4月30日，再生医学治疗糖尿病领域迎来重大突破，海军军医大学第二附属医院（上海长征医院）殷浩教授团队联合中国科学院分子细胞科学卓越创新中心程新教授团队，在Cell Discovery发表题为《Treating a type 2 diabetic patient with impaired pancreatic islet function by personalized endoderm stem cell-derived islet tissue》的研究成果。
This is the first case report in the world that the use of stem cell-derived autologous regenerative islet transplantation has successfully cured diabetes with severely impaired islet function.
The first beneficiary of the technique was 59 years old, had a 25-year history of type 2 diabetes, and developed end-stage diabetic nephropathy (uremia). In July 2021, the patient underwent autologous regenerative islet transplantation in Shanghai Changzheng Hospital, and was gradually weaned off exogenous insulin after surgery, and the dose of oral hypoglycemic drugs (baisampicin and metformin) was also gradually reduced, and the drug was completely withdrawn at weeks 48 and 56, and the patient has been completely weaned from insulin for 33 months.
This achievement marks the discovery of a new treatment for type 2 diabetes mellitus (T2D), which not only opens up a new treatment path for patients, but also brings new hope to hundreds of millions of diabetes patients around the world.
Diabetes is known as a "chronic cancer" and is mainly due to insufficient insulin secretion or an insufficient response of cells to insulin, which leads to high blood sugar. Long-term poor blood sugar control can lead to serious complications such as blindness, kidney failure, cardiovascular and cerebrovascular accidents, amputation, hypoglycemic coma and even death.
There are three main types of diabetes: type 1, type 2, and gestational diabetes. Among them, the number of pancreatic β cells in type 2 diabetes mellitus gradually decreases and degenerates, and their insulin secretion function declines, resulting in about 30% of type 2 diabetes patients having to rely on insulin injection for life.
According to the 2021 IDF Global Diabetes Map (10th Edition) released by the International Diabetes Federation (IDF), as of 2021, 537 million adults aged 20-79 have diabetes worldwide, which is expected to rise to 783 million by 2045, and an estimated 240 million people worldwide have undiagnosed diabetes, which means that nearly half of adults with diabetes do not know they have the disease.
In China, 140 million adults aged 20-79 have diabetes, the highest number of people living with diabetes in the world, and this number is expected to increase to 174 million by 2045.
Digital intelligence
01 Anthropic推出手机应用 Claude
On May 1, Claude, a large model of American artificial intelligence startup Anthropic, officially landed on the Apple App Store, providing the same experience as the website, including syncing users' chat history and supporting taking and uploading photos. Users can use photos from the gallery, take new photos, or upload files for real-time image analysis, contextual understanding, and use cases for moving scenes. The app will be available free of charge to all users of Claude, including free users, Claude Pro users, and Claude Team plan users.
https://www.anthropic.com/news/team-plan-and-ios
• Comments: OpenAI launched the ChatGPT mobile app in May 2023, and Microsoft also released the Copilot iOS mobile app in January 2024. (Tong Ka Lok)
02 Microsoft's nearly US$4 billion deployment of AI infrastructure in Southeast Asia
From April 30 to May 2, Microsoft announced three investments: it will invest $1.7 billion over the next four years to expand its cloud services and AI business in Indonesia, including the construction of a data center; building its first cloud data and AI hub in Thailand to improve the availability of its cloud services; It will invest US$2.2 billion over the next four years to build cloud and AI infrastructure in Malaysia. Microsoft has pledged to provide AI training opportunities to 2.5 million people in Southeast Asian and ASEAN member states by 2025.
• Comments: In three days and three places, Microsoft has made a big layout in Southeast Asia and has high hopes for this emerging market. Although Southeast Asia is still relatively early in the field of AI, research by global consulting firm Kearney shows that AI will contribute nearly $1 trillion to Southeast Asia's GDP by 2030, with huge growth potential. (Zhang Jin)
03 Apple launched the strongest chip, M4
On May 8, Apple launched new iPad Pro, iPad Air, M4 chip, Apple Pencil Pro and other new products at the spring new product launch conference. Among them, the M4 chip plays the slogan "strong enough to be proud of today's AI PCs".
The chip uses the second-generation 3nm process technology, has 28 billion transistors, a unified memory bandwidth of 120GB/s, and a new display engine, which can bring better accuracy, color and brightness to the ultra-fine retina XDR screen. The new CPU has up to 4 performance cores and 6 efficiency cores, all equipped with next-generation machine learning accelerators, and the CPU speed is up to 50% faster than the M2 chip in the previous iPad Pro. The next-generation 10-core GPU architecture features dynamic caching, bringing hardware-accelerated mesh shading and ray tracing to the iPad for the first time, up to four times faster than the M2.
• Comments: A year after the iPad product line was discontinued, skipping the M3 in the iPad Pro series with the M4 chip, making the performance of the tablet better than the laptop MacBook, which may mean that Apple will give priority to embracing AI from the mobile terminal. (Zhang Jin)
04 "Origin Wukong" successfully achieved the integration of four calculations
On May 5, China's third-generation independent superconducting quantum computer "Origin Wukong" was connected to the Wuhu cluster computing power public service platform of the Yangtze River Delta hub, realizing the "four calculations in one" of general computing, intelligent computing, supercomputing and quantitative computing. The platform supports cross-computing center heterogeneous resource scheduling, and users can directly use the "Origin Wukong" computing power through this platform.
"Origin Wukong" is equipped with a 72-bit independent superconducting quantum chip and will be launched in early 2024. As of May 5, it has attracted more than 7.77 million visits from 119 countries around the world, and has successfully completed more than 178,000 computing tasks.
• Comments: "Four-in-one" represents a new computing paradigm and a new direction of evolution, which will promote the modularization and interconnection of quantum chips, making them more practical faster. (Zhang Jin)
Biomedicine
05 谷歌推出AlphaFold 3
On May 8, Google's DeepMind and Isomorphic Labs jointly released AlphaFold 3, the latest artificial intelligence model in the field of proteins, in Nature, which can predict the structure of proteins, DNA, RNA and other biomolecules and how they interact, which can help scientists develop more effective therapeutics for disease mechanisms more precisely.
AlphaFold 3 is at least 50% more accurate at predicting protein-protein interactions with other molecule types than existing prediction techniques, doubling the accuracy in some key interaction classes, and surpassing all existing technologies in predicting molecular interactions.
• Comments: AlphaFold 3 builds on AlphaFold 2, which made a fundamental breakthrough in protein structure prediction in 2020. AlphaFold 3 is a leap forward in technology that not only optimizes the accuracy of protein structure prediction, but also predicts the complex interaction patterns between proteins and key biomolecules such as DNA and RNA, which are critical to understanding protein function, regulatory mechanisms, and even the mysteries of cellular life. (Zhang Yue)
06 诺华超17亿美元收购Mariana Oncology
On May 2, Novartis announced the acquisition of Mariana Oncology, a U.S. biotechnology company, with an upfront payment of $1 billion and a potential milestone payment of $750 million.
Founded in 2021, Mariana Oncology has developed a pipeline of novel peptidyl radiopharmaceuticals targeting a wide range of solid tumor cancers, covering multiple solid tumor indications such as breast, prostate, lung, and more, and has invested in manufacturing capabilities, an extensive isotope supply chain, and new formulations. The Company's lead program, MC-339, is a novel actinium-based RLT that is being studied for small cell lung cancer and is expected to enter the clinical phase in 2024.
• Comments: Compared with traditional tumor radiotherapy, radioligand therapy (RLT) has the advantages of precise targeting, strong killing, and limited damage, and pharmaceutical giants have adopted acquisitions: in October 2023, Eli Lilly acquired Point, a radiopharmaceutical company, for $1.4 billion, followed by BMS's acquisition of RayzeBio, a nuclear pharmaceutical company, for $4.1 billion, and AstraZeneca's acquisition of Fusion, a next-generation radiopharmaceutical company, for $2.4 billion in March 2024. Novartis' acquisition has once again boosted the attention of radiopharmaceuticals. Novartis, which currently has two approved RLT therapies – Pluvicto and Lutathera, strengthens its RLT pipeline and expands its research infrastructure and clinical supply capabilities to bring new cancer treatments to patients faster. (Luo Xianxian)
07 Harvard University and Google jointly released a nanoscale human brain map
On May 10, a research team led by scientists from Harvard University and Google released a nano-resolution atlas of the human brain, showing that the 1 cubic millimeter cerebral cortex contains 57,000 cells and 150 million synapses, which is the largest and most detailed replica of the human brain to date, and shows the synaptic connection network in the brain for the first time. The findings were published in Science in the paper A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution.
The human brain sample was obtained from the temporal cortex tissue of a 45-year-old female epilepsy patient, with a size of about 1 cubic millimeter, and the researchers used an ultramicrotome with an automatic collection device to cut 5,019 serial sections with a thickness of about 33.9 nanometers using an ultramicrotome of an automatic collection device, and then used a multibeam scanning electron microscope to image each slice at a resolution of 4×4 nanometers per pixel to obtain raw 2D image data with a total size of about 1.4 petabytes.
After the original data collection was completed, the researchers used computational tools to stitch and align these 2D images and reconstruct the 3D voxel data. They used the flood-filling networks (FFN) algorithm to segment the neuronal morphology of the whole voxel, and by manually correcting the segmentation errors, they finally reconstructed the three-dimensional morphology of all cells, synapses, blood vessels and other structures in the 1 cubic millimeter brain tissue. At the same time, they also use machine learning models to automatically identify synaptic locations and distinguish between excitatory and inhibitory synapses.
• Comments: This study will classify and quantify human brain cells, blood vessels, and synapses, and visualize the complex structure of the brain, which will provide further insight into how the brain works. (Tong Ka Lok)
Southern Weekly Science and Technology Innovation Research Center
Editor-in-charge: Huang Jinping