#25 | Gene Therapy For Rare Condition

Quote of the Week 💬 

Gene Therapy Halts Progression Of Rare Genetic Condition In Boy

⌛ The Seven Second Summary: SickKids announced encouraging clinical trial results about a 4-year-old boy with a rare genetic condition who received a personalized gene therapy treatment.

🔬 How It Was Done:

• The boy suffers from spastic paraplegia type 50 (SPG50). This condition affects nerve cells and leads to cognitive issues, speech impairments, seizures, and progressive paralysis.

• SPG50 can be caused by a variation in the AP4M1 gene. This gene produces a protein that helps nerve cells properly sort, package and transport proteins and other essential components within a cell. Without this protein, nerve cells struggle to function and communicate with other cells.

• To treat his condition, the researchers administered a working copy of the AP4M1 gene in the boy’s spinal fluid.

🧮 Key Results:

• 12 months after receiving the gene therapy treatment, the boy has yet to show any serious adverse events.

• Furthermore, the patient’s condition appears to be stabilizing, and may be modestly improving in certain motor skills. For example, the boy is now able to stand with his heels on the ground, which was not possible before.

💡 Why This May Matter: 10,000+ rare diseases affect nearly 400 million people worldwide, and less than 5% of these diseases have treatments approved by the United States’ Federal Drug Administration. Many of these diseases cause considerable disabilities, premature death and disproportionately impacts healthcare systems. While not every rare disease is caused by a gene mutation, we regularly highlight stories where researchers are able to craft promising treatments through gene therapies because the results are astounding when they work.

🔎 Elements To Consider: The 4-year-old was the only patient in this clinical trial, so the team’s therapy treatment still needs to succeed in several more trials with many more patients before any drug approvals.

📚 Learn More: SickKids. Nature.

Stat of the Week 📊 

New Drug Provides Total Protection From HIV In Recent Trial

100%

⌛ The Seven Second Summary: A new antiviral drug called lenacapavir went through a Phase III clinical trial in Africa and has given young women total protection from HIV infection.

🔬 How It Was Done:

• HIV occurs from a viral infection that affects CD4 cells and weakens the body’s immune system.

• The virus can be prevented with ~99% efficacy by taking pre-exposure prophylaxis (PrEP) drugs, which helps to block an enzyme the virus needs in order to make copies of itself and spread throughout the body.

• Unlike common PrEP drugs, which are oral pills that require daily dosing for optimal effectiveness, lenacapavir is an injectable medication that is administered every 6 months. The clinical trial’s goal was to compare lenacapavir’s efficacy with two daily pill medications.

🧮 Key Results:

• 0 of the 2,134 women who received lenacapavir contracted HIV

• By comparison, 16 of 1,068 women (1.5%) who took Truvada and 39 of 2,136 women (1.8%) who received Descovy contracted HIV.

• These results are so convincing the trial was halted early at the recommendation of an independent review committee, which said all participants should be offered the injection because it clearly provides superior protection against HIV.

💡 Why This May Matter: The Global HIV Prevention Coalition makes it clear that despite significant progress in recent years, the world is not on track to end the AIDS epidemic by 2030 that it set out for itself back in 2017. However, since the lenacapavir trial showcases superior efficacy and convenience, it can accelerate HIV prevention efforts and help get the global response to the epidemic back on track.

🔎 Elements To Consider: Gilead charges $42,250 per patient per year for lenacapavir in the United States, where it is already approved as a treatment for HIV. However, PrEP medications need to cost less than $54 a year per patient for South Africa, for example, to afford them. Thus, generic alternatives need to be produced, and partnerships with Gilead need to be sorted out so financial hurdles do not restrict critical medical access around the world.

📚 Learn More: NYT. Gilead. PURPOSE Study.

AI x Science 🤖

The Value of Tacit Data To Improve A Model’s Reasoning Skills

Last week, we shared the framework from a team that holds the current world record at solving the ARC challenge. The basic idea behind their AI system is they designed it to extrapolate and infer patterns within a single visual reasoning question, generate many synthetic examples based the context of the specific question it received, and train itself on the self-generated examples it produced until the system becomes proficient at solving the visual task.

This approach is novel because it requires an AI system to problem-solve primarily with real-time data rather than rely on training data and historical examples to inform its reasoning decisions. The early success this framework has seen at solving the ARC challenge means it may grow in popularity in the months and years ahead. As always, we will keep you updated on notable progress we see in the field. That being said, there is one other framework we have seen to develop a model’s reasoning skills that we want to bring to your attention, because its potential is significant.

A couple weeks ago, Scientific American published the transcript of a forthcoming interview discussion with Terence Tao, a mathematics professor at UCLA. During the discussion, he mentions the importance of tacit data — the implicit knowledge and intuition experts possess but do not explicitly write down in scientific papers and formal documents. Tao mentions how experts build a corpus of tacit data from years of experience and countless discussions about a variety of topics with colleagues, mentors and mentees. Over time, this data shapes and sharpens a person’s intuition and reasoning abilities and helps them become an expert at a particular field.

Tao also mentions that this data is not easily accessible because scientific papers and most published documents are not appropriate mediums to convey all this information at once; essential elements are shared, but valuable knowledge such as heuristics, taxonomies, nuanced perspectives, practical tips and lessons learned are often omitted in favor of brevity.

This presents an opportunity. If researchers make concerted efforts to transcribe their tacit knowledge in explicit ways over text, voice, or video formats, then AI systems will be able to learn from them in more comprehensive ways. Documenting the world’s tacit data in useful formats for model training will be a long, gradual process. But the benefits may ultimately lead to more reliable AI systems, and we would be surprised if the largest AI research labs are not already experimenting down this path.

Our Full AI Index

• Doctors & Wearable Devices: The Wall Street Journal published a fascinating article about how the Apple Watch is being used by doctors and researchers to diagnose and manage diseases despite not being fully approved for these applications by regulatory bodies. A growing body of research is demonstrating data from the wearable device is useful for clinical purposes, such as detecting irregular heart rhythms, monitoring patient recovery after surgery, and anticipating serious heart problems like silent ischemia. It will be interesting to see how wearable devices advance over time and contribute to early disease diagnosis. WSJ Archive.

• Learning Molecular Representations In A Cell: Researchers at the University of Notre Dame and The Board Institute of MIT and Harvard shared a new approach to help scientists understand how molecules work in cells. Their system attempts to connect the behavior of various molecules to the effects they have on different types of cells. In tests, their InfoAlign system either performed comparably or outperformed 19 other systems at predicting how molecules would behave. It was also able to predict a molecules effect on the shape and structure of cells without receiving any training data on these sets of questions, and it was the only system capable of doing so. arXiv. Github.

• Finding GPT-4’s Mistakes With GPT-4: The OpenAI team published a paper with details about CriticGPT, a model they trained on GPT-4 to detect errors in code generated by ChatGPT. Their model was trained using Reinforcement Learning from human feedback, and they found the system’s bug critiques are preferred by trainers over actual human feedback 63% of the time during blind comparison tests. The caveat here is the team grades these results with a low confidence level, but it’s an indication of how helpful AI agents are at improving developer code quality. OpenAI. Paper.

Other Observations 📰

China’s Space Probe Returns + Phosphate Found In An Asteroid

China's Chang'e-6 mission successfully returned to Earth with the first-ever samples from the Moon's unexplored far side. The spacecraft managed to land on a crater near the Moon's south pole, collect about 2 kilograms (4.4 pounds) of lunar material using a scoop and a drill, and then dock with an orbiter to return to Earth after 53 days.

The returned samples are expected to help scientists answer key questions about the Moon's formation and evolution, such as why the Moon's two faces are so different, what the composition of deeper lunar structures is, and when the Moon’s South Pole basin formed. The samples may also provide clues about the Moon's birth and evolution. For example, there is a longstanding debate about whether or not the Earth and Moon are a mixture of the same meteorites, and these samples may help to confirm or disprove the theory.

It is appropriate the news about China’s lunar achievement occurred during the same week NASA’s OSIRIS-REx mission shared some of the sample results it collected from the Bennu asteroid last fall. The team discovered the asteroid contains all the essential components for life on Earth: carbon, nitrogen, water-soluble phosphates, and other organic compounds. While this does not confirm the presence of life out in the universe, it does suggest the ingredients that allows life to flourish on Earth also exist on another celestial body in the universe. YouTube BBC. Cell. NASA. Wiley.

Our Full Science Index

• Honeybees “Smell” Lung Cancer: Researchers at Michigan State University discovered honeybees can detect chemical biomarkers for lung cancer in human breath. To simulate the breath of someone with lung cancer in one experiment, they created a synthetic odor composed of 6 different chemical compounds associated with the disease. In other studies, they grew different types of human lung cancer in cell cultures and tested whether the insects could distinguish between the odors. When they recorded the neural signals from the honeybees' brains, they found the insects showed neural activity in response to the cancer odors even when present in a mixture at concentrations as low as 1 part per billion. They also noticed distinct neural activity fired when presented with different types of lung cancer odors, which suggests the honeybees have an incredible sensitivity to the disease, and similar techniques may be emulated to advance early diagnosis efforts. Michigan State University. ScienceDirect.

• Eliminating Trans Fats From Food Supplies: The World Health Organization (WHO) shared a new report documenting significant progress to eliminate global trans fat from food supplies. 53 countries have now implemented best practices from the WHO, which covers 46% of the world's population. This is a substantial increase from just 6% population coverage in 2018, and this improvement is expected to save 183,000 lives per year. The report estimates if 8 additional countries implement the WHO’s best practices, 90% of global deaths associated with trans fat can be eliminated. WHO.

• Predicting Parkinson’s: Researchers from University College London used a machine learning-based system to analyze a panel of eight blood-based biomarkers from 99 different patient blood samples, and the system was 100% accurate at diagnosing Parkinson's disease and predicting if somebody would ultimately develop condition. With the help of their system, the team correctly predicted disease diagnoses 3.5 years before symptoms presented themselves on average, though the earliest prediction came 7.3 years before symptom onset. University College London. Nature.

Media of the Week 📸 

Advancing The Development Of Tactile Robot Hands

A team of researchers from the University of Bristol developed a four-fingered robotic hand with tactile fingertips that can rotate objects in any direction and orientation. We love to showcase robots with impressive dexterity skills in our newsletter, and this video does not disappoint. University of Bristol. Science.

Researchers Map A Fly’s Nervous System In Breathtaking Detail

Researchers at UW Medicine used machine learning, simulation techniques, electron microscopy, and other tools to craft a detailed wiring diagram of the motor circuits in the central nervous system of fruit flies. They mapped over 14,000 neuronal cell bodies and ~45 million synapses, which led to some unexpected discoveries. For example, ~70 motor neurons control a fly's six legs, and only 29 motor neurons govern the muscles of a fruit fly wing. Despite this relative simplicity, the researchers also found each motor neuron has thousands of synaptic connections from hundreds of other neurons. UW Medicine. 1st Nature Paper. 2nd Nature Paper.

Star Clusters When The Universe Was 460 Million Years Old

The James Webb Space Telescope's incredible sensitivity continues to amaze. An international team of researchers used the device to discover 5 young massive star clusters in a galaxy that existed just 460 million years after the Big Bang. This is the oldest star cluster researchers have found to date, and the clusters in this photo are massive, dense, and emit substantial amounts of powerful ultraviolet light. ESA. Nature.

This Week In The Cosmos 🪐

July 5: a new moon. The best time to stargaze!

+

At 05:06 UTC the Earth will reach aphelion — the point in its orbit farthest from the Sun.