If you want to give AI In the field of 2021 The most breakthrough Award , Who will you choose ？《science》 and 《nature》 The answers are 「 Protein structure prediction 」.
as everyone knows , Long chain amino acids in proteins are distorted 、 Fold and interweave into a complex three-dimensional structure , These structures can be difficult to , It's not even possible to decipher . For decades, scientists have been hoping to simply predict the structure and shape of proteins through gene sequences , To open a new world with insight into the mechanism of life , But progress has been slow .
until DeepMind announce , For the first time, people have found a way to predict protein structure by calculation . Even without knowing the similar structure ,AI It can also accurately predict the structure of proteins at the atomic level .
DeepMind Express AlphaFold Protein structure can be predicted periodically with atomic accuracy , Technically, multi sequence alignment and deep learning algorithm are used to design , Combined with the physical and biological knowledge of protein structure, the prediction effect is improved .AlphaFold Our breakthrough research results will help researchers explore the mechanisms that cause some diseases , And design drugs for 、 The crops have increased , And biodegradable plastics 「 super enzyme 」 R & D paves the way .
AlphaFold In the paper 7 Published in June 《nature》 The journal , Address of thesis ：https://www.nature.com/articles/s41586-021-03819-2
In recent days, Alphafold One of the founders of John Jumper It was also named 《nature》2021 Top ten scientific figures of the year .
from 2018 Early generation AlphaFold In the international protein structure prediction competition （CASP） Cut a striking figure , To 2021 year AlphaFold2 Official open source ,John Jumper led DeepMind Our research team overcame many difficulties , Just let AlphaFold2 Realized 2/3 Outstanding achievements in protein structure prediction .
It is also the subject of protein structure prediction RoseTTAFold Also be 《science》 The selection of 2021 The annual breakthrough .
RoseTTAFold By the Institute of protein design, University of Washington School of Medicine （Institute for Protein Design） United with Harvard University 、 University of Texas Southwest Medical Center 、 University of Cambridge 、 A protein prediction tool based on deep learning developed by Lawrence Berkeley National Laboratory and other institutions .RoseTTAFold Achieved comparable results AlphaFold2 High accuracy of , And faster 、 The computer processing power required is also lower .
RoseTTAFold Published in 《science》 The journal , Address of thesis ：https://www.science.org/doi/abs/10.1126/science.abj8754
Structurally ,RoseTTAFold It's a three track （three-track） neural network , It means that it can take into account the pattern of protein sequence 、 How amino acids interact and the possible three-dimensional structure of proteins . In this structure , A one-dimensional 、 A two-dimensional 、 Three dimensional information flows back and forth , So that the neural network can centrally infer the chemical part of protein and its folding structure .
As people marvel at it , More than a decade ago, some scientists thought that the problem of protein structure prediction could never be solved , But today this has become a reality . The biggest breakthrough brought about by artificial intelligence is to 「 impossible 」 Turned into 「 Probably 」.
Expand on , It's not just a change in protein structure prediction ,AI There is still a lot of potential to be tapped in the whole scientific research field , This is also AI for Science The reason why this topic has attracted much attention this year , Such as AI + mathematics 、AI + chemical 、AI + medicine .
Maybe , There will be more in the next two years AI + Breakthroughs in scientific research , You can focus on .
author[Heart of machine],Please bring the original link to reprint, thank you.