The 2024 Nobel Prize in Chemistry is deservedly awarded to Professor David ·, director of the Institute for Protein Design at the University of Washington in United States, for his outstanding achievements in computational protein design, and the DeepMind team for breakthrough protein structure prediction.
As a PhD student and postdoc at Beck, I would like to talk about my understanding of the field of computational structures and how I see Professor Baker as well.
Picture丨The author of this article, Shen Hao (left) and Professor David · (Source: Shen Hao)
I learned in my biochemistry course at university that sequencing the human genome revealed a large number of DNA sequences, and the protein sequences encoded by these sequences were largely determined.
However, the function of the vast majority of proteins is not well understood, because there is a more important layer beyond the sequence information, and that is the three-dimensional structure of the protein.
One of the major discoveries of the Nobel Prize in Chemistry in 1972 was that the three-dimensional structure of proteins is determined by one-dimensional sequences. Through a long evolutionary process, nature has screened out a series of protein sequences that spontaneously fold in water to form protein structures through a rational arrangement of hydrophobic and hydrophilic amino acids.
However, the complexity of physically calculating this process is extremely high, and the energy state is difficult to accurately describe, which has always been an unsolved mystery in science and is known as the real "code of life".
Baker received his undergraduate from Harvard University in United States, majoring in philosophy and minoring in biology. After graduating from university, he took a year off from school to travel around the world, during which he also visited China, and can be said to be the first batch of pioneer international friends who "checked-in" China after China's reform and opening up.
Later, he decided to find a way to accumulate research and joined the laboratory of Randy ·Schekman, a professor at the University of United States California, Berkeley, and winner of the 2013 Nobel Prize in Physiology and Medicine, to study protein trafficking within cells.
As a postdoctoral fellow, he joined the lab of David · at the University of California, San Francisco, United States, where he studied protein structures.
When Baker first came to the faculty at the University of Washington, his research focused on protein folding. His idea was that if the process of protein folding could be slowed down, subtle changes might be observed.
However, experiments have found that this method does not work, and once the protein reaches the critical point of folding, the folding process is instantaneous. But this did not discourage him, because in the process, he stumbled upon some short peptide chains that could form stable local structures.
At this time, he had an idea: since it is too complicated to calculate the folding of the entire protein, it is better to start with the local peptide structure and see if the whole folding process can be done by assembling.
So, Baker and his team broke up the proteins in the protein database (PDB) into small peptides, and used the energy function to determine whether to accept each step by randomly inserting the assembly method, and developed the Rosetta protein structure prediction software like building blocks.
这个软件在 2004 年第六届蛋白质结构预测大赛（CASP，Critical Assessment of Structure Prediction）上“一战成名”，成为当时的标杆。 但由于能量函数的准确性和搜索空间的限制，蛋白质结构预测依然是个难题。
So Baker and Brian Kuhlman, then ·postdoc, proposed:
Now that Rosetta's protein conformational space search and energy assessment have made some progress, why not do the opposite? Starting from the structure, the protein is designed, and then the protein sequence that can be folded into the structure is calculated.
In 2003, they designed the first Top7 protein with a new folding method, ushering in a new era of computational protein design.
Since then, the Baker laboratory has designed a wide range of proteins with a wide variety of functions and morphologies, from chemical catalytic enzymes, to drug-binding protein targets, small molecule binding proteins, to nanomaterials.
In the process, artificial neural networks for deep learning have gradually emerged.
2014 年，贝克的博士生谢尔盖·奥夫奇尼科夫（Sergey Ovchinnikov）和博士后赫图南丹·卡米塞提（Hetunandan Kamisetty）基于深度学习原理，利用同源序列共进化信息改进了蛋白质结构预测，显著提高了准确性。
Later, the DeepMind team took it a step further and used the co-evolution information directly as an "energy function" to develop AlphaFold1, which improved the prediction accuracy to nearly 60% in the 2018 CASP competition.
Subsequently, the optimization of AlphaFold2 in 2020 further improved the accuracy to more than 90%, and most of the protein structures were successfully predicted.
Photo丨Professor David · Baker (Source: Shen Hao)
Baker Labs has also continued to make progress in the field of machine learning, developing Message Passing Neural Networks (MPNN) based on language models, which can generate optimal sequences based on protein structure and complete important steps in protein design.
In addition, the Baker team developed their own open-source program, RosettaFold, based on the AlphaFold framework, and launched RFDiffusion in combination with the diffusion model to generate novel protein structures. The combination of the two becomes two core tools for protein design.
Of course, the field of protein design is vast, and different application scenarios require different structural designs, which not only requires unbridled imagination, but also requires teams with in-depth understanding of specific problems and solid experimental techniques.
In my opinion, Baker was a pure scientist who put all his mind on science. He rarely travels to conferences, and spends most of his time walking around the lab, discussing cutting-edge scientific issues and giving directions.
One of his characteristics is his lack of patience, which may not sound like a virtue at first glance, but it turns into an advantage when applied to scientific research: he has the courage to innovate and take great strides forward, focusing only on important and challenging problems.
In scientific research, we encounter many difficulties every day, but Baker is positive and pragmatic, and the difficulties in scientific research are like boulders encountered on the mountaineering road, going around and turning over, in short, nothing can stop him from moving forward.
Picture丨Professor David · Baker and members of the research group in the Rainier Snow Mountain (Source: Shen Hao)
In addition to his rapid progress in scientific research, Baker is also an outdoor expert in life: he works hard in the middle of the week, enjoys hiking, camping, and skiing on the weekends, and I also set new records for single-day climbing and step counting. Not only does Baker love to explore new routes, but he always finds a different challenge and never seems to go off the beaten path.
Baker is very smart, clear-thinking, and always able to get to the heart and essence of the problem through the details. He is a strong learner and always full of curiosity, even if it is a complex problem across fields, he can learn quickly and touch on the class. He is also excellent in writing and verbal skills, and is often able to write and revise articles quickly and accurately.
In addition, his high emotional intelligence, broad-mindedness, and ability to collaborate and share have attracted a large number of talented people to join his research group, forming a vibrant protein design family.
In my opinion, Baker's lab is very flat, and he leads a team of 100 people by himself, and the team members spontaneously form groups based on research interests. He regularly meets with each member one-on-one to keep track of everyone's progress and keep the project moving.
He also attaches great importance to communication and cooperation within the team, believing that collective wisdom is far more important than individual strength, and that in this cross-learning and free and open environment, everyone can maximize their creativity.
I first met Baker 12 years ago, when Brian Kobilka, a United States·professor at Stanford University at Stanford University who had just won the Nobel Prize in Chemistry that year, came to Seattle University of Washington to give a lecture.
When I was a third-year exchange student at Tsinghua University, I was already familiar with Beck's research and his outstanding contributions to protein structure prediction and design.
After the lecture, I plucked up the courage to introduce myself to Baker and expressed my desire to join his research group. Although Baker felt that a semester was too short to do much work, he offered me to join the lab and study with young PhD students. In this passionate laboratory, I feel the joy of scientific research.
After that, I successfully applied back to the University of Washington at Seattle to pursue a Ph.D. in Baker's group. For the first time, my Ph.D. project realized the de novo design of self-assembled protein fibers.
During my postdoc, I continued to design controllable fiber assemblies, including through pH, small peptide chain conformational changes, multi-component protein fibers, and functional fibers.
In the future, my research group will continue to focus on protein fibers, including the design of protein nanowires, controllable assembly, and new nanomaterials such as protein machines.
In addition, it can also be used for signal amplification for single-molecule detection, cryo-EM structure elucidation assisted by helical symmetry, purification and separation of carbon nanotubes, functional molecular arrangement, antigen presentation-enhancing antibody generation, and cell biology research.
Resources:
https://www.bakerlab.org/future-faculty/#hao-shen
Operation/Typesetting: He Chenlong