Researchers at the UC Santa Cruz Genomics Institute and Google Research have launched DeepSomatic, an AI-powered tool designed to identify cancer-driving genetic mutations with higher accuracy. By analyzing both short- and long-read sequencing data, the system aims to improve diagnostic power for complex cancers, including pediatric leukemia and glioblastoma. The study detailing the tool was published this week in Nature Biotechnology, marking a significant step toward integrating long-read sequencing into routine clinical care.
Closing the Accuracy Gap in Cancer Genomics
Detecting somatic variants—genetic changes present in tumors but absent in healthy cells—is essential for understanding how a patient’s cancer grows and responds to treatment. While some cancers have hereditary components, these somatic variants largely drive tumor progression. Despite the widespread use of genomic sequencing in clinical settings, technical limitations have historically hampered the ability to identify these mutations with the high confidence required for medical decision-making.
Most existing tools rely on short-read sequencing, a technique that is highly accurate for short DNA segments but limited in its ability to map complex and repetitive regions of the genome where many harmful variants reside. While long-read sequencing techniques developed over the last decade have shown promise in mapping these complex regions, their application for detecting cancer variations has remained largely untapped. DeepSomatic bridges this gap by using deep learning to interpret data from both short- and long-read technologies and cross-validating results between them. The result is a system that identifies known cancer-driving variants with greater precision while uncovering new variants that were previously undetectable.
With this tool, we’re overcoming the technical barriers that limit the accuracy of genomic sequencing when used in cancer care,
said Benedict Paten, professor of biomolecular engineering and a core member of the UC Santa Cruz Genomics Institute. Our goal is to improve genome sequencing for cancer to give a more complete picture of the mutations present, ultimately improving diagnostic power and helping uncover new molecular mechanisms.
Clinical Validation and Real-World Application
To demonstrate the tool’s effectiveness in a clinical environment, researchers tested DeepSomatic on samples from pediatric blood cancer and glioblastoma cases. The study, conducted in collaboration with Children’s Mercy and the Translational Genomics Research Institute (TGen), showed that the AI could accurately identify key mutations even in samples preserved in formalin—a common clinical preservative that often produces technical challenges for sequencing. By increasing the sensitivity and confidence of mutation detection, DeepSomatic could help clinicians more reliably match patients to targeted therapies or clinical trials.
The model is built on the DeepVariant framework, originally developed at Google and extended by the UCSC team to recognize patterns unique to tumor DNA. Unlike traditional methods that rely on rigid statistical models, DeepSomatic learns directly from vast sets of labeled sequencing data, allowing it to distinguish true variants from noise even in complex regions of the genome. In benchmark tests, the tool outperformed all existing tools, achieving higher accuracy across all sequencing platforms for both single-nucleotide variants—where a single letter in a DNA string is swapped—and small insertions or deletions.
Expanding the Toolkit for Structural Analysis
DeepSomatic is the latest development in a broader initiative to refine cancer diagnostics. It functions as a partner to Severus, a complementary method for detecting larger, structural changes in cancer genomes, which was also recently published in Nature Biotechnology. These two tools provide a complete picture of cancer genomes,
Paten noted. “With Severus, we can detect complex rearrangements, and with DeepSomatic, we can resolve the smaller but equally important variants. Together, they bring us closer to the comprehensive, multi-scale view of cancer genomes that will enable true precision oncology.”
The release of DeepSomatic comes as Google marks a decade of its dedicated genomics research initiative. According to the company, the effort began in 2015 as a foundational project to apply deep learning to sequencing challenges, evolving from a small team into a global initiative. As of October 24, 2025, this work continues to expand through partnerships with institutions worldwide to accelerate scientific discovery and advance healthcare.