Google's AlphaGenome Atlas Maps All 9 Billion Single-Letter Genetic Variants, Extending AlphaFold-Era Push Into Disease Detection
核心洞察
Google (搜索) has mapped all 9 billion possible single-letter genetic changes across the human genome with AlphaGenome Atlas (搜索) and made the resource openly available to researchers worldwide.
The company reported that AI detected 25% of interval cancers previously missed in mammograms of 175,000 women in a study with Imperial College London and the UK's NHS.
Google (搜索)'s tuberculosis (搜索) chest X-ray tool has screened over 25,000 X-rays across 40 locations in six nations, while its diabetic retinopathy (搜索) model has supported more than 1.15 million screenings.
Google (搜索) has mapped all 9 billion possible single-letter genetic changes across the human genome with a new resource called AlphaGenome Atlas (搜索), making the dataset openly available to researchers, the company said. The release extends Google's genetics portfolio beyond its Nobel Prize-winning AlphaFold (搜索), which has predicted all 200 million protein structures known to science and is now used by 4 million researchers in 190 countries, and AlphaMissense (搜索), which helps researchers predict disease-causing genetic mutations.
According to the company, AlphaGenome Atlas (搜索) offers scientists predictive insights into how genetic variations alter cellular behavior, building on the foundations laid by AlphaFold (搜索) and AlphaMissense (搜索). Google (搜索) CEO Sundar Pichai outlined the work in a blog post published on September 15 and shared on X, describing AI as a means of accelerating scientific research and translating it into tangible benefits.
Cancer Detection and Global Screening Programs
The company reported results from a breast cancer (搜索) study conducted with Imperial College London and the UK's NHS showing that AI can detect 25% of interval cancers previously missed in mammograms of 175,000 women. Google (搜索) said it is also making progress on tools to help detect lung cancer (搜索), colorectal cancer (搜索), and genetic mutations in tumor cells.
In infectious disease, Google (搜索)'s chest X-ray tool, used by Nexus Intelligence (搜索), has screened over 25,000 X-rays across 40 locations in six nations for tuberculosis (搜索) — a disease where roughly 40% of infected people worldwide go undiagnosed, according to the company. Google is additionally using bioacoustic models to detect TB through coughs via Health Acoustic Representations.
For diabetic retinopathy (搜索), a treatable but growing cause of preventable blindness, the company's model developed with partners has supported more than 1.15 million screenings globally, with plans to expand to 6 million over the next decade.
Tools for Researchers and Clinicians
Google (搜索) said its collaborative AI tools, including Co-Scientist (搜索), are helping researchers accelerate core steps of the scientific method such as generating and validating novel hypotheses — for example, identifying new therapeutic applications for existing drugs in acute myeloid leukemia (搜索). The company has also open-sourced tools including DeepConsensus, DeepVariant, and DeepPolisher, which over the last decade have assisted in completing the human genome, drafting the first pangenome, and enabling ongoing work as part of the Human Pangenome Reference Consortium.
Through AMIE (搜索) (Articulate Medical Intelligence Explorer), Google (搜索) said it is continuing to work on prospective evidence in real-world settings, collaborating with academic and medical institutions such as Beth Israel Deaconess Center (搜索) and conducting a first-of-its-kind nationwide trial in real-world care settings. The company positions AMIE as assisting frontline care providers to free up doctors' time with patients.
Google (搜索) is also pioneering the use of everyday smartphones and wearables for early detection of cardiovascular disease, insulin resistance, hypertension, loss of pulse, and passive heart rate monitoring, and is working with leaders in Arkansas to develop a blueprint for improving health outcomes in rural areas.
Planetary Prediction Engine Deployed in Ebola Outbreak
Beyond human health, Google (搜索) introduced the Earth AI Planetary Prediction Engine, an autonomous AI system that uses simple language instructions to predict global crises such as disease outbreaks, food shortages, and climate risks. In the ongoing Ebola (搜索) outbreak in the Democratic Republic of the Congo, the company said the system pinpointed 83% of emerging hotspots ahead of time, beating current forecasting systems. In Nigeria, it doubled food security forecasting accuracy at the local district level, and in the U.S. it outperformed traditional models in identifying vulnerable communities across 21 CDC health indicators.
The engine consolidates data on global health, food security, and socioeconomics, and has also been used in the U.S. to identify vulnerable communities across the same 21 CDC health indicators.
Weather, Language, and Access Commitments
Google (搜索) introduced WeatherNext 3 (搜索), which it described as its most advanced and accurate global weather model, delivering 50% more accurate precipitation forecasts a day or more ahead. The company said the model combines real-time satellite observations with AI to deliver high-resolution hourly forecasts without immense supercomputing power — a step forward for data-sparse regions. Authorities in Jamaica used WeatherNext to predict Hurricane Melissa's path, securing early disaster funding, according to the company.
On language access, Google (搜索) said its technologies now support more than 300 languages spoken by 7 billion people, representing 86% of the global population, with more than 1 billion people translating around 1 trillion words each month. The company's stated goal is to support the world's 1,000 most-spoken languages. It has partnered with Makerere University, University of Ghana, and Digital Umuganda (搜索) to create an open dataset for 27 African languages spoken by over 100 million people, and with IISC-Bangalore and ARTPARK (搜索) on Project Vaani to open-source speech and image datasets for 109 Indic languages.
Responsibility and Collaboration
Google (搜索) said its approach pairs bold ambitions with responsibility, noting distinct challenges in applying AI to science: the need for detection, reporting, and mitigation tools to address risks around chemical, biological, radiological, and nuclear misuse, alongside the risk of "missed use" — failing to apply AI where it would provide benefit and potentially worsening existing societal gaps.
The company said advancing scientific discovery requires deep collaboration with academia, national laboratories, philanthropic foundations, non-governmental organizations, governments, other companies, and local partners. Google (搜索) said it has given more than $1 billion across 1,700+ research institutions worldwide since 2006, and has provided more than $1 billion globally in training and skilling initiatives, helping over 100 million people gain digital and AI skills.