In 2023, a new dawn ignites on the horizon of science. From electrodes in living bodies to the "singing" of gravitational waves; from exploration at the single-atom level to the Chinese's own laboratories in the vast space; from human beings' understanding of their own cellular level to artificial intelligence truly entering our lives... 2024 is about to begin, and scientists who are constantly moving forward are approaching the door of a new era of science and technology step by step.
1. Electrodes “grow” from living tissue
Injectable gel tested on microfabricated circuits. Image source: Thor Bakshid/Science.com
The physical boundaries between organisms and technology are blurring.
Swedish researchers have successfully grown electrodes in living tissue for the first time by injecting a gel with enzymes as "assembly molecules" and using human molecules as triggers. The results, published in the journal Science in February, paves the way for fully integrated electronic circuits in living organisms.
A team of researchers from Linköping University, Lund University and Gothenburg University in Sweden have connected neural tissue with electronic devices. Often, a mismatch between rigid electronics and soft tissue can damage fragile living systems. But the team used an injectable gel to create soft electrodes directly inside the body. After injection into living tissue, enzymes in the gel break down endogenous metabolites in the body, triggering enzymatic polymerization of organic monomers in the gel, converting them into stable, soft, conductive electrodes. The researchers demonstrated this process by injecting the gel into zebrafish and medicinal leeches. The gel polymerized in both organisms and "grown" the electrodes within the tissue.
This method of creating electronic circuits directly within living tissue provides a way to treat diseases through electrical signals in the nervous system or by modulating neural circuits.
2. Male mice produce functional eggs
Male mice produce functional egg cells. Image source: "Nature" website
This is research that could inspire or advance future fertility.
A paper published in the journal Nature in March reported a blockbuster result in stem cell research: converting male mouse stem cells into female cells and producing functional egg cells. About 1% of the embryos obtained after fertilization of these egg cells can produce healthy offspring.
Male and female gametes—sperm and oocyte (egg), respectively—are produced from a type of stem cell called primordial germ cells. These stem cells differentiate into gametes, which require sex chromosomes to function properly.
Previous studies have explored the possibility of changing the sex of primordial germ cells, and found that the production of gametes may be reduced and only cells with low fertility can be produced. But this time, Japan's Kyushu University Lin Katsuhiko's team reported that it is possible to use pluripotent stem cells to produce more healthy egg cells. The team used skin cells from the tails of mature male mice (which carry XY chromosomes) and transformed these cells into induced pluripotent stem cells. They cultured these stem cells in vitro, and this process produced a rare subset of cells (accounting for about 6% of cultured cells) that lacked the Y chromosome, known as XO cells.
Continued development of these XO cells in culture media induces X chromosome replication. Treating cells with reversin drugs that interfere with cell division can improve the replication efficiency of the X chromosome. The resulting double
Although the impact of changing male cells into female cells on genome stability still needs to be more rigorously assessed, this major finding is important for future research and applications.
3. Reconstruction of the double-slit experiment in the time dimension
Original double-slit experimental art drawing. Image source: "Nature" website
British scientist Thomas Young's observation of light wave interference in the 19th century is one of the most iconic experiments in the history of physics and had a profound impact on quantum physics. Now, it has a new development.
In April this year, British scientists used a "metamaterial" that can change properties within a femtosecond (a quadrillionth of a second) to recreate the famous double-slit experiment in the dimension of time rather than space. The latest experiments reveal more about the fundamental properties of light and lay the foundation for creating the ultimate material that can finely control light on spatial and temporal scales.
This experiment originally involved the diffraction of light through a pair of "slits" in space, but the new study shows that it is possible to achieve the equivalent effect in time using double slits. The research team at Imperial College London used a thin film of indium tin oxide in the experiment. On an ultra-fast time scale such as femtoseconds, the reflectivity of this material will be changed by laser, creating "slits" for light. The researchers achieved this by turning the reflectivity of a semiconductor mirror on and off twice in rapid succession and recording interference fringes along the spectrum of the light reflected from the mirror. Their experiments found that interference occurs between waves of different frequencies, rather than between different locations in space.
This result may have a variety of future applications, such as optical switches for signal processing and communications or optical computing.
4. International team announces epoch-making discovery of gravitational wave background radiation
A pair of supermassive black holes (top left) emit gravitational waves that ripple through the fabric of space-time (artist’s impression). Image source: North American Nanohertz Gravitational Wave Observatory
If the gravitational wave background is compared to an ancient and mysterious song, then the "chorus" is performing at different frequencies every day. Now, through monitoring pulsars, scientists have finally heard the song. In other words, they have obtained the first evidence of the gravitational wave background.
After 15 years of data collection, in June this year, scientists for the first time "heard" the eternal chorus of gravitational waves rippling in the universe, and the sound was much louder than expected. This is an epoch-making discovery of the gravitational wave background.
The gravitational wave background radiation is the superposition of many different gravitational wave sources. Their frequencies and intensities are different, but they are all very low. They should exist around us and may tell us the important information it has hidden forever. But unfortunately, its existence and composition have always been the product of theorizing.
The scientists report their results in a series of new papers published in June in The Astrophysical Journal Letters. The most likely source of the gravitational wave background detected this time is a pair of supermassive black holes trapped in a "death spiral." These black holes are so massive that they can reach billions of solar masses. Because almost all galaxies, including the center of the Milky Way, have such a black hole monster. So when two galaxies merge, their supermassive black holes meet and begin orbiting each other. Once two black holes are close enough, they may be observed by a pulsar timing array.
The North American Nanohertz Gravitational Wave Observatory team stated that currently they can only measure the overall gravitational wave background, but cannot measure the radiation of a single "singer" or "instrument". Even so, it was enough to surprise the entire astrophysics community, because "the gravitational wave background is about twice as loud as expected." Mingarelli, an assistant professor at Yale University in the United States, said this is the upper limit of the models that people can create from supermassive black holes.
Scientists are also excited because this research is opening up uncharted territory. Due to current experimental limitations, they cannot estimate whether something else is also producing powerful gravitational waves. If so, then there may be other explanations for the mechanisms predicted by string theory and even the birth of the universe.
5. Single-atom X-ray signal detected for the first time
Schematic diagram of the experiment that first detected single-atom X-rays. Image source: Physicist Organization Network
To achieve a historic breakthrough in materials detection methods, it is not enough to rely solely on equipment upgrades. Scientists need to innovate from the atomic level.
In June, scientists from Ohio University, Argonne National Laboratory, University of Illinois at Chicago, and others captured single-atom X-ray signals for the first time. This breakthrough achievement is expected to completely change the way people detect materials.
The research team used synchrotron X-rays to image individual atoms. The minimum sample size that can be analyzed using synchrotron X-ray scanning tunneling microscopy is attogram, which is approximately 10,000 atoms. This is because the X-ray signal produced by a single atom is so weak that conventional detectors are not sensitive enough to detect it. To solve this problem, the team added a sharp metal tip to a traditional X-ray detector, which is placed just 1 nanometer above the sample to be studied. As the sharp tip moves across the sample surface, electrons pass through the space between the tip and the sample, creating an electric current. This essentially detects each element's unique "fingerprint," allowing the researchers to combine the ultra-high spatial resolution of scanning tunneling microscopy with the chemical sensitivity provided by intense X-ray illumination.
The technology can track toxic materials to extremely low levels, allowing applications in materials design and environmental science.
6. Human Y chromosome assembly and analysis completed
The Y chromosome is the last of the 24 human chromosomes to be sequenced. Image source: National Human Genome Research Institute
This is the first truly complete sequence of the human Y chromosome and the last human chromosome to be completely sequenced.
Two papers published in Nature in August announced the assembly and analysis of the human Y chromosome. This study, involving more than 100 scientists from around the world, fills many gaps in the current Y chromosome reference and brings insights into the evolution and variation of different populations.
The human Y chromosome has been difficult to sequence and assemble due to its complex structure. More than half of the Y chromosome is missing from the current human reference genome assembly, resulting in an incomplete understanding of the Y chromosome and limiting understanding of its composition, complexity, and differences among different populations. As part of the "Telomere to Telomere" consortium, led by the National Human Genome Research Institute and including scientists from Johns Hopkins University, University of California, Santa Cruz and other institutions, this time reported the complete 62,460,029 base pair sequence of the human Y chromosome. This assembly corrected multiple errors regarding the Y chromosome in the current human reference genome assembly. It also added more than 30 million base pairs to the reference genome, revealed the complete structure of multiple gene families, and confirmed 41 new protein-coding genes.
In a separate paper, the joint team assembled human Y chromosomes from 43 men representing 21 different populations around the world. These assembly results explain in more detail the genetic differences of the Y chromosome in the 183,000-year history of human evolution. The researchers will integrate this new insight into primate studies to dig deeper into the evolution of the Y chromosome and analyze clinically relevant genes that may influence cancer and a variety of other diseases, thereby aiding personalized medicine.
7. Neural networks design new proteins
Visual example of designing protein biomaterials. Image credit: Marcus Buehler/Journal of Applied Physics
Proteins have always been difficult to model, especially when people want to "operate backwards" - converting the desired function into a protein structure, which is a difficult challenge.
A team from the Massachusetts Institute of Technology announced in August that it would combine attention neural networks with graph neural networks to better understand and design proteins. The method combines the best of both worlds, geometric deep learning and language models, not only to predict existing protein properties but also to envision new proteins that nature has not yet designed. This new model reassembles these natural building blocks by modeling the fundamental principles that serve as the foundation for everything nature has invented. When the team trained the model, it predicted the sequence, solubility and amino acid composition of different proteins based on their functions. Then, after receiving initial parameters for the new protein's function, the model got creative and generated an entirely new structure.
Coincidentally, the company "Deep Thinking" also announced a new generation of "Alpha Folding" this year, which not only significantly improved its accuracy, but also expanded its prediction scope from proteins to other biological molecules, including ligands. The model can predict almost all molecules in the Protein Data Bank (PDB) with atomic-level accuracy.
8. China’s National Space Laboratory is officially operational
The Wentian experimental module of the Chinese space station assembly display module taken at the 14th China Air Show. Photo by Xinhua News Agency reporter Liu Dawei
This year marks the 20th anniversary of the success of China's first manned mission. On August 18, good news came from the China Manned Space Engineering Office: China’s National Space Laboratory was officially operational, and space applications were being carried out in an orderly manner with frequent results.
Lin Xiqiang, spokesman for China Manned Space Engineering and deputy director of the China Manned Space Engineering Office, said that the current scientific experimental facilities of the space station have basically completed on-orbit testing, the on-orbit operation is stable and reliable, and it has the ability to carry out large-scale space scientific research. So far, the space station has carried out more than 60 experimental projects, tens of thousands of on-orbit experiments, obtained nearly 60TB of original experimental data, and downloaded more than 300 scientific experiment samples.
As the largest and long-term manned space experiment platform in China's aerospace history, the National Space Laboratory will take advantage of the environment in space to carry out scientific research, most of which cannot be simulated on earth. The multiple experimental cabinets deployed in the Wentian Experimental Module, Mengtian Experimental Module, and Tianhe Core Module will carry out thousands of scientific experiments, explore the mysteries of the universe, and transform the incubated scientific and technological achievements into real applications to benefit the lives of ordinary people on the earth.
9. The most complete human brain cell atlas released to date
"Science" magazine cover article. Image source: "Science" website
With the development of biomedicine so far, what can we rely on to gain a new understanding of the identity of the human species? One answer is brain science.
In October, 21 papers were simultaneously published in the American magazines "Science", "Science Advances" and "Science Translational Medicine", announcing and explaining the most complete human brain cell map to date. This series of studies involving scientists from many countries has revealed the characteristics of more than 3,000 brain cell types, which will help to deeply understand the uniqueness of the human brain and advance research on brain diseases and cognitive abilities. The "Nature" website quoted Anthony Hannan, an expert from Australia's Florey Institute of Neuroscience and Mental Health, as saying that this series of studies has mapped the human brain at the single-cell level for the first time, showing its complex molecular interactions and laying the foundation for a better understanding of the human brain.
Among them, the team of neuroscientist Kimberly Siletti of Utrecht University Medical Center in the Netherlands conducted RNA (ribonucleic acid) sequencing on more than 3 million cells covering 106 locations in the human brain. The analysis recorded 461 major categories of brain cells, including more than 3,000 subtypes. Research has shown that neurons, the cells that send and receive signals in the brain and nervous system, vary greatly in different parts of the brain. In particular, the brainstem region that connects the brain to the spinal cord contains a particularly large number of neuron types. This difference reveals different functions and developmental histories. This is an unprecedented granular analysis of the organizational structure of the human brain from the single cell level, including the adult brain and the developing human brain during the embryonic stage. It identifies and depicts the astonishing diversity of human brain cell types, and provides clues for understanding the mechanisms of human mental and neurological diseases.
10. Large-scale language models are continuously iteratively upgraded
Gemini can handle text, audio and video. Image source: Google Inc.
2023 is the “Year of Generative Artificial Intelligence”.
This year, GPT-4’s performance was deemed “comparable to humans.” About four months after the release of the chatbot ChatGPT, OpenAI, the company behind ChatGPT, announced the official release of GPT-4, a more powerful next-generation technology that supports ChatGPT. It has image recognition capabilities, advanced reasoning skills, and the ability to process 25,000 words, and its performance is no less than that of humans in some tests.
On December 6, Google announced the launch of a new artificial intelligence model called Gemini, claiming that the model performed better than the GPT-4 model and "expert-level" humans in a series of intelligence tests. Google claims that the mid-range Pro version of Gemini beat some other models, such as OpenAI's GPT3.5, but the more powerful Ultra exceeds the capabilities of all existing AI models. It scores 90% on the industry-standard MMLU benchmark, while "expert" humans are expected to achieve 89.8%. It's the first time an artificial intelligence has beaten humans in the test, and it's the highest score among existing models.
The test involves a range of tough questions including logical fallacies, ethical issues in everyday scenarios, medical issues, economics and geography. In the same test, GPT-4 scored 87%, LLAMA-2 scored 68%, and Claude2 scored 78.5%. Gemini beat all of these models on 8 out of 9 other common benchmarks.