{
  "id": 4921259,
  "title": "New AI approaches to help understand complex biological data",
  "url": "https://urgent.news/2026/09/01/new-ai-approaches-to-help-understand-complex-biological-data",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-01T18:00:07.000Z",
  "source": {
    "name": "Phys.org",
    "slug": "phys-org",
    "url": "https://phys.org/news/2026-09-ai-approaches-complex-biological.html"
  },
  "original_language": "en",
  "account": "Two new artificial intelligence techniques presented at the International Conference on Machine Learning aim to better understand complex biological datasets. Dr. You Zhou, senior lecturer at Cardiff University's School of Medicine, explained that modern AI struggles to comprehend the intricate relationships within biological data. Biological systems exhibit complex geometries and patterns that emerge at various levels of scale, but current AI architectures struggle to capture these nuances. To address these limitations, researchers developed two novel AI models: Vector Bundle Attention (VBA) and Dynamic Fractal Mamba (DF-Mamba). VBA represents biological data using a geometric framework, allowing AI to better interpret spatial relationships between cells and their environment. This model particularly excelled in single-cell RNA sequencing tasks and spatial transcriptomics, demonstrating its potential for biomedical applications. DF-Mamba, on the other hand, tackles the challenge of scale invariance, enabling AI to analyze data across vastly different levels of granularity without requiring retraining. By applying the same learned rules at multiple scales, DF-Mamba efficiently integrates information from small-scale patterns to broader contexts. Both AI models represent significant advancements in understanding and interpreting complex biological data, providing scientists with more powerful tools to unravel the intricacies of living systems.",
  "summary": "Researchers at Cardiff University have presented two studies at the 2026 International Conference on Machine Learning (ICML 2026) that address fundamental challenges in modern AI: understanding the complex geometry and relationships within data and recognizing patterns that occur across very different scales.",
  "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."
}