ＡＩが予測したヒトタンパク質の画期的データベース、オンラインで公開

【パリＡＦＰ＝時事】生体を構成する主成分であるタンパク質の構造をＡＩ（人工知能）に学習・予測させたこれまでで最も網羅的なデータベースが先週、オンライン上で公開された。画期的な成果は「生物学研究を根本から変える」と注目を集めている。（写真はＡＩ「アルファフォールド」が予測したヒトタンパク質の立体構造の一例。カレン・アーノット氏作成。欧州分子生物学研究所提供）
　公開されたのは、ヒトのゲノム（全遺伝情報）によって発現するタンパク質、約２万種類のデータベース。米ＩＴ大手グーグルの親会社アルファベット傘下のＡＩ開発企業ディープマインドと、欧州分子生物学研究所（ＥＭＢＬ）が無料で一般公開した。
　生物の細胞はどれもタンパク質が絶え間なく発する指令によって、健康を維持したり、感染を防いだりするための機能を発揮させる。
　ゲノムとは異なり、プロテオーム（発現し得る全タンパク質）は遺伝的指令や環境の刺激によって常に変化している。
　タンパク質が細胞内でどのように作用するか、つまり、「折りたたみ」と呼ばれる立体構造のどの形に落ち着くかを理解する研究は、何十年にもわたって科学者たちを魅了してきた。
　しかし、それぞれのタンパク質の正確な機能を実験によって直接明らかにすることは非常に困難で、過去５０年間で、ヒトのプロテオームのアミノ酸のうちわずか１７％しか明らかになっていない。アミノ酸はタンパク質の構成成分。
　今回のデータベースを構築するために科学者らは、アミノ酸配列に基づいてタンパク質の形状を正確に予測する最新の機械学習プログラムを使用。既知のタンパク質構造１７万種類のデータベースをＡＩ「アルファフォールド」に学習させ、人間で発現し得る全タンパク質のうち５８％の構造を予測した。
　遺伝性疾患や抗菌剤耐性の研究など、応用できる可能性は極めて大きいと期待される。【翻訳編集ＡＦＰＢＢＮｅｗｓ】

〔ＡＦＰ＝時事〕（2021/08/03-11:24）

Scientists on Thursday unveiled the most exhaustive database yet of the proteins that form the building blocks of life, in a breakthrough observers said would fundamentally change biological research.
Every cell in every living organism is triggered to perform its function by proteins that deliver constant instructions to maintain health and ward off infection.
Unlike the genome -- the complete sequence of human genes that encode cellular life -- the human proteome is constantly changing in response to genetic instructions and environmental stimuli.
Understanding how proteins operate -- the shape in which they end up, or fold into -- within cells has fascinated scientists for decades.
But determining each protein's precise function through direct experimentation is painstaking.
Fifty years of research have until now yielded only 17 percent of the human proteome's amino acids, the subunits of proteins.
On Thursday, researchers at Google's DeepMind and the European Molecular Biology Laboratory (EMBL) unveiled a database of 20,000 proteins expressed by the human genome, freely and openly available online.
They also included more than 350,000 proteins from 20 organisms such as bacteria, yeast and mice that scientists rely on for research.
To create the database, scientists used a state-of-the-art machine learning programme that was able to accurately predict the shape of proteins based on their amino acid sequences.
Instead of spending months using multi-million dollar equipment, they trained their AlphaFold system on a database of 170,000 known protein structures.
The AI then used an algorithm to make accurate predictions of the shape of 58 percent of all proteins within the human proteome.
This more than doubled the number of high-accuracy human protein structures that researchers had identified during 50 years of direct experimentation, essentially overnight.
The potential applications are enormous, from researching genetic diseases and combating anti-microbial resistance to engineering more drought-resistant crops.
- 'Protein-folding problem' -
Paul Nurse, winner of the 2001 Nobel Prize for Medicine and director of the Francis Crick Institute, said Thursday's release was a great leap for biological innovation.
With this resource freely and openly available, the scientific community will be able to draw on collective knowledge to accelerate discovery, ushering in a new era for AI-enabled biology, he said.
John McGeehan, director for the Centre for Enzyme Innovation at the University of Portsmouth, whose team is developing enzymes capable of consuming single-use plastic waste, said AlphaFold had revolutionised the field.
What took us months and years to do, AlphaFold was able to do in a weekend. I feel like we have just jumped at least a year ahead of where we were yesterday, he said.
The ability to predict a protein's shape from its amino acid sequence using a computer rather than experimentation is already helping scientists in a number of research fields.
AlphaFold is already being used in research into cures for diseases that disproportionately affect poorer countries.
One US-based team is using the AI prediction to study ways of overcoming strains of drug-resistant bacteria.
Another group is using the database to better understand how SARS-CoV-2, the virus that causes Covid-19, bonds with human cells.
Venki Ramakrishnan, winner of the 2009 Nobel Prize for Chemistry, said Thursday's research, published in the journal Nature, was a stunning advance in biological research.
He said AlphaFold had essentially solved the so-called protein-folding problem, which argued that the 3D structure of a given protein should be determinable from its amino acid sequence, and which had puzzled scientists for half a century.
Given that the number of shapes a protein could theoretically take is astronomically large, the protein-fold problem was partly one of processing power.
The task was so daunting that in 1969 US molecular biologist Cyril Levinthal famously theorised that it would take longer than the age of the known universe to enumerate all possible protein configurations using brute calculation.
But with AlphaFold capable of performing a mind-dizzying number of calculations every second, the problem stood no chance when faced with AI and algorithms.
It has occurred long before many people in the field would have predicted, Ramakrishnan said.
It will be exciting to see the many ways in which it will fundamentally change biological research.