Recently, Michael Levitt, Nobel Laureate in Chemistry and honorary dean of the Institute of Complex Systems and Multiscale Research of Fudan University, walked into the Guanghua Building of Fudan and gave two classes for undergraduates. This elective course is called "Introduction to the Frontiers of Quantitative Biophysics" and is taught by five professors, including Levitt and Ma Jianpeng. Quantitative biophysics, also known as computational biology, has a strong interdisciplinarity, and young people majoring in biology, physics, chemistry, mathematics, pharmacy, computer science and other majors can learn and become the reserve talents of this cutting-edge interdisciplinary discipline.
The Jiefang Daily Shangguan News reporter found that the undergraduate students who took this course came from a number of science and engineering majors, and many graduate students also came to listen to the lectures. "The accurate prediction of protein structure by 'AlphaFold' has upgraded computational biology from an auxiliary discipline to a dominant discipline, and its scientific and applied value is very large." Levitt's words sowed the seeds of engaging in cutting-edge basic research in the hearts of Fudan students.
Breakthroughs have been made in protein structure prediction
Levitt, vice president of the World's Leading Scientists Association, shared the Nobel Prize in Chemistry in 2013 with two other computational biology pioneers for "Creating Multiscale Models for Complex Chemical Systems." The award that year marked that the direction of using computer software to study biochemistry was highly recognized by the scientific community; in 2020, "Alpha Fold 2" was born, which marked the practical value of computational biology in protein structure prediction, drug research and development and other fields.
The Nobel laureate told the students that proteins are made of amino acids folded, and how the long chain of amino acids is spontaneously folded into a three-dimensional structure of proteins is a question that can be called the "crown jewel" of molecular biology. It is relatively easy for scientists to determine amino acid sequences, but it is difficult to resolve protein structures, because protein structure depends on the interaction forces between the atoms of thousands of amino acids. Can we develop algorithmically powerful computer software to accurately predict protein structure? Levitt studied the problem for 55 years.
With the maturity of artificial intelligence technologies such as deep learning, the problem of protein structure prediction has ushered in a breakthrough. Last November, The Alpha Fold 2, developed by Google's DeepMind, won the international protein structure prediction competition with a prediction accuracy very close to the level of experimental measurements. This incident caused a sensation in the scientific community, surpassing the "Alpha Go" defeat of Ke Jie and Li Sedol, and was named one of the "Top Ten Scientific Breakthroughs of 2020" by the American "Science" magazine.
The protein structure of the "Alpha Fold 2" prediction (blue) and the experimental assay (green) almost exactly coincided. Source: DeepMind
"I didn't expect to see this breakthrough in my lifetime!" Levitt sighed. In his view, the company's huge R & D team, the support of strong computing resources and the accumulation of previous research results are the reasons for the success of Google and deep thinking.
Sharing the "fate" of scientific research with college students
How do you build a huge computational biology team? Talent training is the first priority. Ma Jianpeng, dean of the Institute of Complex Systems and Multiscale Research at Fudan University, said that in order to cultivate young talents, this semester the Institute led the opening of the "Introduction to the Frontiers of Quantitative Biophysics". In addition to Levitt and him, three professors from the School of Mathematical Sciences, the Department of Chemistry, the School of Biological Medicine and the School of Life Sciences also teach, allowing undergraduates to feel the charm of cutting-edge interdisciplinarity.
As the biggest professor, Levitt's lectures are macro-oriented and very thoughtful. In addition to the history of computational biology and protein folding, he also shared his life insights such as the successful experience of winning the Nobel Prize and the environmental conditions for cultivating Nobel Prize-level talents. In response to a student's question about "how to find your scientific interests," he uttered a word— serendipity. He used his own life experience to explain what is the "fate" on the road of scientific research.
In the 1960s, After watching the Nobel Prize-winning John Kendrew tv show "The Thread of Life: An Introduction to Molecular Biology," Levitt wrote to Kendlow, who was working at Cambridge University, hoping to be his Doctoral student. Kendrew wrote back, however, that the number of doctoral students was full. A friend of Levitt's who learned of this told him: "In business, when someone says 'no' to you, you have to give him another option that can say 'yes'. So he wrote to Kendrew: "If you can't be your PhD student next year, can you do it later?" Seeing that the young man was so persistent, Kendrew invited him to Cambridge for an interview and recommended him to the Weizmann Institute of Science to study computational biology. At the institute, Levitt met Ariaye Vaschel, a collaborator with whom he shared the Nobel Prize 46 years later. "That's serendipity. You must be passionate about your career, persevere, be good at innovation, and be a kind person. ”
The Nobel Laureate's vivid and philosophical lectures made the students harvest full. "This class was eye-opening, for example, making me realize that the central laws of molecular biology can be understood from the perspective of reinforcement learning." Xiao Yangfan, a senior biology major, told reporters, "Michael's lecture method is also very distinctive, every quarter of an hour, we will discuss around the content we just heard for two or three minutes, which is very helpful for improving learning efficiency." ”
How to accelerate the development of computational biology
It is understood that Levitt's main job after coming to Shanghai in March this year is to guide the Fudan team to study computational biology. At present, the "OPUS-Fold" software developed by the Institute of Complex Systems multi-scale research of Fudan University surpasses the "Alpha Fold 2" in the prediction accuracy of protein side chain structure, and the relevant paper will be published in the Uk "Bioinformatics Bulletin". Ma Jianpeng introduced that most of the drug molecules targeting proteins interact with side chains, and only by obtaining high-precision side chain structures can drug molecules be designed, so this breakthrough is of great significance to the development of new drugs.
The Fudan team modeled the side chain based on the known main chain: the target structure in blue and the prediction structure in red.
However, from the perspective of the whole system, the "Alpha Fold 2" is still in the global leading position. Demis Hassabis, CEO of DeepIns, recently announced that the company has partnered with the European Institute of Bioinformatics to establish a protein structure database, providing data access to more than 350,000 protein structures. DeepThinkering plans to expand the database to more than 100 million protein structures, covering almost every protein known to man.
Alpha Fold 2 put computational biology in the fast lane, and many research teams are developing this kind of software. In July, the University of Washington team published a paper in the journal Science introducing the open-source software RoseTTAFold, which predicts not only protein structure, but also the binding form of multiple proteins. In August, the Stanford team published a science cover paper describing the breakthrough in artificial intelligence algorithms predicting the three-dimensional structure of RNA (ribonucleic acid).
Many scientists believe that the accurate prediction of protein structure is a scientific and technological revolution that will trigger industrial changes such as biomedicine and new materials. Yin Weihai, deputy dean of the Med-X Research Institute at Shanghai Jiao Tong University, said: "From predicting protein structure, finding drug targets to drug molecule design, to clinical trial design, artificial intelligence is fully entering the pharmaceutical industry and is beginning to play a key role. Scientists can also use artificial intelligence software to develop new materials, such as designing protein materials that do not exist in nature, for use in chemical, energy, environmental protection and other industries."
In Levitt's view, china can have more research universities offering computational biology courses for undergraduates, stimulating their interest and attracting them to this cutting-edge field of science and technology in the future. He also suggested that Shanghai accelerate the development of computational biology, provide more powerful computing resources for scientific research teams, and expand the scale of scientific research teams to accelerate the iteration of software with independent intellectual property rights such as "work folding".
Column Editor-in-Chief: Huang Haihua Text Editor: Yu Taoran
Source: Author: Yu Taoran