{
  "id": 15164198,
  "title": "Why AI Isn’t Ready to Tackle Our Convoluted Genomes",
  "url": "https://urgent.news/2026/10/17/why-ai-isnt-ready-to-tackle-our-convoluted-genomes-15164198",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-17T09:00:00.000Z",
  "source": {
    "name": "Wired Science",
    "slug": "wired-science",
    "url": "https://www.wired.com/story/why-ai-isnt-ready-to-tackle-our-convoluted-genomes/"
  },
  "original_language": "en",
  "account": "The human genome is far from a straightforward script; it is a complex puzzle that grows more difficult to unravel as scientists peer closer. The sequencing of the entire genome, a project completed between 1990 and 2003, has not provided all the answers. In fact, just 2 percent of our DNA consists of actual genes, the sequences that code for proteins. Understanding how these genes are regulated is a far more challenging question that involves much of the remaining noncoding DNA.\n\nCells can create different types from the same genetic material by switching genes on or off. They also regulate gene expression in response to constant signals from their surroundings and neighboring cells. However, the processes governing gene regulation are proving to be exceedingly complex. Some biologists wonder if our limited understanding of this intricate system will ever be within reach.\n\nResearchers are turning to artificial intelligence to help analyze the genome. Genomic foundation models, such as Evo 2, Genos, and Google DeepMind's AlphaGenome, have been trained on vast amounts of genomic data. These algorithms make predictions about how variations in DNA sequences affect biological processes and an organism's traits, including disease risk. While these models may be useful for making predictions, they do not delve into the complex regulatory mechanisms that govern gene expression.\n\nThe assumption of these AI models is that the genome functions like a straightforward input-output system. However, the genome is more akin to a living organism with its own intricacies. Unlike a blueprint or a simple algorithm, the genome is not a linear sequence of instructions. Its complexity lies in the ways genes are regulated and expressed.\n\nTranscription factors, proteins that control gene expression, bind to specific DNA sequences called enhancers. These enhancers act as gateways, delivering the signal for polymerase enzymes to transcribe the DNA into messenger RNA (mRNA), which then gets translated into proteins. Mapping these enhancers to their corresponding genes is a daunting task, given the thousands or millions of enhancers present in the human genome.\n\nThe regulatory logic in our genome is far more nuanced than the simple \"OR\" logic seen in bacteria. In humans, gene regulation follows a \"AND\" logic, where multiple signals must be integrated to make a regulatory decision. This complexity allows for more responsive and context-sensitive regulation, akin to fine-tuning knobs rather than simple on/off switches.\n\nWhile AI models may prove useful for making predictions, they fall short of providing the deeper understanding needed to comprehend the full workings of the genome and life itself. The genome does not conform to the simplistic input-output model that such algorithms assume. Instead, it is a dynamic and intricate system, where understanding requires unraveling the complex interplay of genes, enhancers, and regulatory mechanisms that shape the diversity of life.",
  "summary": "The DNA inside our cells isn’t just a series of instructions, but something far more dynamic. Like scientists, AI models don't yet have the data to properly decode it.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Wired",
        "title": "Why AI Isn’t Ready to Tackle Our Convoluted Genomes",
        "url": "https://urgent.news/2026/10/17/why-ai-isnt-ready-to-tackle-our-convoluted-genomes",
        "published": "2026-10-17T09:00:00.000Z"
      }
    ]
  },
  "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."
}