- The Aurorean
- Posts
- The World's First Partial Heart Transplant Is Still Promising | #1
The World's First Partial Heart Transplant Is Still Promising | #1
+ building the world's largest immunological dataset and much more
Hello!
Welcome to the inaugural edition of The Aurorean 🥳
We’re excited to have you here on this journey with us.
Foundational is the theme sprinkled throughout this week’s news and research.
Curious why? Wondering what science discovered last week? Let’s find out.
Quote of the Week 💬
World's First Partial Heart Transplant Recipient Is Still Healthy
“You could potentially double the number of hearts that are use for the benefit of children with heart disease”
⌛ The Seven Second Summary: it has been more than a year since the world’s first pediatric patient underwent a partial heart transplant, and they remain in good health
🔬 How It Was Done:
a donated heart was unsuitable for a traditional transplant, but upon examination, its arteries and valves were still considered safe for use
in spring 2022 the world's first partial heart transplant took place for an infant requiring a heart valve replacement
🧮 Key Results:
a year after the procedure, Owen, the child who was the first patient to receive this type of surgery, has a fully functional heart valve
the transplanted heart is also reported to be growing with Owen’s body, which will prevent the need for multiple heart valve replacement surgeries
💡 Why This May Matter: since Owen’s partial heart transplant, there have been 12 additional procedures performed around the world. This innovation is an advancement in congenital heart surgery, and may one day eliminate the need for repeated surgeries for many of the 330,000 children worldwide with heart valve defects
🔎 Elements To Consider: if this novel procedure eventually becomes a new medical standard, then a single heart donation will have the potential to save two children’s lives in the future
🧵 Thematic Thread: the achievements in this story marks a foundational step towards progressing this innovative procedure towards clinical trials, where it can be tested for safety and efficacy on a larger scale of children in need
📚 Learn More: Duke Health. JAMA Network.
Stat of the Week 📊
Project To Build The World’s Largest Immunological Dataset
2 trillion
⌛ The Seven Second Summary: The Human Immunome Project aims to create the world's largest immunological database and use this data to train a predictive AI model to forecast individual immune system responses to various stimuli
🔬 How It Will Be Done:
In 2024: the project will collect data from 7-10 clinical research centers, studying about 500 people volunteers at each site. This focus will measure immune variables, such as different types of immune cells, gene activity, concentrations of metabolic molecules, and more
In 2027: the project aims to begin global data collection, involving up to 300 collection sites across all continents. This data collection will include volunteer individuals from diverse populations, covering various socioeconomic levels, age groups, and health conditions to create a comprehensive immune database representing all of human diversity
2027 Onwards: the project will leverage the collected data to build a predictive AI model to provide better insights about how individual immune systems may respond to drugs, vaccines, diseases and more
🧮 Key Results: nearly 2 trillion data points of human immunology will be cataloged from this project and made publicly available
💡 Why This May Matter: if the project is successful, then hyper personalized medical practices may emerge not long after. The insights this AI model may provide can help medical experts worldwide maximize treatment efficacy while minimizing patient side effects
🔎 Elements To Consider: the project’s approach to AI ethics, data security, data privacy and commitment to transparency is of immense importance if it hopes to foster trust with people worldwide and build a broadly applicable AI tool with minimal bias
🧵 Thematic Thread: this ambitious project wants to build a foundational resource to expand science’s knowledge of human immunology, which may significantly accelerate society’s path towards personalized medicine
📚 Learn More: Science. Human Immune Project.
AI x Science 🤖
Credit: Possessed Photography on Unsplash
Google Deepmind Unveils New AI Methods To Train Robots
Google DeepMind has unveiled three new innovations in robotics research and development: AutoRT, SARA-RT, and RT-Trajectory. AutoRT utilizes two sets of models to create a system that can simultaneously direct multiple robots to carry out a wide range of tasks in various settings to better understand its environment. For example, over the course of 7 months, AutoRT orchestrated as many as 20 robots simultaneously to build a dataset of 77,000 robotic trials across 6,650 unique tasks. Meanwhile, SARA-RT is a new system to make robotics models faster and more efficient by reducing the computational demands of the model it is fine-tuning. In fact, the best SARA-RT system resulted in 10.6% more accurate results and 14% faster performance from Google’s state-of-the-art robotics model. Finally, RT-Trajectory is a model to help robots better interpret how to perform tasks assigned to it. When a robot is given a task, RT-Trajectory automatically interprets the motion of the task within the robot’s environment and adds a visual sketch to describe the motion that the robotics model should have its robot perform. Simply adding RT-Trajectory’s visual hints to a robotics model more than doubled the performance of existing state-of-the-art models to successfully complete tasks the robot has not been trained on. Each breakthrough mentioned in this report is noteworthy in its own right. Together, these advancements establish a promising step towards the development of more capable and helpful robots in the scientific community. Google Deepmind. AutoRT Github. SARA-RT. RT-Trajectory Github.
Our Full AI Index
Business: new research finds Google's Gemini Pro and OpenAI's GPT-4V have comparable visual capabilities. GPT-4V outperformed on IQ and reasoning while Gemini outperformed at retrieving relevant images. The Decoder. Arxiv. Arxiv.
Research: in response to concerns about image manipulation in scientific papers, the Science family of journals is adopting a computer vision and AI-powered image-analysis tool to detect image alterations before papers are published. Science.
Satellite Imagery: new research utilizing machine learning and satellite imagery estimates ~75% of the world’s industrial fishing vessels are not publicly tracked. This research is possibly the first widely available global map of large vessel traffic, which may assist marine resource management and conservation efforts in the future. Global Fishing Watch. Nature. Github.
🧵 Thematic Thread: Deepmind’s robotics models are improving rapidly, and its latest open source contributions may soon become foundational tools for teams to build state-of-the-art robots
Other Observations 📰
Credit: Photo by Clay Banks on Unsplash
Encoding Plants To Become More Efficient At CO2 Capture
Plants are known for photosynthesis, which is the ability to convert CO2 from the air into chemical energy to fuel their growth. However, the primary enzyme plants use for their CO2 process is not the fastest CO2 capture enzyme known to science. Thus, a team of scientists saw an opportunity for research. They built a biosynthetic process where they paired some of the fastest CO2 capture enzymes known to science together, optimized the functionality and yield of this new process with machine learning, and encoded parts of this new biosynthetic process into different E. coli cells. The rest of the story is TBD because the team has yet to have its biosynthetic process operate end-to-end within a single E. coli. However, in spite of its preliminary results, this research appears noteworthy because of some of the broader scientific questions it may conceptually answer as the research continues: can we successfully encode enhanced biosynthetic processes into living organisms? If we can, will those organisms still live healthy lives, and will these bioenhanced processes still exist in their offspring? Max Planck Institute. Nature.
Our Full Science Index
Public Health: a new NIH study finds patients treated with Ozempic and Wegovy, two popular drugs used to manage obesity and type 2 diabetes, have a lower risk of first-time or recurring suicidal thoughts compared to other medications. NIH. Nature. Nature.
Climate: Germany’s CO2 emissions are at their lowest in seven decades + the UK’s use of gas and coal for electricity are at their lowest levels in six decades. AP News. Carbon Brief.
Cultural Events: a 13-year-old is believed to be the first person to ever beat Tetris. His reaction to winning is priceless. The Guardian.
🧵 Thematic Thread: the ongoing research to build and encode a biosynthetic CO2 capture process in plants is a foundational exploration of how advancing technologies can be harnessed to augment and potentially redefine the limits of biological processes seen in nature
Media of the Week 📸
A Housekeeping Robot To Cook & Clean
A team of researchers designed a robot to complete several household tasks for only $32,000, and recorded it in action. Check out this video to see its dexterity and finesse at cooking eggs, loading a dishwasher and handling laundry duties. Github. Arxiv.
Robotic Exosuit Improves Walking For Parkinson’s Patients
A research team built a wearable robot to help a person living with Parkinson’s walk without freezing. Watch the video to see how significant the person’s mobility is with and without the assisted technology. Harvard SEAS. Nature.
Color-Corrected Images of Neptune & Uranus
Credit: Patrick Irwin, University of Oxford
The Voyager 2 space program captured images of Neptune and Uranus in 1986 and 1989, respectively. Now, a team of astronomers reprocessed those images and utilized additional data and modern image modeling techniques to produce a far more accurate representation of the planets’ true colors. Oxford. Royal Astronomical Society.
The Aftermath Of Two Star Explosions
Credit: X-ray: NASA/CXC/Penn State Univ./L. Townsley et al.; Optical: NASA/STScI/HST; Infrared: NASA/JPL/CalTech/SST; Image Processing: NASA/CXC/SAO/J. Schmidt, N. Wolk, K. Arcand
A team of astronomers used x-ray observatory data to discover the aftermath of at least two supernova explosions from 160,000 light years away. The remnants of the explosion left a colorful gas cloud at least 130 light-year wide NASA. The Astronomical Journal.
This Week In The Cosmos 🪐
Jan 11: a new moon. The best time to stargaze!
Credit: Gabe Hobbs on Unsplash
That’s all for this week! Thanks for reading.
Until Next Time 💭
What did you think of this week's newsletter?
What do you want to see more of next time?
What frontier projects are you working on?
Send us an email and let us know!