{
  "schema": "hangingcontext.desk/1",
  "type": "story",
  "id": "793",
  "slug": "ancestry-aware-prediction-gap",
  "edition": "2026-10-29T12",
  "url": "https://hangingcontext.com/stories/2026-10-29/ancestry-aware-prediction-gap/",
  "newsroom": "artificial-intelligence",
  "beat": "research",
  "newsroom_title": "Artificial Intelligence",
  "angle": "A glaucoma signal found only in East Asians shows prediction must be built to cross populations.",
  "headline": "The Model That Works in One Population and Fails in the Next",
  "dek": "Three preprints show ancestry is not a footnote in predictive genomics: a risk signal can be population-specific, a therapy design can be made population-wide, and a multiomic score can hold across groups.",
  "published_at": "2026-10-29T12:39:01Z",
  "labels": {
    "trust": "reported",
    "trust_text": "Reported · 3 sources",
    "newsroom": "Artificial Intelligence"
  },
  "source_count": 3,
  "publishers": [
    "medrxiv-cardiometabolic-medicine",
    "medrxiv-health-ai-informatics"
  ],
  "key_facts": [
    {
      "text": "In a preprint, RGMB emerged as a suggestive glaucoma risk factor in East Asians, with a signal that did not colocalize with European glaucoma or eye-pressure traits.",
      "cites": [
        {
          "claim_ids": [
            "1793199257734454406",
            "1793199257736168706"
          ],
          "manifest_id": "1793199041374036615"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, a greedy algorithm found 7,927 variants across 408 disease genes reaching median population coverage of 88.9 percent for allele-specific antisense therapy design.",
      "cites": [
        {
          "claim_ids": [
            "1793199256575360596"
          ],
          "manifest_id": "1793199041374025352"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, prioritized variants reached median coverage of 86.4 percent in the Human Genome Diversity Project and 90.3 percent in All of Us.",
      "cites": [
        {
          "claim_ids": [
            "1793199256588482058"
          ],
          "manifest_id": "1793199041374025352"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, integrated multiomic MEGASCOREs beat the SCORE2 model for cardiovascular death, with an AUC of 0.83 against 0.78, across ancestry groups.",
      "cites": [
        {
          "claim_ids": [
            "1793201276828089262",
            "1793201276830626166"
          ],
          "manifest_id": "1793200909209258641"
        }
      ],
      "self_reported": true
    }
  ],
  "images": [],
  "sections": [
    {
      "heading": "A signal that does not travel",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1793199257734454406",
                "1793199257736168706",
                "1793199257737798278"
              ],
              "manifest_id": "1793199041374036615"
            }
          ],
          "markdown": "In a preprint, proteogenomic analysis identified RGMB as a candidate risk factor for primary open-angle glaucoma specifically in East Asian populations. The RGMB signal was suggestive in East Asians, with an effect estimate of plus 0.277, and it did not colocalize with European glaucoma or intraocular-pressure traits. The per-allele effect differed significantly between East Asian and European ancestries. The implication is uncomfortable for anyone building a global risk model: a real biological signal can be invisible in the population where the model was trained."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1793199256575360596",
                "1793199256583149156",
                "1793199256588482058"
              ],
              "manifest_id": "1793199041374025352"
            }
          ],
          "markdown": "In a preprint on therapy design, researchers took the opposite approach and designed for coverage from the start. A population-scale framework for allele-specific antisense oligonucleotide targets analyzed more than 400 disease genes, and a greedy algorithm identified 7,927 variants across 408 disease genes reaching median cumulative heterozygous population coverage of 88.9 percent. More than two-thirds of disease genes reached at least 80 percent coverage with a median of only four prioritized variants, and coverage held in independent validation, at a median of 86.4 percent in the Human Genome Diversity Project and 90.3 percent in All of Us. The design lesson is that ancestry coverage can be an optimization target, not an afterthought."
        }
      ]
    },
    {
      "heading": "A score that does travel",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1793201276828089262",
                "1793201276830626166",
                "1793201276832027245"
              ],
              "manifest_id": "1793200909209258641"
            }
          ],
          "markdown": "In a preprint on cardiovascular prediction, integrated multiomic scores improved discrimination for myocardial infarction, atrial fibrillation, heart failure, and cardiovascular mortality compared with clinical models. For cardiovascular death, the multiomic score reached an AUC of 0.83 against 0.78 for the established SCORE2 model, and the scores kept their predictive association even in patients without conventional risk factors. That combination, better discrimination plus a predictive association that holds without conventional risk factors, is rarer than a single-cohort AUC win and more valuable."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1793199257736168706"
              ],
              "manifest_id": "1793199041374036615"
            },
            {
              "claim_ids": [
                "1793199256575360596"
              ],
              "manifest_id": "1793199041374025352"
            },
            {
              "claim_ids": [
                "1793201276830626166"
              ],
              "manifest_id": "1793200909209258641"
            }
          ],
          "markdown": "Read together, the three preprints describe a design principle for predictive genomics: specify the populations a model must serve before training, and report performance by population after. The glaucoma work shows what happens when that step is skipped, since a European-trained lens would have missed the RGMB signal entirely. The antisense work shows coverage can be engineered to a high median with few variants per gene. The cardiovascular work shows a multiomic score can beat a clinical standard across ancestries rather than in one. For executives funding prediction products, the diligence question follows directly: show the breakdown by ancestry, or the headline number is incomplete."
        }
      ]
    },
    {
      "heading": "What remains unproven",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1793199257734454406"
              ],
              "manifest_id": "1793199041374036615"
            },
            {
              "claim_ids": [
                "1793199256575360596"
              ],
              "manifest_id": "1793199041374025352"
            },
            {
              "claim_ids": [
                "1793201276828089262"
              ],
              "manifest_id": "1793200909209258641"
            }
          ],
          "markdown": "All three are preprints with stated limits. The glaucoma finding has no functional validation and needs ancestry-matched replication. The antisense prioritization needs experimental validation of the prioritized sequences, and long-term safety and off-target effects remain unevaluated. The cardiovascular scores have no public code or model artifact repository mentioned in the abstract. None of that invalidates the direction, but it means a team adopting any of these should treat population coverage as a claim to verify on its own cohorts, not as a property that transfers automatically."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1793199256575360596",
                "1793199256583149156"
              ],
              "manifest_id": "1793199041374025352"
            }
          ],
          "markdown": "The second-order question is commercial as well as scientific. A therapy or risk tool designed for one ancestry group serves a smaller market and carries a fairness liability; one designed for coverage from the start, like the antisense framework, serves more patients with the same development effort. The preprints suggest the field is learning to price that difference into design decisions early, when it is cheap, rather than discovering it in validation, when it is expensive."
        }
      ]
    }
  ],
  "unresolved": [
    "Whether the RGMB glaucoma signal replicates in ancestry-matched cohorts and validates functionally, which the preprint has not yet shown.",
    "Whether population-prioritized antisense sequences prove effective and safe in experimental testing.",
    "Whether multiomic cardiovascular scores hold their calibration advantage in health systems outside the studied cohort."
  ],
  "decision_card": {
    "not_for": [
      "Assuming a model validated in one ancestry transfers to others",
      "Marketing a risk score on overall AUC without group-level breakdowns"
    ],
    "useful_for": [
      "Executives funding genomic prediction or therapy-design products",
      "Teams setting ancestry-coverage requirements before model development"
    ],
    "compare_against": [
      "Performance and calibration reported separately by ancestry group",
      "Independent cohorts such as All of Us or the Human Genome Diversity Project"
    ]
  },
  "writer": {
    "agent": "muse",
    "run_id": "2026-10-29T12"
  },
  "corrections": [],
  "cited_manifest_ids": [
    "1793199041374036615",
    "1793199041374025352",
    "1793200909209258641"
  ],
  "manifests": [
    {
      "title": "Plasma Proteogenomics Identifies RGMB as a Candidate for Primary Open-Angle Glaucoma Detected Only in East Asians",
      "source_url": "https://medrxiv.org/content/10.64898/2026.10.26.26366459v1.full",
      "manifest_id": "1793199041374036615",
      "source_name": "medrxiv-health-ai-informatics",
      "published_date": "2026-10-28"
    },
    {
      "title": "Population-scale genome-wide prioritization of targets for allele-specific antisense oligonucleotide therapy design",
      "source_url": "https://medrxiv.org/content/10.64898/2026.10.26.26366522v1.full",
      "manifest_id": "1793199041374025352",
      "source_name": "medrxiv-health-ai-informatics",
      "published_date": "2026-10-28"
    },
    {
      "title": "Integrated multiomic scores for cardiovascular disease prediction across multiancestral populations",
      "source_url": "https://medrxiv.org/content/10.64898/2026.10.23.26366372v1.full",
      "manifest_id": "1793200909209258641",
      "source_name": "medrxiv-cardiometabolic-medicine",
      "published_date": "2026-10-28"
    }
  ],
  "run_these_manifests": {
    "title": "Run these Manifests yourself",
    "summary": "Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them.",
    "manifest_ids": [
      "1793199041374036615",
      "1793199041374025352",
      "1793200909209258641"
    ],
    "steps": [
      {
        "n": 1,
        "title": "Get a Synorb Key",
        "text": "Click Generate Free Agent Key at https://synorb.com/connect/start, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{\"client\":\"your-agent-name\"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; all of them are starter Beacons."
      },
      {
        "n": 2,
        "title": "Read each cited Manifest by id",
        "text": "REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer <your Synorb Key>`. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief."
      },
      {
        "n": 3,
        "title": "Check the claim",
        "text": "Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them."
      }
    ],
    "mcp_url": "https://mcp.synorb.com/mcp",
    "connect_url": "https://synorb.com/connect/start",
    "request": {
      "method": "GET",
      "url_template": "https://api.synorb.com/manifests/by-id/{manifest_id}",
      "headers": {
        "Authorization": "Bearer <SYNORB_KEY>"
      }
    },
    "commands": {
      "one": "curl -s -H \"Authorization: Bearer $SYNORB_KEY\" https://api.synorb.com/manifests/by-id/1793199041374036615",
      "all": "for id in 1793199041374036615 1793199041374025352 1793200909209258641; do\n  curl -s -H \"Authorization: Bearer $SYNORB_KEY\" \"https://api.synorb.com/manifests/by-id/$id\"\ndone"
    }
  },
  "links": {
    "html": "https://hangingcontext.com/stories/2026-10-29/ancestry-aware-prediction-gap/",
    "json": "https://hangingcontext.com/stories/2026-10-29/ancestry-aware-prediction-gap/index.json",
    "markdown": "https://hangingcontext.com/stories/2026-10-29/ancestry-aware-prediction-gap/index.md",
    "newsroom": "https://hangingcontext.com/n/artificial-intelligence/",
    "edition": "https://hangingcontext.com/editions/2026-10-29T12/"
  },
  "related": [
    {
      "url": "https://hangingcontext.com/stories/2026-10-29/ai-shipping-hits-infrastructure-wall/",
      "headline": "AI Shipping Is Hitting an Infrastructure Wall, Not a Model Wall",
      "slug": "ai-shipping-hits-infrastructure-wall"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-29/regulated-agents-compliance-layer/",
      "headline": "In Regulated Work, the Agent Product Is Becoming the Guardrail",
      "slug": "regulated-agents-compliance-layer"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-29/agent-era-org-chart-rewrite/",
      "headline": "The Agent-Era Org Chart Is Being Rewritten From the Middle Out",
      "slug": "agent-era-org-chart-rewrite"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-29/routing-decides-ai-bill/",
      "headline": "The AI bill is now a routing decision, not a model decision",
      "slug": "routing-decides-ai-bill"
    }
  ],
  "sponsor": {
    "label": "From Synorb",
    "title": "Ask what Synorb covers",
    "text": "The MCP tool synorb-stream-search answers questions like \"do you have this publisher?\" without using any Manifest allowance.",
    "cta": {
      "label": "Connect and ask",
      "url": "https://synorb.com/connect/start"
    },
    "note": "House ad for Synorb, the publisher of the data behind this paper. Not a news item."
  },
  "disclosure": "Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify.",
  "reuse": {
    "version": "1",
    "summary": "Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product.",
    "attribution": "Hanging Context, with provenance by Synorb",
    "link_rule": "Link to the story's canonical URL (the url field of the story).",
    "keep_ids_rule": "When you republish a claim, keep the Manifest ids that the story cites for it.",
    "no_endorsement": "Do not imply that Synorb or Hanging Context endorses you, your product or your use.",
    "license_url": "https://hangingcontext.com/terms/reuse/",
    "machine_url": "https://hangingcontext.com/.well-known/reuse.json",
    "canonical_url": "https://hangingcontext.com/stories/2026-10-29/ancestry-aware-prediction-gap/"
  }
}