{
  "id": 18107850,
  "title": "Reference-Guided De Novo Retrieval of 3D Chromatin Organizational Features with Representation Learning",
  "url": "https://urgent.news/2026/11/04/reference-guided-de-novo-retrieval-of-3d-chromatin-organizational",
  "topic": "science",
  "section": "Science",
  "published": "2026-11-04T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.10.31.763357v1?rss=1"
  },
  "original_language": "en",
  "account": "Chromosome conformation maps offer insights into genome organization by revealing recurring contact patterns, some linked to known architectural features and others tied to less understood biological processes. The task of identifying instances of a specific pattern within these maps can be crucial for understanding its biological significance and consistency across different experimental conditions. However, traditional feature detection methods often rely on predefined patterns, which may not be effective when researchers possess examples of a desired feature but lack a specialized detection approach. To address this challenge, researchers have developed Chromatin Fetch (C-Fetch), a reference-guided retrieval framework that leverages examples of a desired feature and a contrasting class to rank and assign contact-map patches at predetermined candidate coordinates. C-Fetch employs a weakly KL-regularized 3D autoencoding technique to learn a reusable representation of structurally related, non-temporal Hi-C and Micro-C patches without the need for architectural feature labels.\n\nThe researchers demonstrate the effectiveness of C-Fetch in retrieving biologically-defined contacts, patterns specified visually, and established architectural features. For instance, polycomb-guided retrieval successfully identifies contacts enriched for PRC2-associated signals, surpassing enrichment observed in loop-guided assignments at a distance of 5 kb. Furthermore, queries resembling condensed domains lead to progressively stronger rectangular enrichment at higher ranks when utilizing C-Fetch.\n\nMoreover, C-Fetch showcases the utility of loop-guided retrieval, which enhances aggregate focal enrichment compared to complete Mustache proposal sets across four held-out maps. In cases where selected contacts undergo cohesin depletion, they exhibit greater focal loss compared to matched remaining candidates. The applicability of C-Fetch is not limited to a single technology, as it also extends to Region Capture Micro-C and HiChIP datasets. The ranking agreement observed across biological replicates and under reduced sequencing depths further underscores the robustness and versatility of C-Fetch in analyzing contact-map features of interest. Overall, the introduction of C-Fetch establishes a flexible methodology for specifying contact-map features using reference examples, facilitating biological investigations, exploration of visual patterns, and refinement of existing calls across diverse datasets and technologies.",
  "summary": "Chromosome conformation maps reveal spatial genome organization through recurring contact patterns, some associated with established architectural features and others with less understood biological processes. Identifying instances of a pattern across these maps is necessary to investigate its biological associations and recurrence across experimental settings. However, conventional feature…",
  "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."
}