February 3, 2023

Medical Trend

Medical News and Medical Resources

Science丨Breakthrough in 2021: Protein Structure Prediction

Science丨Breakthrough in 2021: Protein Structure Prediction


Science丨Breakthrough in 2021: Protein Structure Prediction. 


The balance in 2021 has been less than half a month, and the Science editorial department issued an article entitled Protein structures for all , unveiling a major breakthrough in the scientific research community this year: the use of artificial intelligence to predict the structure of proteins .


Science丨Breakthrough in 2021: Protein Structure Prediction
Christian Anfinsen, winner of the Nobel Prize in Chemistry in 1972


In his acceptance speech for the Nobel Prize in 1972, American biochemist Christian Anfinsen once proposed an idea: one day, we can predict the structure of any protein only by the sequence composition of amino acids.

There are thousands of proteins in the human body, and the prediction of protein structure will help provide new insights for basic biology and reveal new drug targets with clinical significance.

In recent years, researchers have discovered that artificial intelligence-driven software can produce thousands of protein structures, thus turning Anfinsen’s dream into reality.

It is also an annual breakthrough in the scientific community in 2021.

In the past, protein structure could only be determined by laborious laboratory analysis.

Protein is the main force in the realization of biological functions, and is involved in the processes of muscle contraction, energy conversion, oxygen transport, and resistance to microbial invasion.

Although the functions of proteins are different, the basic composition is similar. 20 different amino acids form peptide chains, which are further assembled into a unique and complex three-dimensional structure.

The three-dimensional structure of proteins determines how these proteins interact with other molecules, as well as their role in cells.

In the 1950s, researchers began to analyze and draw the three-dimensional structure of proteins through X-rays, which has gradually become one of the mainstream techniques in the field of protein structure analysis today.

Based on the X-ray diffraction structure, the Protein Data Bank (http://www.rcsb.org/pdb/) , a database in the field of protein structure, was constructed , which contains about 185,000 laboratory-analyzed protein structures [1] .

But the structure of each protein requires several years, and it also requires hundreds of thousands of experimental funds.


Science丨Breakthrough in 2021: Protein Structure PredictionPDB database


In order to accelerate the progress of this process, scientists began to build computer models in the 1970s to predict how a given protein would fold.

At first, this was only possible for small proteins or short fragments of large proteins.

By 1994, the computer model had been developed enough to initiate a biennial protein structure prediction competition (Critical Assessment of Protein Structure Prediction, CASP) .

The organizer provided the modeler with the amino acid sequences of dozens of proteins.

At the end of the event, the modeler’s results were compared with the latest experimental data of X-ray crystallography and emerging technologies such as nuclear magnetic resonance spectroscopy and cryo-electron microscopy.

A score of more than 90 points was considered to be the same as the experimentally solved structure. But in the beginning, everyone’s score for structural prediction was less than 60 points.

By 2018, the modeler’s score is usually only around 70 points.


Science丨Breakthrough in 2021: Protein Structure Prediction


Subsequently, the artificial intelligence software AlphaFold debuted. This program was developed by Google’s sister company, DeepMind, which trains itself through experiments to solve the structured database.

In the first game, its median score was close to 80 points, and in 90 games with other algorithms, it won 43 games. In 2020, its successor AlphaFold2 performed even better [2] .

AlphaFold2 is driven by a network of 182 processors optimized for machine learning, and its average score is 92.4 points, which is equivalent to experimental technology.

Scientists at DeepMind reported that they did the same experiment on 350,000 proteins found in the human body, accounting for 44% of all known human proteins [3] .

In the next few months, they expect their database to increase to 100 million proteins, covering almost all species, half of the total number of proteins available.

The next step for DeepMind is to predict which proteins work together and how they interact.

Subsequently, in the October preprint, the organization’s scientists announced 4433 protein-protein complexes, revealing how proteins interact.

AlphaFold2’s code is currently publicly available, helping other scientists to make better use of the tool.

In November last year, researchers in Germany and the United States used AlphaFold2 and cryo-electron microscopy to map the structure of the nuclear pore complex, which is composed of 30 different proteins that control the way into the nucleus.

In August of this year, Chinese researchers used AlphaFold2 to map the structure of nearly 200 proteins that bind to DNA.

These proteins may be involved in everything from DNA repair to gene expression (for details, please refer to: Science | Ride the Wind and Waves!

AI deep plough protein complex prediction ;

14 expert comments丨Nature shockedly released the 98.5% human protein structure prediction results completed by AlphaFold ;

Nature pays special attention to AlphaFold2—”It’s here, it’s here, it’s coming with the source code”, the predicted protein structure reaches the atomic level accuracy ).

Now, scientists studying SARS-CoV-2 are also using AlphaFold2 to simulate the impact of the new coronavirus Omicron variant on the spike protein mutation [4] .

Mutations change the shape of the protein, which may be enough to prevent antibodies from binding to it and neutralize the virus, which may become the starting point for new vaccines or drug development.

The structure of a protein is not static, it will produce bending and various deformations during the working process, and modeling these changes is still a challenge.

In addition, it is not easy to visualize the structure of large multi-protein complexes in cells. Therefore, the field of protein structure prediction and analysis is still promising.

Technological advances driven by artificial intelligence this year have provided an unprecedented perspective, a moment that will change biology and medicine forever.



Breakthrough selection for 2021


1. The era of DNA extraction from ancient soils has arrived, revealing the evolution and migration activities of ancestor humans and species.


2. The COVID-19 antiviral drugs are on the stage:

the COVID-19 vaccination is gradually being rolled out. This fall, major manufacturers Merck and Pfizer respectively announced the results of clinical trials of anti-coronavirus oral drugs developed and demonstrated good efficacy.

Merck developed Molnupiravir as a nucleoside analogue, which reduced hospitalization and mortality of COVID-19s by 30% (50% was the number declared for the first time, and Merck later corrected this number to 30%).

Pfizer’s oral antiviral drug PAXLOVID (PF-07321332+Ritonavir), PF-07321332 is a new coronavirus 3CL protease inhibitor, and Ritonavir (aspartic protease inhibitor) can reduce the hospitalization rate and mortality rate by 89%.


3. “Psychedelic rock” in the treatment of traumatic stress syndrome:

In May this year, Nature Medicine published a study: talk therapy and ecstasy (MDMA) helped PTSD patients survive difficult times and achieved significant results.

At present, hallucinogenic drugs enter clinical trials to help treat depression, anxiety and addiction and other diseases.


Science丨Breakthrough in 2021: Protein Structure Prediction


4. Artificial antibodies “tame” infectious diseases:

Because of advances in clone screening, animal models and crystal structure analysis, antibody screening and optimization have become more convenient.

Antibody therapy and research and development have entered the field of infectious diseases. This year, a number of COVID-19 antibody drugs have shown their strength.

In addition, antibody drugs against influenza, Zika virus, respiratory syncytial virus, HIV and Plasmodium are being developed.

The FDA has approved antibody treatments against Ebola, inhalation anthracnose, Clostridium difficile infection, respiratory syncytial virus in infants, and HIV-infected people who have failed all drugs.


5. Direct CRISPR gene editing therapy in vivo:

In June this year, NEJM magazine published the experimental results of Intellia Therapeutics and Regeneron Pharmaceuticals.

For hereditary transthyretin amyloidosis, four weeks after the gene editing system entered the body, the blood of six patients was transferred Thyroid protein was reduced by 52% and 87%, respectively.

In September, Editas Medicine announced the results of gene therapy for congenital amaurosis.

Through direct eye injection of the virus packaging gene editing system, some effects were observed in two patients, which may help the patients to see the light again in the future.


6. Breakthrough in in vitro embryo culture helps to explore the secrets of early embryo development: through a special rotating instrument, scientists have extended the in vitro culture time of mouse embryos from 3-4 days to 11 days. Scientists can also obtain blastocysts from human embryonic stem cells and induced pluripotent stem cells. The research results of these stem cells will help us further explore the secrets of early embryonic development.


There are three other studies entering the annual breakthrough candidate:

7. The nuclear fusion reaction is at a new height, and the gains and losses are broken even.

9. “Insight” Mars rover revealed the internal geological structure of Mars.

10. Muzi’s behavior is abnormal, and the standard model of particle physics has cracks.


There are still many disappointments in science in 2021. Science specifically pointed out three:

1. Global climate cooperation is still struggling.

2. The first Alzheimer disease-targeted drug aducanumab was approved, but it did not usher in cheers, and received many doubts and great controversies.

3. Many scientists are subject to great hostility, personal attacks and threats.


In the end, the COVID-19 has swept the world for two years, and there are still many low- and middle-income countries that cannot obtain enough vaccines. This battle is long and difficult, and victory is difficult to hope. I hope that more and better drugs will appear in the breakthrough in the coming year!






Original link:

1. Berman, HM et al. The Protein Data Bank. Nucleic acids research 28, 235-242, doi:10.1093/nar/28.1.235 (2000).
2. Cramer, P. AlphaFold2 and the future of structural biology. Nature structural & molecular biology 28, 704-705, doi:10.1038/s41594-021-00650-1 (2021).
3. Tunyasuvunakool, K. et al. Highly accurate protein structure prediction for the human proteome. Nature 596, 590-596, doi:10.1038/s41586-021-03828-1 (2021).
4. Kryshtafovych, A. et al. Modeling SARS-CoV-2 proteins in the CASP-commons experiment. Proteins 89, 1987-1996, doi:10.1002/prot.26231 (2021).

Science丨Breakthrough in 2021: Protein Structure Prediction. 

(source:internet, reference only)

Disclaimer of medicaltrend.org