cell-type table**: single-cell comparisons, with tissue, fold change, q-value and the share of cells expressing the gene in disease vs normal tissue. - When no single-cell data exist for the disease, the closest related disease is shown and labelled. Here there is no ulcerative colitis single-cell dataset, so the table shows Crohn's disease ileum. - **For cancer** the panel compares tumour with matched normal tissue (TCGA), by tumour subtype where available. Interpreting the result - Agreement across independent studies counts more than one large fold change. TNFSF15 is raised in two separate ulcerative colitis biopsy studies. - A log2 fold change of 1 = twice as high; 2.5 = about 5.7 times. - "61% vs 0% of cells" (disease vs normal) means the gene is switched on in disease in cells that do not use it normally, a strong signal of a disease-specific programme. - Up in disease plus risk-raising genetics in the same tissue (Genetic Evidence panel) is a consistent case for inhibition. Methods Bulk studies compare tissue biopsies from patients and controls. Single-cell comparisons are done per cell type, with correction for testing many genes (q-value). Only results with a named source study are shown. Data sources - EBI Expression Atlas (curated bulk studies, here E-MEXP-2083 and E-MTAB-5464) - Public single-cell disease datasets - TCGA and GTEx for cancer Definitions | Term | Meaning | | --- | --- | | log2 fold change | Ratio of disease to normal on a log2 scale; 0 = no change | | p-value | Chance of a difference this large with no real effect | | q-value | p-value corrected for testing thousands of genes; below 0.05 is significant | Limits - Expression change can be a consequence of inflammation rather than a cause; genetics speaks to cause. - Biopsy composition varies: more immune cells in a sample can raise immune genes without any change per cell. - A related-disease single-cell result is supporting context, not a measurement in this disease. ## How to read: Mechanism What this panel answers How does this gene plausibly cause the disease: which cells, which partners, which signalling route? How to read it - **Diagram**: read left to right. Each box names a step and shows the evidence for it (expression level, interaction score, pathway, genetic result). - **Layout depends on the target type.** For a ligand such as TNFSF15: producing cells → ligand → receptor → signalling → responding cells. For a receptor: its ligands → receptor → transcription factors → gene programme. For an enzyme, transcription factor, circulating protein, ion channel or cancer target, the columns change to match. - **The highlighted box** is the target. A decoy receptor (here DcR3) is shown alongside the signalling receptor. - **Bottom band**: the disease, with the genetic direction and nearby disease risk genes. - **Partner table**: each direct interaction partner with its role, score and source database. Interpreting the result - A coherent chain, where every step has its own evidence, is stronger than any single data point. For TNFSF15: macrophages and gut epithelium make it, it binds DR3 on innate lymphoid cells and T cells, DR3 signals through NF-κB. - Interaction scores near 1 are experimentally confirmed; around 0.7 often reflect co-mention or co-expression. - Known disease risk genes one step away (NOD2, LACC1, TNFRSF14 for TNFSF15) place the target inside the disease's genetic architecture. - A decoy receptor is a natural brake and can hint at how the body limits the pathway. Methods The chain is assembled from separate evidence layers: cell-type expression for producer and responder cells, curated protein interactions, pathway membership and the genetic results. Each step carries the source that supports it. When the target type cannot be determined, a generic chain (protein → pathway → cell function → tissue → symptoms) is shown instead. Data sources - BioGRID and STRING (protein interactions) - OmniPath and CellPhoneDB (ligand–receptor pairs) - Reactome and Gene Ontology (pathways and functions) - Single-cell reference atlas; genetic evidence from panel 1 Definitions | Term | Meaning | | --- | --- | | Ligand | A signalling molecule that binds a receptor on another cell | | Decoy receptor | Binds the ligand without signalling, soaking it up | | Interaction score | Confidence (0–1) that two proteins interact | | NF-κB | A central inflammatory signalling pathway | Limits - The diagram is a hypothesis built from evidence, not a validated pathway. - Steps without data are left out rather than guessed, so a short chain may reflect missing data. ## How to read: Protein Networks What this panel answers How close is this gene to the genes known to drive the disease, in the network of protein interactions? How to read it - **Graph**: the target sits in the centre, with its one- and two-step neighbourhood around it. Colours mark the role of each gene: the selected gene, disease seed genes (the disease's anchor genes), direct interactors that are also seeds, approved and clinical drug targets, and other direct interactors. - **Stat chips**: network proximity score, number of direct partners, number of disease anchor genes, how many anchors lie one or two steps away, shortest and median path to an anchor. - **Claim card**: names the partners that matter most, such as receptors or known disease genes. Interpreting the result - A shortest path of one step to an anchor gene, and many anchors within two steps, mean the target sits inside the disease neighbourhood. TNFSF15 has 42 anchors within two steps. - Neighbours that are approved or clinical drug targets show the neighbourhood is druggable and validated. - A proximity score above about 0.1 is in the top quarter of target–disease pairs; above about 0.04 is above median. TNFSF15 scores 0.25. - Network evidence supports genetics; it rarely carries a case alone. Methods The disease's anchor genes are those with the strongest independent evidence for the disease. A random walk from those anchors through the protein network measures how often the target is reached; the more often, the closer it sits. Path lengths are counted directly in the graph. Data sources - Platform knowledge graph built from curated interaction databases (including BioGRID, STRING, SIGNOR) - Disease anchor gene sets from genetics and curated sources - Drug-target records for approved and clinical anchors Definitions | Term | Meaning | | --- | --- | | Anchor (seed) gene | A gene with strong independent evidence for the disease | | Step (hop) | One direct protein–protein interaction | | Network proximity score | How often a random walk from the anchors reaches the target (0–1) | Limits - Well-studied proteins have more recorded interactions, so they look more central than they are. - Interactions are often measured outside the disease's cell types. - Proximity does not say whether the effect is up or down. ## How to read: Phenotypes What this panel answers Do the clinical signs linked to this gene match the signs of the disease, or of a closely related one? How to read it - **Match summary**: the disease whose signs best match the gene's profile, with the number of shared signs, a specificity-weighted score and the most specific shared sign. - **Shared-sign table**: each shared sign with its specificity (IC) and how often it occurs in patients with that disease. - **Bar chart**: the most common signs across all diseases linked to the gene. - **For inherited (Mendelian) diseases** the panel compares the gene's signs directly with the disease and is a core piece of evidence. - **For common diseases** without a curated symptom profile, it shows the closest related disease instead. For TNFSF15 that is primary sclerosing cholangitis, a bile-duct disease that most often occurs with ulcerative colitis. - **For cancer** the panel shows prognosis data instead. Interpreting the result - Specific shared signs count more than generic ones: "abnormal biliary tract morphology" (IC 15.4) says more than "fatigue". - A high specificity-weighted score with few, rare signs beats many common ones. - A related-disease match is context, not direct evidence for the pair on the page. Methods Genes and diseases are each described by sets of clinical signs from the Human Phenotype Ontology. Matches are scored by the shared signs, weighting each by how rare it is across all diseases, so specific signs count more. Data sources - Human Phenotype Ontology (HPO) annotations - OMIM and Orphanet (via HPO) Definitions | Term | Meaning | | --- | --- | | Phenotype / sign | An observable clinical feature, such as jaundice | | Information content (IC) | How specific a sign is; rare signs score higher | | Frequency band | How often a sign occurs in patients: occasional 5–29%, frequent 30–79%, very frequent 80–99% | Limits - About a quarter of human genes have curated sign annotations; an empty profile means no data, not no effect. - Profiles describe inherited syndromes best; common diseases are poorly described in HPO. ## How to read: Literature What this panel answers How much has been published on this gene in this disease, and which papers established the link? How to read it - **Trend chart**: papers per year that mention both the gene and the disease. The marked point is the most-cited (landmark) paper. - **Stat chips**: total papers, papers this year, reviews. - **Most-cited table**: year, journal, citation count (bar) and title, with the share of all citations held by the top five papers. Interpreting the result - A landmark genetics paper at the top (for TNFSF15, the 2011 and 2015 IBD genome-wide studies in Nature Genetics) shows the link came from unbiased discovery, not from following a hypothesis. - A rising trend after a quiet period often follows clinical news. - Few papers with strong genetics can mean an underexplored target; many papers with weak genetics can mean a popular but less causal one. Methods A paper counts only when it is linked to both the exact gene and the exact disease concept; broader or narrower disease terms are not added. Citation counts come from the platform's citation index. Literature is shown for context: it is not used in the biology or priority scores. Data sources - PubMed / MEDLINE - Platform citation index Definitions | Term | Meaning | | --- | --- | | Co-mention | A paper that names both the gene and the disease | | Landmark paper | The most-cited paper in the set | Limits - Counts measure attention, not truth; a popular gene is not necessarily a causal one. - Recent papers have had little time to collect citations. - Synonyms and older gene names may be missed (TNFSF15 is also published as TL1A). ## How to read: Clinical Trials What this panel answers Which clinical trials have tested drugs against this target, or closely related targets, in this disease, and how far did they get? How to read it - **Timeline**: one row per gene, sponsor and drug. Markers show each trial's phase (P1, P2, P3) at its start date; a cross marks a discontinued trial. - **Rows for related targets** (proteins one interaction away or in the same pathway) are labelled with their own gene name, here TNFSF14. - **Stat chips**: number of trials on the target, trials on related targets, active, completed and discontinued trials, highest phase reached on the target and on related targets. - **Claim card**: the current clinical position in plain words, including news after the data snapshot where relevant. Interpreting the result - Phase 3 trials mean the target has already passed early efficacy tests in patients. - Discontinued trials need a reason before you draw a conclusion: safety, efficacy and business reasons look the same here. - Trials on related targets show whether the pathway itself is being tested, even if this exact protein is not. - This panel counts trials; the Competition panel counts drugs. One drug often has several trials. Methods Trials are linked to the target through the drugs they test, using drug–target mappings. Related-target trials are added through protein interactions and shared pathways. Status and phase come directly from the registry. Data sources - ClinicalTrials.gov via the AACT database, refreshed weekly - Drug–target mappings (ChEMBL, curated antibody catalogues) - STRING and Reactome for related targets Definitions | Term | Meaning | | --- | --- | | Phase 1 / 2 / 3 | Safety / early efficacy and dose / confirmatory efficacy in large patient groups | | Discontinued | Terminated, withdrawn or suspended before completion | | Related target | A different protein that interacts with the target or shares its pathway | Limits - Registry records are often incomplete or late; results are rarely in the registry itself. - Trials outside ClinicalTrials.gov (some regional registries) may be missing. ## How to read: Protein Structure What this panel answers What does the protein look like, which parts are reachable by a drug, and is there a structure to design against? How to read it - **Domain architecture**: protein length and its functional domains (for TNFSF15, a tumour necrosis factor domain). - **Stat chips**: length in amino acids, experimental structure ID (PDB) and predicted structure ID (AlphaFold). - **2D topology**: the sequence drawn as beads across the cell membrane. The band is the membrane; beads above it are outside the cell, below are inside. Coloured beads mark binding sites, modifications, variants and disulfide bonds (see legend). - **3D structure**: the experimental structure, with its method and resolution (here PDB 2RJL, X-ray, 2.0 Å). - **Claim card**: what the structure means for drug design. Interpreting the result - A large domain outside the cell supports antibodies; a defined pocket supports small molecules. - Resolution below about 2.5 Å is good enough for structure-based design. - Coverage matters: a structure of only part of the protein may miss the region you want to target. - AlphaFold models are reliable where confidence (pLDDT) is above 70 and should be treated with care below that. Methods Domains and topology come from protein annotation databases; the 2D view is drawn with Protter. The best experimental structure is chosen by coverage and resolution; AlphaFold is shown when no experimental structure exists or as a complement. Data sources - UniProt and InterPro (domains, topology, features) - Protein Data Bank (experimental structures) - AlphaFold Protein Structure Database - Protter (2D topology drawing) Definitions | Term | Meaning | | --- | --- | | Domain | A self-contained, folded part of a protein with its own function | | Transmembrane region | The stretch that crosses the cell membrane | | Resolution (Å) | Level of detail in an experimental structure; lower is sharper | | pLDDT | AlphaFold's confidence per residue, 0–100 | Limits - Structures show one state; proteins move and change shape on binding. - Membrane proteins and complexes are often only partly resolved. ## How to read: Perturbation What this panel answers When this gene is knocked down in cells, does the cell move away from the disease state or towards it? How to read it - **Dot plot**: one dot per disease expression signature. Left of zero = knocking the gene down reverses the disease signature, which favours inhibiting the target. Right of zero = knockdown mimics the disease, which favours activating it. - **Coloured dots** are significant (p below 0.05); grey dots are not. A larger dot marks a combined signature across several studies (here the combined IBD signature). - **Stat chips**: signatures tested, how many favour inhibition or activation, the strongest signal, and how many survive correction for multiple testing. - **For cancer** this panel shows CRISPR knockout effects on tumour cell growth by lineage instead. Interpreting the result - Consistent dots on one side, ideally surviving correction, are a useful direction check alongside genetics. - A split pattern, as for TNFSF15 (three signatures each way, none surviving correction), means knockdown data cannot settle the direction; the panel says so. - Agreement with the genetic direction (panel 1) strengthens the case; disagreement is a reason to look closer, not to drop the target. Methods Gene-expression changes after knocking down the gene in cell lines are compared with gene-expression changes in patient tissue. A negative concordance score means the two run in opposite directions. Scores are tested per signature and corrected across all genes tested. Data sources - Public gene-knockdown expression screens (including LINCS L1000) - Disease expression signatures from public patient studies Definitions | Term | Meaning | | --- | --- | | Knockdown | Reducing a gene's expression in cells, e.g. with shRNA or CRISPR | | Disease signature | The set of genes up and down in patient tissue vs healthy tissue | | Concordance score | Agreement between knockdown and disease signatures; negative = knockdown reverses the disease | | Multiple-testing correction | Adjusting p-values because many tests are run at once | Limits - Knockdown happens in single cell lines, which often do not express the gene or its partners. - For signalling molecules that act between cells (ligands such as TNFSF15), knockdown in one cell type is a weak model of their biology. - Results describe expression programmes, not clinical outcomes. ## How to read: Experimental Models What this panel answers Which cells, tissues or cell lines should be used to test this target in the lab? How to read it - **Recommended system table**: the role each model plays (source of the target, responding cells, tissue barrier, combined read-out), the model, and the evidence behind it. - **Cell line table**: cell lines from the relevant lineages that express the gene, with lineage and expression level (TPM), out of all lines screened. - **Claim card**: the recommended set-up and the read-out to use, taking into account whether cell lines depend on the gene for survival. - **Stat chips**: whether cancer cell lines depend on the gene for survival (mean dependency effect), and its tolerance to loss of function (pLI). - **For cancer** the table ranks cell lines by lineage match, expression and dependency instead. Interpreting the result - For non-cancer diseases, primary cells and organoids that match the disease cell types are usually more informative than tumour lines. For TNFSF15: monocyte-derived macrophages as the source, ILC3, Th17 and regulatory T cells as responders, colon organoids for the gut barrier. - For ligands, a co-culture of producer and responder cells tests the whole axis in one assay. - If the gene is not a survival dependency, measure signalling or cytokine release, not cell death. - For cancer, a selective dependency (strong in the tumour lineage, weak elsewhere) marks lines where the target matters. Methods Recommendations combine the producer and responder cell types from the Cell-Type Expression panel with cell-line expression and dependency data. Cell lines are scored on lineage match to the disease, expression within that lineage and dependency. Data sources - Cancer Cell Line Encyclopedia (CCLE) expression - DepMap CRISPR dependency (Chronos) - Single-cell reference atlas (cell types) Definitions | Term | Meaning | | --- | --- | | Dependency (Chronos score) | Effect of knocking out the gene on cell growth; 0 = no effect, −1 = as strong as typical essential genes | | Organoid | A 3D mini-tissue grown from stem cells or biopsies | | Co-culture | Two or more cell types grown together to model their interaction | | Lineage | The tissue of origin of a cell line | Limits - Cell lines are mostly tumour-derived and drift from normal cell behaviour. - Expression in a line does not guarantee the receptor partners or pathway are present. - Primary-cell recommendations come from reference data; availability and donor variation need checking. ## Read the methodology. Run a target through the platform. The whitepaper covers every pipeline, every reference, and the limitations explicitly. Send us a target and indication you are working on, and we will run the assessment and send you the result. Download the methodology paper [Request an assessment](mailto:information@euretos.com?subject=Disease%20Atlas%20-%20Target%20assessment%20request) × ### Methodology & Validation, 2026 60 pages. Four pipelines. Every reference cited by PMID. Enter your details below and we will send the PDF directly. We do not sell or share your address. One follow-up at most. Disease Atlas | Euretos  ·  [information@euretos.com](mailto:information@euretos.com) ![](https://px.ads.linkedin.com/collect/?pid=3328428&conversionId=4680788&fmt=gif)