Connect with us

Uncategorized

DeepMind’s protein-folding AI has solved a 50-year-old grand challenge of biology

Published

on

For years DeepMind has notched up a streak of wins, showcasing AIs that have learned to play a variety of complex games with superhuman skill, from Go and StarCraft to Atari’s entire back catalogue. But Demis Hassabis, DeepMind’s public face and co-founder, has always stressed that these successes were just stepping stones towards a larger goal: AI that actually helps us understand the world.

Today DeepMind and the organizers of the long-running Critical Assessment of protein Structure Prediction (CASP) competition announced an AI that will have a huge impact on science. The latest version of DeepMind’s AlphaFold, a deep-learning system that can accurately predict the structure of proteins to within the width of an atom, has cracked one of biology’s grand challenges. “It’s the first use of AI to solve a serious problem,” says John Moult at the University of Maryland, who leads the team that runs CASP.

A protein is made up of a ribbon of amino acids that folds itself up with many complex twists and turns and tangles. This structure determines what it does. And figuring out what proteins do is key to understanding the basic mechanisms of life, when it works and when it doesn’t. Efforts to develop vaccines for covid-19 have focused on the virus’s spike protein, for example. The way the coronavirus snags onto human cells depends on the shape of this protein and the shapes of the proteins on the outsides of those cells. The spike is just one protein among billions across all living things; there are tens of thousands of different types of protein inside the human body alone.      

In this year’s CASP, AlphaFold predicted the structure of dozens of proteins with a margin of error of just 1.6 angstroms—that’s 0.16 nanometers. This far outstrips all other computational methods and for the first time matches the accuracy of experimental techniques to map out the structure of proteins in the lab, such as cryo-electron microscopy, nuclear magnetic resonance and x-ray crystallography. These techniques are expensive and slow: it can take hundreds of thousands of dollars and years of trial and error for each protein. AlphaFold can find a protein’s shape in a few days.

But identifying a protein’s structure is very hard. For most proteins, researchers have the sequence of amino acids in the ribbon but not the contorted shape they fold into. And there are typically an astronomical number of possible shapes for each sequence. Researchers have been wrestling with the problem at least since the 1970s, when Christian Anfinsen won the Nobel prize for showing that sequences determined structure.

The breakthrough could help researchers design new drugs and understand diseases. In the longer term, predicting protein structure will also help design synthetic proteins, such as enzymes that digest waste or produce biofuels. Researchers are also exploring ways to introduce synthetic proteins that will increase crop yields and make plants more nutritious.

Shocking advance

“It’s a very substantial advance,” says Mohammed AlQuraishi, a systems biologist at Columbia University who has developed his own system for predicting protein structure. “It’s something I simply didn’t expect to happen nearly this rapidly. It’s shocking, in a way.”

“This really is a big deal,” says David Baker, head of the Institute for Protein Design at the University of Washington and leader of the team behind Rosetta, a family of protein analysis tools. “It’s an amazing achievement, like what they did with Go.”

The launch of CASP in 1994 gave the field of predicting protein structures a boost. Every two years, the organizers release 100 or so amino acid sequences for proteins whose shapes have been identified in the lab but not yet made public. Dozens of teams from around the world then compete to find the correct way to fold them up using software. Many of the tools developed for CASP are already used by medical researchers. But progress was slow, with two decades of incremental advances failing to produce a shortcut to painstaking lab work.   

When DeepMind entered the competition in 2018 with its first version of AlphaFold, it gave CASP the jolt it was looking for. It still could not match the accuracy of a lab but it left other computational techniques in the dust. Researchers took note: soon many were adapting their own systems to work more like AlphaFold.

This year more than half of the entries use some form of deep learning, says Moult. The accuracy overall was higher as a result. Baker’s new system, called trRosetta, uses some of DeepMind’s ideas from 2018. But it still came a “very distant second,” he says.

In CASP, results are scored using what’s known as a global distance test (GDT), which measures on a scale from 0 to 100 how close a predicted structure is to the actual shape of a protein identified in lab experiments. The latest version of AlphaFold scored well for all proteins in the challenge. But it got a GDT score above 90 for around two thirds of them. Its GDT for the hardest proteins was 25 points higher than the next best team, says John Jumper, who heads up the AlphaFold team at DeepMind. In 2018 the lead was around six points.

A score above 90 means that any differences between the predicted structure and the actual structure could be down to experimental errors in the lab rather than a fault in the software. It could also mean that the predicted structure is a valid alternative configuration for a protein to the one identified in the lab, within the range of natural variation.

According to Jumper, there were four proteins in the competition that the independent judges had not finished working on in the lab and AlphaFold’s predictions helped them get the correct structure.

AlQuraishi thought it would take researchers 10 years to get from AlphaFold’s 2018 results to this year’s, where the margin of error is as small as a single atom. This is close to the physical limit for how accurate you can get, he says. “These structures are fundamentally sloppy. It doesn’t make sense to talk about resolutions much below that.”

Puzzle pieces

AlphaFold builds on the work of hundreds of researchers around the world. DeepMind also drew on a wide range of expertise, putting together a team of biologists, physicists and computer scientists. Details of how it works will be released this week at the CASP conference and in a peer-reviewed article in a special issue of the journal Proteins next year. But we do know that it uses a form of attention network, a deep-learning technique that allows an AI to train by focusing on parts of a larger problem. Jumper compares the approach to assembling a jigsaw: it pieces together local chunks first before fitting these into a whole.

DeepMind trained AlphaFold on around 170,000 proteins taken from the protein data bank, a public repository of sequences and structures. It compared multiple sequences in the data bank and looked for pairs of amino acids that often end up close together in folded structures. It then uses this data to guess the distance between pairs of amino acids in unknown structures. Training took “a few weeks”, using computing power equivalent to between 100 and 200 GPUs.

Dame Janet Thornton at the European Bioinformatics Institute in Cambridge, UK, has been working on the structure and function of proteins for 50 years. “That’s really as long as this problem has been around,” she said in a press conference last week. “I was beginning to think it would not get solved in my lifetime.”

Many drugs are designed by simulating their 3D molecular structure and looking for ways to slot these molecules into target proteins. Of course, this can only be done if the structure of those proteins is known. This is the case for only a quarter of the roughly 20,000 human proteins, says Thornton. That leaves 15,000 untapped drug targets. “AlphaFold will open up a new area of research.”

DeepMind says it plans to study leishmaniasis, sleeping sickness and malaria, all tropical diseases caused by parasites, because they are linked to lots of unknown protein structures.

One drawback of AlphaFold is that it is slow compared to rival techniques. AlQuraishi’s system, which uses an algorithm called a recurrent geometrical network (RGN), can find protein structures a million times faster—returning results in seconds rather than days. Its predictions are less accurate, but for some applications speed is more important, he says.

Researchers are now waiting to find out exactly how AlphaFold works. “Once they describe to the world how they do it then a thousand flowers will bloom,” says Baker. “People will be using it for all kinds of different things, things that we can’t imagine now.”

Even a less accurate result would have been good news for people working on enzymes or bacteria, says AlQuraishi: “But we have something even better, with immediate relevance to pharmaceutical applications.”

Lyron Foster is a Hawaii based African American Musician, Author, Actor, Blogger, Filmmaker, Philanthropist and Multinational Serial Tech Entrepreneur.

Continue Reading
Comments

Uncategorized

The biggest step the Biden administration took on climate yesterday wasn’t rejoining the Paris Agreement

Published

on

While the Biden Administration is being celebrated for its decision to rejoin the Paris Agreement in one of its first executive orders after President Joe Biden was sworn in, it wasn’t the biggest step the administration took to advance its climate agenda.

Instead it was a move to get to the basics of monitoring and accounting, of metrics and dashboards. While companies track their revenues and expenses and monitor for all sorts of risks, impacts from climate change and emissions aren’t tracked in the same way. Now, in the same way there are general principals for accounting for finance, there will be principals for accounting for the impact of climate through what’s called the social cost of carbon.

Among the flurry of paperwork coming from Biden’s desk were Executive Orders calling for a review of Trump era rule-making around the environment and the reinstitution of strict standards for fuel economy, methane emissions, appliance and building efficiency, and overall emissions. But even these steps are likely to pale in significance to the fifth section of the ninth executive order to be announced by the new White House.

That’s the section addressing the accounting for the benefits of reducing climate pollution. Until now, the U.S. government hasn’t had a framework for accounting for what it calls the “full costs of greenhouse gas emissions” by taking “global damages into account”.

All of this is part of a broad commitment to let data and science inform policymaking across government, according to the Biden Administration.

Biden writes:

“It is, therefore, the policy of my Administration to listen to the science; to improve public health and protect our environment; to ensure access to clean air and water; to limit exposure to dangerous chemicals and pesticides; to hold polluters accountable, including those who disproportionately harm communities of color and low-income communities; to reduce greenhouse gas emissions; to bolster resilience to the impacts of climate change; to restore and expand our national treasures and monuments; and to prioritize both environmental justice and the creation of the well-paying union jobs necessary to deliver on these goals.”

The specific section of the order addressing accounting and accountability calls for a working group to come up with three metrics: the social cost of carbon (SCC), the social cost of nitrous oxide (SCN) and the social cost of methane (SCM) that will be used to estimate the monetized damages associated with increases in greenhouse gas emissions.

As the executive order notes, “[an] accurate social cost is essential for agencies to accurately determine the social benefits of reducing greenhouse gas emissions when conducting cost-benefit analyses of regulatory and other actions.” What the Administration is doing is attempting to provide a financial figure for the damages wrought by greenhouse gas emissions in terms of rising interest rates, and the destroyed farmland and infrastructure caused by natural disasters linked to global climate change.

These kinds of benchmarks aren’t flashy, but they are concrete ways to determine accountability. That accountability will become critical as the country takes steps to meet the targets set in the Paris Agreement. It also gives companies looking to address their emissions footprints an economic framework to point to as they talk to their investors and the public.

The initiative will include top leadership like the Chair of the Council of Economic Advisers, the director of the Office of Management and Budget and the Director of the Office of Science and Technology Policy (a position that Biden elevated to a cabinet level post).

Representatives from each of the major federal agencies overseeing the economy, national health, and the environment will be members of the working group along with the representatives or the National Climate Advisor and the Director of the National Economic Council.

While the rule-making is proceeding at the federal level, some startups are already developing services to help businesses monitor their emissions output.

These are companies like CarbonChainPersefoni, and SINAI Technologies. And their work compliments non-profits like CDP, which works with companies to assess carbon emissions.

Biden’s plan will have the various agencies and departments working quickly. The administration expects an interim SCC, SCN, and SCM within the next 30 days, which agencies will use when monetizing the value of changes in greenhouse gas emissions resulting from regulations and agency actions. The President wants final metrics will be published by January of next year.

The executive order also restored protections to national parks and lands that had been opened to oil and gas exploration and commercial activity under the Trump Administration and blocked the development of the Keystone Pipeline, which would have brought oil from Canadian tar sands into and through the U.S.

“The Keystone XL pipeline disserves the U.S. national interest. The United States and the world face a climate crisis. That crisis must be met with action on a scale and at a speed commensurate with the need to avoid setting the world on a dangerous, potentially catastrophic, climate trajectory. At home, we will combat the crisis with an ambitious plan to build back better, designed to both reduce harmful emissions and create good clean-energy jobs,” according to the text of the Executive Order. “The United States must be in a position to exercise vigorous climate leadership in order to achieve a significant increase in global climate action and put the world on a sustainable climate pathway. Leaving the Key`12stone XL pipeline permit in place would not be consistent with my Administration’s economic and climate imperatives.”

Continue Reading

Uncategorized

Ars online IT roundtable today: What’s the future of the data center?

Published

on

Ars online IT roundtable today: What’s the future of the data center?

Enlarge

If you’re in IT, you probably remember the first time you walked into a real data center—not just a server closet, but an actual raised-floor data center, where the door wooshes open in a blast of cold air and noise and you’re confronted with rows and rows of racks, monolithic and gray, stuffed full of servers with cooling fans screaming and blinkenlights blinking like mad. The data center is where the cool stuff is—the pizza boxes, the blade servers, the NASes and the SANs. Some of its residents are more exotic—the Big Iron in all its massive forms, from Z-series to Superdome and all points in between.

For decades, data centers have been the beating hearts of many businesses—the fortified secret rooms where huge amounts of capital sit, busily transforming electricity into revenue. And they’re sometimes a place for IT to hide, too—it’s kind of a standing joke that whenever a user you don’t want to see is stalking around the IT floor, your best bet to avoid contact is just to badge into the data center and wait for them to go away. (But, uh, I never did that ever. I promise.)

But the last few years have seen a massive shift in the relationship between companies and their data—and the places where that data lives. Sure, it’s always convenient to own your own servers and storage, but why tie up all that capital when you don’t have to? Why not just go to the cloud buffet and pay for what you want to eat and nothing more?

Read 4 remaining paragraphs | Comments

Continue Reading

Uncategorized

Transforming the energy industry with AI

Published

on

For oil and gas companies, digital transformation is a priority—not only as a way to modernize the enterprise, but also to secure the entire energy ecosystem. With that lens, the urgency of applying artificial intelligence (AI) and machine learning capabilities for optimization and cybersecurity becomes clear, especially as threat actors increasingly target connected devices and operating systems, putting the oil and gas industry in collective danger. The year-over-year explosion in industry-specific attacks underscores the need for meaningful advancements and maturity in cybersecurity programs.

However, most companies don’t have the resources to implement sophisticated AI programs to stay secure and advance digital capabilities on their own. Irrespective of size, available budget, and in-house personnel, all energy companies must manage operations and security fundamentals to ensure they have visibility and monitoring across powerful digital tools to remain resilient and competitive. The achievement of that goal is much more likely in partnership with the right experts.

MIT Technology Review Insights, in association with Siemens Energy, spoke to more than a dozen information technology (IT) and cybersecurity executives at oil and gas companies worldwide to gain insight about how AI is affecting their digital transformation and cybersecurity strategies in oil and gas operating environments. Here are the key findings:

  • Oil and gas companies are under pressure to adapt to dramatic changes in the global business environment. The coronavirus pandemic dealt a stunning blow to the global economy in 2020, contributing to an extended trend of lower prices and heightening the value of increased efficiency to compensate for market pressures. Companies are now forced to operate in a business climate that necessitates remote working, with the added pressure to manage the environmental impact of operations growing ever stronger. These combined factors are pushing oil and gas companies to pivot to new, streamlined ways of working, making digital technology adoption critical.
  • As oil and gas companies digitalize, the risk of cyberattacks increases, as do opportunities for AI. Companies are adding digital technology for improved productivity, operational efficiency, and security. They’re collecting and analyzing data, connecting equipment to the internet of things, and tapping cutting-edge technologies to improve planning and increase profits, as well as to detect and mitigate threats. At the same time, the industry’s collective digital transformation is widening the surface for cybercriminals to attack. IT is under threat, as is operational technology (OT)—the computing and communications systems that manage and control equipment and industrial operations.
  • Cybersecurity must be at the core of every aspect of companies’ digital transformation strategies. The implementation of new technologies affects interdependent business and operational functions and underlying IT infrastructure. That reality calls for oil and gas companies to shift to a risk management mindset. This includes designing projects and systems within a cybersecurity risk framework that enforces companywide policies and controls. Most important, they now need to access and deploy state-of-the-art cybersecurity tools powered by AI and machine learning to stay ahead of attackers.
  • AI is optimizing and securing energy assets and IT networks for increased monitoring and visibility. Advancements in digital applications in industrial operating environments are helping improve efficiency and security, detecting machine-speed attacks amidst the complexity of the rapidly digitalizing operating environments.
  • Oil and gas companies look to external partners to guard against growing cyberthreats. Many companies have insufficient cybersecurity resources to meet their challenges head-on. “We are in a race against the speed of the attackers,” Repsol Chief Information Officer Javier García Quintela explains in the report. “We can’t provide all the cybersecurity capabilities we need from inside.” To move quickly and address their vulnerabilities, companies can find partners that can provide expertise and support as the threat environment expands.

Cybersecurity, AI, and digitalization

Energy sector organizations are presented with a major opportunity to deploy AI and build out a data strategy that optimizes production and uncovers new business models, as well as secure operational technology. Oil and gas companies are faced with unprecedented uncertainty—depressed oil and gas prices due to the coronavirus pandemic, a multiyear glut in the market, and the drive to go green—and many are making a rapid transition to digitalization as a matter of survival. From moving to the cloud to sharing algorithms, the oil and gas industry is showing there is robust opportunity for organizations to evolve with technological changes.

In the oil and gas industry, the digital revolution has enabled companies to connect physical energy assets with hardware control systems and software programs, which improves operational efficiency, reduces costs, and cuts emissions. This trend is due to the convergence of energy assets connected to OT systems, which manage, monitor, and control energy assets and critical infrastructure, and IT networks that companies use to optimize data across their corporate environments.

With billions of OT and IT data points captured from physical assets each day, oil and gas companies are now turning to built-for-purpose AI tools to provide visibility and monitoring across their industrial operating environments—both to make technologies and operations more efficient, and for protection against cyberattacks in an expanded threat landscape. Because energy companies’ business models rely on the convergence of OT and IT data, companies see AI as an important tool to gain visibility into their digital ecosystems and understand the context of their operating environments. Enterprises that build cyber-first digital deployments similarly have to accommodate emerging technologies, such as AI and machine learning, but spend less time on strategic realignment or change management.

Importantly, for oil and gas companies, AI, which may have once been reserved for specialized applications, is now optimizing everyday operations and providing critical cybersecurity defense for OT assets. Leo Simonovich, vice president and global head of industrial cyber and digital security at Siemens Energy, argues, “Oil and gas companies are becoming digital companies, and there shouldn’t be a trade-off between security and digitalization.” Therefore, Simonovich continues, “security needs to be part of the digital strategy, and security needs to scale with digitalization.”

To navigate today’s volatile business landscape, oil and gas companies need to simultaneously identify optimization opportunities and cybersecurity gaps in their digitalization strategies. That means building AI and cybersecurity into digital deployments from the ground up, not bolting them on afterward.

Download the full report.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

Continue Reading

Trending