Precise Capture of Cellular Dynamics in Disease Progression: UNAGI Generative Platform Accurately Simulates Drug Perturbations}
UNAGI, a deep generative model, accurately simulates cellular dynamics and drug effects in disease progression, advancing personalized medicine and virtual drug screening.
Editor: ScienceAI
On July 2nd, a scientific initiative was launched in a lecture hall in London, establishing the Meta-Science Alliance. Comprising over 25 funding agencies, academic groups, companies, and other organizations, its core mission is to promote meta-science—the use of scientific methods to understand and improve scientific research itself.
The alliance’s formation coincides with a growing community and broader recognition of meta-science. During the Meta-Science 2025 conference, over 830 participants from more than 65 countries attended, leading organizers to turn away some applicants.
Amid unprecedented changes in research, meta-science is maturing. AI is rapidly reshaping research, with budgets tightening and science facing politicization and attacks. Meta scientists can contribute by undertaking ambitious research and applying their expertise to benefit society broadly.
For decades, researchers in scientific and technical institutions have focused on science itself. Since the 2010s, meta-science has flourished, driven by increased awareness of reproducibility and research integrity issues, prompting calls for reform in research evaluation and publication models.
Meta-science now broadly encompasses peer review, reproducibility, research assessment, impact, open science, and citation analysis. It also involves exploring career paths, funding mechanisms, and addressing inequality and fairness in science.
In the past year, the field has exploded. Tim Errington, senior director of research at the Center for Open Science in Washington, D.C., supports the alliance’s efforts.
AI in science was a major theme at last week’s conference. Researchers are using large language models to accelerate everything from funding allocation to evidence synthesis. Meta-science can help document how AI changes science and assist funders and policymakers in strategizing AI’s impact across research fields.
Last year, the UK launched the Meta-Science Unit, the first team within the Department for Science, Innovation and Technology and UK Research and Innovation. It funds and commissions research to analyze the UK’s complex research environment and find ways to improve it.
Meta-science faces risks of misuse, raising the delicate issue of how to communicate the difficulties of reproducibility without undermining science’s credibility.
Hiding problems can be worse. Science is under scrutiny, requiring researchers to openly discuss shortcomings and address them with rigorous methods and data.
Meta-scientists should also explore how trust in science diminishes in some regions and how to rebuild it. They should help researchers improve transparency in explaining principles and results, countering efforts to weaken scientific credibility.
Currently, meta-science is at an exciting yet challenging stage. UK Science Minister Patrick Vallance calls for daring to challenge the scientific system while actively responding to public needs—an important responsibility.