#33 | Stem Cell Treatment For Diabetes

Quote of the Week 💬

Stem Cell Therapy Shows Promise In Treating Type 1 Diabetes

⌛ The Seven Second Summary: The Washington Post shared the personal account of a Type 1 diabetic patient who participated in Vertex’s Phase 1 / 2 clinical trial earlier this year and has seen remarkable results.

🔬 How It Was Done: 

• Vertex researchers developed a method to take embryonic stem cells in a lab and expose these cells to a sequence of environmental growth factors and chemicals that mimic the natural developmental stages a cell goes through to produce insulin in a person’s pancreas.

• These lab-grown insulin-producing cells were then transplanted through a blood vessel into the livers of the 12 clinical trial patients.

• Afterwards, each patient was given immunosuppressive drugs to prevent their body from rejecting the transplanted cells.

🧮 Key Results:

• After the cell transplant, all 12 patients had normal blood glucose levels for more than 70% of the time.

• 11 out of 12 patients significantly reduced their dependence on insulin.

• 3 patients with at least 12 months of follow-up no longer had severe low blood sugar episodes and became insulin independent.

💡 Why This May Matter: Most Type 1 diabetic treatments only attempt to improve how well patients can monitor and manage their autoimmune disease. In contrast, this stem cell-based therapy has the potential to be a functional cure for Type 1 diabetes because it attempts to restore the body’s natural ability to produce insulin.

🔎 Elements To Consider: A few more years will need to pass before there is enough data about the long-term safety and efficacy of Vertex’s treatment. Also, while this type of treatment may eliminate the need for insulin pumps, pens and injections, patients may need to take lifelong immunosuppressants to keep their bodies from rejecting the implanted cells, and these drugs carry their own risks and side effects.

📚 Learn More: The Washington Post. Vertex.

Stat of the Week 📊

Gene Therapy Shows Remarkable Vision Improvement

100x

⌛ The Seven Second Summary: A Phase 1 / 2 clinical trial of a gene therapy for a rare inherited form of blindness called Leber congenital amaurosis 1 (LCA1) has resulted in dramatic vision improvements.

🔬 How It Was Done:

• Researchers created the gene therapy by replacing a harmless virus' genes with a functional copy of the gene that is missing or dysfunctional in patients with this form of blindness.

• The therapy was surgically injected under the retina of 15 patients with LCA1 at either low, medium or high doses.

• Afterwards, the patients' vision was evaluated for a year with different sorts of exams, such as eye charts, light sensitivity measurements, and tests in different lighting conditions.

🧮 Key Results:

• 3 out of 6 patients who received high doses of the gene therapy achieved the maximum possible score in a mobility course under varying light levels.

• On average, patients experienced a 100x improvement in their vision, with many seeing results within a month of treatment.

• 2 of the 9 patients who received the highest dosage of the gene therapy saw a 10,000x improvement in light sensitivity.

  • For context, this scale of improvement meant a patient who needed bright indoor lighting to see anything before the clinical trial could see their outdoor surroundings on a moonlit night after treatment.

💡 Why This May Matter: Earlier this year, we highlighted a separate study utilizing a different type of gene therapy to treat a group of children diagnosed with a rare case of hereditary deafness. Some of these preliminary trials in hearing and vision are showcasing remarkable results, and we hope they are meaningful advances in how scientists may the 10,000+ rare diseases affecting nearly 400 million people worldwide in the years ahead.

🔎 Elements To Consider: The average patient exam score is skewed by some of the outlier results from patients who received high doses of the therapy treatment. Several patients who received low and medium doses saw their vision improve over the course of the trial, but their results were more modest.

📚 Learn More: University of Pennsylvania. The Lancet.

AI x Science 🤖

The Potential Of AI-Driven Protein Design & Structure Predictions

We frequently reference how AI-driven solutions are advancing fields such as drug discovery, biological research and healthcare systems in isolation. However, we have not previously mentioned how these different systems may converge and integrate together to allow scientists to treat all sorts of intractable illnesses and diseases.

First, let’s consider AlphaFold 3 and an open-source version of the tool that was released last week. The latest version of AlphaFold has far more robust capabilities than its inaugural version. The product not only predicts the structure of nearly every protein known to science, the model now also considers and predicts how these proteins might interact with different amino acid sequences, chemical bonds, and other molecules in various settings.

This means the scientific community can more comprehensively predict how disease-related proteins might interact with our body’s biological systems, and they can use the same tool to predict how a potential treatment might interact with the disease and the rest of our body in wanted and unwanted ways.

Next, there is AlphaProteo, a new AI-based system Google Deepmind announced last week. AlphaProteo is designed to create protein binders, which are molecules that can tightly attach to specific target proteins. The team trained their system to understand the intricate ways that proteins interact with one another by feeding it vast amounts of protein data, some of which included data about protein structures predicted by AlphaFold. 

With AlphaProteo and AlphaFold 3 working in tandem, scientists can now theoretically predict the structure of a disease-related protein using AlphaFold 3, then use AlphaProteo to design a custom protein to bind to and potentially neutralize the disease-causing protein. For instance, this combination could be used to develop a treatment for a specific type of cancer by designing a protein to precisely target a mutated protein that may responsible for tumor growth. Researchers can then use AlphaFold 3 to predict how this new protein interacts with the body's surrounding environment, so they can select treatment proteins that are less likely to cause unwanted side effects.

Finally, let’s mention OpenCRISPR-1. It is a language model that has been trained on millions of CRISPR-related proteins. Scientists can use this system to design new gene-editing tools. They can also use the product to generate custom instructions for their newly designed tools to target specific genes in whichever way they may need to produce the sorts of clinical outcomes they want.

When these three models are put together, the potential for breakthroughs are even more profound. For example, the aforementioned mutated protein example may be the result of some underlying genetic mutation. In this case, OpenCRISPR-1 can be used to develop a gene therapy treatment to correct the genetic mutation and resolve the patient’s health problems at an additional source.

While there is still a lot more biological data science needs to contextualize and understand in order to treat and prevent a host of diseases and illnesses, it is remarkable to see how much faster the scientific community is learning with the advent of these tools. The journey from lab discovery to approved treatment remains long, and these sorts of AI-driven tools do not necessarily speed up end-to-end process all that much. However, they are creating a synergistic effect for researchers to tackle complex health issues from multiple angles, and the eventual treatments of tomorrow may be groundbreaking.

Our Full AI Index

• AI System Detects Signs Of Heart Failure: Researchers from Mayo Clinic developed an AI-enabled digital stethoscope to help detect heart failures in pregnant women. A trial in Nigeria had 1,232 participants, and the team’s AI system was able to detect twice as many cases of heart failure compared to typical care without the tool. The system was also 12x more likely to identify patients with hearts pumping at half its capacity compared to standard care. Nature.

• Predicting Model Performance Without Training: Researchers from Stanford University and UC Berkeley shared an accurate scoring system to determine the performance of a multimodal large language models (MLLMs) with less training. The team discovered a model's ability to understand and describe images has > 95% correlation coefficient to how well its visual representation matches up with its language understanding. Since this was a near linear relationship, the scoring system allowed the researchers to accurately predict top vision models by using a fraction of compute costs on training. AI’s efficiency boon is well underway. arXiv.

• The Value Of The World’s Fastest Supercomputer: Nature shared an interesting purview into ‘a day in the life’ of Frontier, the world's fastest supercomputer, at the Oak Ridge National Laboratory. Frontier operates with 10x more compute power than any other machine on the planet — equivalent to over 100,000 laptops working simultaneously. Researchers around the world are using the system to create highly detailed simulations across various fields, from climate modeling to drug development. For example, the article referenced projects to simulate a 155-billion-molecule water droplet and run global climate models at unprecedented resolution fidelity. The article also mentioned the computer’s water-cooling system, which allows it to operate with 4x more energy efficiency than its predecessors. Nature.

Other Observations 📰

Cleaner Air Means Longer Lives

China declared a "war against pollution" in 2014. A decade later, the country's fine particulate air pollution (PM_2.5) has decreased by 41%. PM_2.5 is particulate matter with a diameter of 2.5 microns or smaller. The lungs can inhale these small particles and pass the pollutants through to the bloodstream. As a result, PM_2.5 shows adverse health impacts in many studies, including the 2024 State of Global Air report. For instance, in 2019, the United States' Environmental Protection Agency released an extensive report where they mention long-term exposure to PM_2.5 is “likely to be causal” to “nervous system effects.” We discuss the impact of PM_2.5 and other lifestyle factors on people’s health more in our Alzheimer’s deep dive article from a few weeks back.

That being said, the nation’s progress against PM_2.5 means the average Chinese citizen can now expect to live 2 years longer compared to a decade ago, provided these reductions are sustained (if not improved) over the course of a person’s lifetime. This is a great example of how targeted environmental policies can have profound impacts on public health in a relatively short timeframe. The Air Quality Life Index’s report about China’s progress is even more impressive to see because China has been experiencing some the fastest industrialization and urbanization shift over the past few decades in human history.

Nonetheless, the data also reveals China's battle with air pollution is far from over. While the country's average particulate pollution now complies with its national standard of 35 micrograms per cubic meter, (μg/m³), it is still 7x higher than the World Health Organization’s (WHO) guideline of 5 μg/m³. This gap means that particulate pollution continues to shorten the average Chinese resident's life expectancy by 2.3 years relative to what it would be if the WHO guideline was met.

The regional variations in air quality across China also paint a more nuanced picture. In Guangdong, China's most populous province with about 110 million residents, the average PM_2.5 concentration is 19.4 μg/m³. This means residents here could gain 1.4 years of life expectancy if PM_2.5 levels met the WHO guideline. Meanwhile, in China's most polluted province, Shijiazhuang in Hebei Province, the average person is losing 4.4 years of life expectancy compared to the WHO guideline.

Our Full Science Index

• The First Metal Part Ever Created In Space: The European Space Agency’s Metal 3D Printer recently produced the first metal part ever created in space aboard the International Space Station. The printed samples will return to Earth shortly for quality analysis to assess the viability of metal 3D printing in microgravity. If the samples are encouraging, this effort may be a massive step towards manufacturing parts and tools on-demand in space, which would reduce a crew’s reliance on Earth and need to return to the planet to resupply long-term missions. ESA.

• A Device To Detect & Reverse Opioid Overdoses: Researchers from MIT developed an implantable device to automatically detect and reverse opioid overdoses. The device is small — about the size of a piece of gum — and monitors vital signs so it can rapidly administer naloxone when an overdose is detected. In a preliminary animal study involving 25 pigs, the team’s device was able to reverse 24 out of 25 pig overdoses within 4 minutes. This may be most helpful for patients who are in locations where first responders cannot reach people quickly or easily. MIT. Cell.

• Reducing The Risk Of Heart Failure Events: Researchers from Brigham and Women’s Hospital reported a new type of heart medication called Finerenone reduced heart failure complications and deaths in patients whose hearts struggle to pump blood effectively. In a trial of 6,000 patients, Finerenone lowered the rate of worsening heart failure events by 18% compared to a placebo, though it increased the risk of high potassium levels in the blood. Brigham and Women’s Hospital. NEJM.

Media of the Week 📸

Performing Surgery On The Other Side Of The World

Researchers at The Chinese University of Hong Kong and ETH Zurich helped surgeons in Switzerland perform a remote endoscopy on a pig in Hong Kong 9,300 km away by using a video game controller and magnetic endoscope. The procedure maintained a latency below 300 milliseconds, which means there was hardly a discernible delay in the system’s response to the surgeons’ controller commands. ETH Zurich. 

A Robot Controlled By Mushroom Threads

Watching a nonchalant robot make its way over a desk or floor would ordinarily not make a cut in our Media of the Week section. But this is no ordinary robot.

Researchers from Cornell University created a robot that is partially controlled by living fungal threads from king oyster mushrooms. The threads of the mushrooms were grown into the robot's hardware so the fungus could transmit electrical signals throughout the robot’s body and control its movements. Since the mushrooms can sense and respond to light, the short flashes seen in the video are the stimulus for the mushroom threads to activate and get the robot moving. We have no idea where the future of biohybrid robots will take us, but we think it’s fascinating to see a living thing that is otherwise stationary have a chance to experience free form movement. Science Robotics.

An Aerial View Of Aurora Streams Over Earth

The International Space Station takes some of the best photos of Earth, and this beautiful image is a perfect representation of our brand’s namesake. Long live all auroras! NASA.

This Week In The Cosmos 🪐

September 18: A partial lunar eclipse will be visible from North and South America, Europe, Africa, and parts of Asia. This will also be a Supermoon, so the moon will appear slightly larger than normal.