{
  "id": 7062581,
  "title": "Explainable Machine Learning and Epigenomic Profiling Decipher the Topological Determinants of Lentiviral Integration and Longitudinal Persistence in SCID-X1 Gene Therapy",
  "url": "https://urgent.news/2026/09/12/explainable-machine-learning-and-epigenomic-profiling-decipher-the",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-12T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.09.05.749598v1?rss=1"
  },
  "original_language": "en",
  "account": "Lentiviral vectors have become a standard tool for SCID-X1 gene therapy, yet the underlying epigenomic and spatial factors that determine precise integration targeting and long-term clonal persistence remain poorly understood. By leveraging extensive clinical multi-omics datasets from SCID-X1 trials, researchers curated a comprehensive dataset of 274,959 unique integration sites across 276,839 clonal records from 10 patients, using a total of 549,918 genomic controls for comparison. By comparing these integration sites to length-weighted genomic controls, five key epigenomic and topological features were identified and analyzed.\n\nUsing machine learning algorithms including Logistic Regression, XGBoost, and a Deep Genomic ResNet, the research team evaluated the ability of these models to predict integration patterns. Through rigorous inner 5-fold chromosome-grouped cross-validation and held-out test chromosomes analysis (chr19-22, chrX), the Deep ResNet model emerged as the most accurate predictor, achieving an impressive ROC-AUC of 0.7857 (95% CI: 0.7672-0.8029) and PR-AUC of 0.8027 (95% CI: 0.7640-0.8337), outperforming baseline linear models. Crucially, these findings were validated against random permutations, confirming the robustness of the model's predictions.\n\nNotably, the study revealed that compact topological and epigenomic factors can effectively predict lentiviral integration without the risk of spatial data leakage. Furthermore, longitudinal tracking of integration sites uncovered 79,587 persistent clones spanning at least two time points. These persistent clones exhibited a significant 31% depletion near proto-oncogenes, suggesting a potential safety concern for SCID-X1 gene therapy. The insertional safety of SIN lentiviral gene therapy is confirmed by the long-term depletion of persistent clones near oncogenes, providing valuable insights into the topological determinants of lentiviral integration and persistence in gene therapy applications.",
  "summary": "Lentiviral vectors (LVs) are clinically established for SCID-X1 gene therapy, yet the quantitative epigenomic and spatial determinants governing integration targeting and long-term clonal persistence across chromosomes remain incompletely characterized. Using clinical multi-omics datasets from SCID-X1 trials, we curated 274,959 unique clinical integration sites (VIS) across 276,839 clonal records…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}