Recently, Academician of the Chinese Academy of Sciences, Director of the Shenzhen Institute of Advanced Medical Sciences, and Director of the Shenzhen Bay Laboratory, Yan Ning, was invited by the School of Artificial Intelligence at Shanghai Jiao Tong University to give an academic report titled "CryoSeek (Cool Seek) - A New Paradigm for Biological Discovery Based on Structure" at the "Shanghai Jiao Tong University Master Lecture | Hui Zhi Lecture" on the Xuhui campus. Yan Ning reviewed her main academic career and discussed hot topics such as the relationship between AI and scientific research.
"It should be able to go into textbooks."
Yan Ning candidly stated that when she first established her laboratory at Tsinghua University in 2007, "there was a very simple goal at the time, which was to make something that could go into textbooks."
Glycolysis is present in every biochemistry textbook, so they focused their research on how glucose transporter proteins obtain energy.
Another focus was the generation and conduction of electrical signals, focusing on sodium ion channels related to this.
She joked that she is a "foodie," because these two directions are sugar and salt (sodium).
Yan Ning also revealed that her "first love" project was actually related to another "foodie" element — proteins related to fat (sterol regulatory element-binding protein, SREBP). However, progress in this field has been very slow so far. She also used this to remind students, "When you are problem-oriented, sometimes due to insufficient technology, you may have to temporarily set it aside, and then come back to study it again when new technologies emerge after many years."
However, the structure of the glucose transporter was also a huge challenge before it was cracked. "In the era dominated by X-ray crystal diffraction technology, membrane proteins were undoubtedly the most difficult targets, so we were very brave to tackle the hard problems."
Transport is not as simple as three steps — opening, passing, and closing. Large molecules have different conformations (conformer), like how a hand can open or clench. Yan Ning said that transporters have many conformations, and their structures need to be deciphered one by one. "Before understanding these details, talking about drug development based on structure is just a fantasy."
The crystal structure of the glucose transporter GLUTs and the improved working model of GLUTs in "Nature"
After these research results were successfully included in textbooks, the next goal became the sodium ion channel. Not to mention conformational changes, there are nine subtypes of the α subunit alone (Nav1.1-1.9). Among them, Nav1.7, which is related to pain perception, has attracted much attention.
Between 2004 and 2006, scientists had just identified the role of Nav1.7 through sequencing of people with congenital insensitivity to pain and pain disorders, and solving its structure was the forefront issue.
In addition, approximately one-third of adults worldwide suffer from some form of pain, and the abuse of painkillers is on the rise, so pharmaceutical companies are also very interested in this potential new drug target.
However, the method of crystal diffraction requires a large amount of very pure protein for crystallization. The sodium ion channel protein itself is very difficult to obtain, and it has glycosylation and phosphorylation. "Getting the protein is already very difficult, let alone using structural analysis."
Yan Ning lamented, "We are fortunate to have welcomed a technological revolution." A representative achievement was in 2013, when David Julius and Cheng Yifan used cryo-electron microscopy to determine the structure of the capsaicin receptor (TRPV1). David won the Nobel Prize in Physiology or Medicine in 2021.
The human Nav1.7-β1-β2 complex structure in "Cell Reports"
Yan Ning introduced many scientists who contributed to advancing cryo-electron microscopy technology, especially the "resolution revolution." The key to improving resolution is not only a significant increase in computing power, but also breakthroughs in materials science, leading to the development of direct electron detection imaging systems that can directly capture high-energy electron signals.
"Previously, achieving a resolution of 10 angstroms was already very good. After this revolution, it has now reached 1 angstrom, which is 0.1 nanometers of resolution, making it possible for electron microscopy to do what was previously only possible with X-ray crystallography and nuclear magnetic resonance." Subsequently, Yan Ning's team applied cryo-electron microscopy technology to sodium ion channels and similar calcium ion channels, and have already solved most of the conformations of these channels.
Can the AI tool Alphafold, which predicts protein 3D structures based on amino acid sequences, predict conformations? Yan Ning said that Alphafold's predictions are currently not meaningful, "not the conformations we want, and it also doesn't have that high accuracy, so we are still working hard."
Some drugs binding to Nav1.7 in "Nature Communications"
To demonstrate the work cycle of a transmembrane transporter, you need resting state, activated state, and inactivated state. Each state may have multiple substates.
"Previously, we mainly obtained the inactivated state, but now we have some clues on how to get the activated state, and it should be made soon. However, studying the resting state is still the hardest right now."
Protein structure research is still a long road with infinite scenery. In Yan Ning's view, the most important thing is the development of methods, especially the efforts of interdisciplinary collaboration. For example, the preparation of graphene grids, the method of applying electric potential on both sides of the membrane, optimization of processing procedures, improvement of resolution, etc. Resolution is particularly important. (Resolution matters.)
She warned, "The common situation is that developing methods may not result in good papers, and the progress is also slow. But once a breakthrough is achieved, it will bring great impact."
Research paradigm shift: From "problem-driven" to "observation-inspired"
During the study of proteins, due to purification, sugar was lost, and the interaction between sugar and protein couldn't be displayed. It is imperative to minimize the purification process.
If we go further, why not observe directly without special purification?
Basic electron microscopy technology is now quite mature. It not only reveals high-resolution structures of known molecules, but it may also help discover new molecules.
Yan Ning believes that in the past, we observed the microscopic world with microscopes, but now the highest resolution microscope in the world is the electron microscope, especially the cryo-electron microscope. "Can we use it to look at this vast world and see what we can find?"
This "observation-inspired" research paradigm is completely different from the previous "problem-driven" approach.
Finally, to avoid "tainting samples from outer space, the moon, or the deep sea," Yan Ning's team chose the lotus pond described by Zhu Ziqing.
They took water samples, filtered out the impurities, concentrated the samples, and put them into the electron microscope — "Suddenly, it was really like opening our eyes to the world."
Lotus pond fibers in PNAS
They found a large number of unique fibers in the water. With the help of AI, most of the proteins were classified, but one type could not be made. "Finally, it was us humans, who looked at it for three hours with our eyes," before realizing it was sugar!
"These structures are entirely supported by sugar. It's really entering a new field." These structures that have never been observed before are called "biological dark matter" by Yan Ning.
Each structure brings up a lot of questions: How do sugars fold? What are their functions? So beautiful structures, so complex sugars, how are they synthesized? Even more serious issues: where do they come from? The protein sequences in these fibers consist of only a few repeating amino acids. How to search for them?
Yan Ning said, "The research paradigm of the laboratory is changing from problem-driven to observation-driven."
The most exciting thing is that they even discovered fibers consisting solely of sugar, without protein. Could it possibly be used for carbon neutrality in the future? Because synthesizing sugar via photosynthesis in living organisms requires many conditions for long-term storage, while this sugar fiber can be stably present in the natural environment, requiring fewer conditions. It can withstand temperature changes and various enzymatic cuts, perhaps becoming a new material, or even a carrier for information storage.
Various fibers in the lotus pond
In this new field, there are also technical developments and applications of AI. "Previously, we did one structure every two weeks, but now we do over a dozen structures in a day. We are trying to develop methods to make the electron microscope a high-throughput technology."
Observation Net also learned that Yan Ning's team published a paper online in October this year introducing a simple and efficient algorithm called Ahaha, specifically designed to determine the absolute chirality of sugar fibers in cryo-electron microscopy images.
Ahaha determines the absolute chirality of sugar fibers in BioRxiv
Cryo-electron microscopy (cryo-EM) combined with "seeking" (seek), this entire new method has a cool name — "CryoSeek (Cool Seek)."
The "impact" of AI is empowerment
Yan Ning said that the current electron microscopy technology is very advanced, "don't be afraid of electron microscopy", nor think it is expensive, many devices are open and shared.
But the large amount of information obtained by "CryoSeek" with cryo-EM still needs interpretation. The solution is to share data online and with global scholars; on the other hand, to train and use AI.
When AlphaFold emerged, some people worried that scholars like Yan Ning who solve protein structure problems might be affected. At the event, she said, "AI is for my use."
Originally, AI didn't know sugar, but now Yan Ning's team has developed an AI model that can automatically build sugar.
Yan Ning said that each technological revolution liberates people from relatively tedious work, allowing them to do what they want most, which is itself a great empowerment.
AI for Science is accelerating the promotion of scientific discovery, and biology may also empower AI (Biology for AI). "Our brain can process information so efficiently because of the billions of years of evolution since the beginning of life, resulting in a highly complex brain structure. Life converts solar energy into chemical energy, and after many years of energy storage, it has created this advanced structure."
Yan Ning envisioned that by referencing the brain, the hardware structure of future AI may also change. The empowerment of various disciplines to AI may be a deeper scientific question.
Original article: https://www.toutiao.com/article/7574706552746017334/
Statement: This article represents the views of the author. Please express your opinion by clicking the [Like/Dislike] button below.