Difference between revisions of "Machine Learning/Life sciences"

From Thalesians Wiki
(Created page with "The importance of machine learning in life sciences continues to grow. This page attempts to collect the most important research papers and developments in this area as mentio...")
 
 
Line 7: Line 7:
==2021.07.15 DeepMind's AI for protein structure is coming to the masses==
==2021.07.15 DeepMind's AI for protein structure is coming to the masses==


Ewen Callaway. Nature. 15 July 2021. <span class="plainlinks">https://www.nature.com/articles/d41586-021-01968-y</span>
Ewen Callaway. Nature. 15 July 2021. https://www.nature.com/articles/d41586-021-01968-y


'''Abstract:''' ''Machine-learning systems from the company and from a rival academic group are now open-source and freely accessible.''
'''Abstract:''' ''Machine-learning systems from the company and from a rival academic group are now open-source and freely accessible.''
Line 13: Line 13:
It's protein structure prediction for the people. Software that accurately determines the 3D shape of proteins is set to become widely available to scientists.
It's protein structure prediction for the people. Software that accurately determines the 3D shape of proteins is set to become widely available to scientists.


On 15 July, the London-based company DeepMind released an open-source version of its deep-learning neural network AlphaFold 2 and described its approach in a paper in ''Nature'' (<span class="plainlinks">https://doi.org/10.1038/s41586-021-03819-2</span>). The network <span class="plainlinks">[https://www.nature.com/articles/d41586-020-03348-4 dominated a protein-structure prediction competition last year]</span>.
On 15 July, the London-based company DeepMind released an open-source version of its deep-learning neural network AlphaFold 2 and described its approach in a paper in ''Nature'' (https://doi.org/10.1038/s41586-021-03819-2). The network [https://www.nature.com/articles/d41586-020-03348-4 dominated a protein-structure prediction competition last year].


Meanwhile, an academic team has developed its own protein-prediction tool inspired by AlphaFold 2, which is already gaining popularity with scientists. That system, called RoseTTaFold, performs nearly as well as AlphaFold 2, and is described in a paper in ''Science'' (<span class="plainlinks">https://doi.org/10.1126/science.abj8754</span>) also published on 15 July.
Meanwhile, an academic team has developed its own protein-prediction tool inspired by AlphaFold 2, which is already gaining popularity with scientists. That system, called RoseTTaFold, performs nearly as well as AlphaFold 2, and is described in a paper in ''Science'' (https://doi.org/10.1126/science.abj8754) also published on 15 July.


The open-source nature of the tools means that the scientific community should be able to build on the advances to create even more powerful and useful software, says Jinbo Xu, a computational biologist at the University of Chicago in Illinois, who was not involved in either effort.
The open-source nature of the tools means that the scientific community should be able to build on the advances to create even more powerful and useful software, says Jinbo Xu, a computational biologist at the University of Chicago in Illinois, who was not involved in either effort.
Line 45: Line 45:
With code now freely available for both RoseTTaFold and AlphaFold 2, researchers will be able to build on both advances, says Xu, and perhaps make the techniques amenable to protein structures that AlphaFold 2 has so far struggled to predict. Two areas of intense interest are predicting the structure of complexes of multiple interacting proteins and applying the software to the design of novel proteins.
With code now freely available for both RoseTTaFold and AlphaFold 2, researchers will be able to build on both advances, says Xu, and perhaps make the techniques amenable to protein structures that AlphaFold 2 has so far struggled to predict. Two areas of intense interest are predicting the structure of complexes of multiple interacting proteins and applying the software to the design of novel proteins.


doi: <span class="plainlinks">https://doi.org/10.1038/d41586-021-01968-y</span>
doi: https://doi.org/10.1038/d41586-021-01968-y


==2021.07.09 China Targets AI Dominance by 2030==
==2021.07.09 China Targets AI Dominance by 2030==


'''Bloomberg Quicktake. July 9th, 2021, 9:48 AM GMT+0100. <span class="plainlinks">https://www.bloomberg.com/news/videos/2021-07-09/china-targets-ai-dominance-by-2030-video</span>'''
'''Bloomberg Quicktake. July 9th, 2021, 9:48 AM GMT+0100. https://www.bloomberg.com/news/videos/2021-07-09/china-targets-ai-dominance-by-2030-video'''


<blockquote>
<blockquote>
Line 69: Line 69:
==2021.07.07 U.K. Deploys Fastest Supercomputer to Fight Dementia, MS==
==2021.07.07 U.K. Deploys Fastest Supercomputer to Fight Dementia, MS==


'''Adeola Eribake. Bloomberg. July 7, 2021, 12:01 AM GMT+1. <span class="plainlinks">https://www.bloomberg.com/news/articles/2021-07-06/nvidia-boots-up-u-k-supercomputer-to-boost-drug-research-nhs</span>'''
'''Adeola Eribake. Bloomberg. July 7, 2021, 12:01 AM GMT+1. https://www.bloomberg.com/news/articles/2021-07-06/nvidia-boots-up-u-k-supercomputer-to-boost-drug-research-nhs'''


'''Abstract:''' ''Developed in partnership with AstraZeneca, GlaxoSmithKline. Nvidia waiting for approval for its $40 billion takeover of Arm.''
'''Abstract:''' ''Developed in partnership with AstraZeneca, GlaxoSmithKline. Nvidia waiting for approval for its $40 billion takeover of Arm.''
Line 105: Line 105:
==2021.07.05 China beats Google to claim the world's most powerful quantum computer==
==2021.07.05 China beats Google to claim the world's most powerful quantum computer==


'''Matthew Sparkes. NewScientist. 5 July 2021. <span class="plainlinks">https://www.newscientist.com/article/2282961-china-beats-google-to-claim-the-worlds-most-powerful-quantum-computer/</span>'''
'''Matthew Sparkes. NewScientist. 5 July 2021. https://www.newscientist.com/article/2282961-china-beats-google-to-claim-the-worlds-most-powerful-quantum-computer/'''


<blockquote>
<blockquote>
Line 115: Line 115:
==2021.04.30 Goldman Sachs is betting on Artificial Intelligence to drive growth==
==2021.04.30 Goldman Sachs is betting on Artificial Intelligence to drive growth==


'''Disha Sinha. Analytics Insight. April 30, 2021. <span class="plainlinks">https://www.analyticsinsight.net/goldman-sachs-is-betting-on-artificial-intelligence-to-dive-growth/</span>'''
'''Disha Sinha. Analytics Insight. April 30, 2021. https://www.analyticsinsight.net/goldman-sachs-is-betting-on-artificial-intelligence-to-dive-growth/'''


<blockquote>
<blockquote>

Latest revision as of 22:33, 21 December 2021

The importance of machine learning in life sciences continues to grow. This page attempts to collect the most important research papers and developments in this area as mentioned in the Press.

Research papers

The Press

2021.07.15 DeepMind's AI for protein structure is coming to the masses

Ewen Callaway. Nature. 15 July 2021. https://www.nature.com/articles/d41586-021-01968-y

Abstract: Machine-learning systems from the company and from a rival academic group are now open-source and freely accessible.

It's protein structure prediction for the people. Software that accurately determines the 3D shape of proteins is set to become widely available to scientists.

On 15 July, the London-based company DeepMind released an open-source version of its deep-learning neural network AlphaFold 2 and described its approach in a paper in Nature (https://doi.org/10.1038/s41586-021-03819-2). The network dominated a protein-structure prediction competition last year.

Meanwhile, an academic team has developed its own protein-prediction tool inspired by AlphaFold 2, which is already gaining popularity with scientists. That system, called RoseTTaFold, performs nearly as well as AlphaFold 2, and is described in a paper in Science (https://doi.org/10.1126/science.abj8754) also published on 15 July.

The open-source nature of the tools means that the scientific community should be able to build on the advances to create even more powerful and useful software, says Jinbo Xu, a computational biologist at the University of Chicago in Illinois, who was not involved in either effort.

Structure to function

Proteins are made of strings of amino acids that, when folded into 3D shapes, determine the function of those proteins in cells. For decades, researchers have used experimental techniques such as X-ray crystallography and cryo-electron microscopy to determine protein structures. But such methods can be time-consuming and costly, and some proteins are not amenable to such analysis.

DeepMind sent shock waves through the scientific world last year, when it showed that its software could accurately predict the structure of many proteins using just the sequence of the proteins alone (which is determined by DNA). Researchers had been working on this challenge for decades, and AlphaFold 2 performed so well in a biennial protein-prediction exercise called CASP that the competition's co-founder declared that "in some sense the problem is solved".

DeepMind—which has a reputation for being cagey about its work—described AlphaFold 2 in a brief presentation at CASP on 1 December. It promised to publish a paper outlining the network in more detail and to make the software available to researchers, but said little else.

"Among academics, there was a fair amount of doom and gloom," says David Baker, a biochemist at the University of Washington in Seattle whose team developed RoseTTaFold. "If someone has solved the problem you're working on but doesn't disclose how they did it, how do you continue working on it?"

"I felt like I lost my job at the time," says computational chemist Minkyung Baek, a member of Baker's team. But DeepMind's presentation also spurred new ideas that Baek couldn't wait to explore. So she, Baker and their colleagues started brainstorming ways to replicate AlphaFold 2's success.

They identified several key advances, including how the network uses information about proteins that are evolutionarily related to the targets researchers are trying to predict, and how the predicted structures of one part of a protein can influence how the network handles sequences corresponding to other parts of the molecule.

RoseTTaFold not only performed nearly as well as AlphaFold 2 — but also much better than other CASP entries (including some from the Baker lab). It's not yet clear why it couldn't equal AlphaFold 2, but one possibility is DeepMind's expertise, says Baek. "We don't have any deep-learning engineers in our lab." Xu is impressed by the efforts of Baek, Baker and their collaborators, and suspects that DeepMind's success was down to its access to engineering expertise and superior computing power.

Speedy structures

DeepMind has also streamlined AlphaFold 2. Whereas the network took days of computing time to generate structures for some entries to CASP, the open-source version is about 16 times faster, says AlphaFold lead researcher John Jumper. It can generate structures in minutes to hours, depending on the size of the protein. That's comparable to the speed of the RoseTTaFold.

Although the source code for AlphaFold 2 is freely available — including to commercial entities — it might not yet be particularly useful for researchers without technical expertise. DeepMind has collaborated with select researchers and organizations, including the non-profit Drugs for Neglected Diseases initiative headquartered in Geneva, Switzerland, to predict specific targets, but it hopes to broaden access, says Pushmeet Kohli, head of AI for science at DeepMind. "There is a lot more we want to do in this space."

As well as making the code for RoseTTaFold freely available, Baker's team has set up a server into which researchers can plug a protein sequence and get a predicted structure. Since it was launched last month, the server has predicted the structure of more than 5,000 proteins submitted by around 500 people, says Baker.

With code now freely available for both RoseTTaFold and AlphaFold 2, researchers will be able to build on both advances, says Xu, and perhaps make the techniques amenable to protein structures that AlphaFold 2 has so far struggled to predict. Two areas of intense interest are predicting the structure of complexes of multiple interacting proteins and applying the software to the design of novel proteins.

doi: https://doi.org/10.1038/d41586-021-01968-y

2021.07.09 China Targets AI Dominance by 2030

Bloomberg Quicktake. July 9th, 2021, 9:48 AM GMT+0100. https://www.bloomberg.com/news/videos/2021-07-09/china-targets-ai-dominance-by-2030-video

Endless cups of coffee, hi-tech tools to help doctors, a masseuse who never gets tired... Just some of the latest innovations on display at this year's World Artificial Intelligence Conference in Shanghai.

"The advantages are first of all its trajectory is a digital record of master-level massage skills. Compared with ordinary masseurs, its techniques are more accurate and precise. Second, with the 3D images it can find different acupuncture points according to different people's body shapes and can massage in a targeted manner."

Meet Walker X, he could be your next butler.

"The functions that we have been able to achieve in the family setting include bringing you a Coke from the refrigerator, opening a drink, bringing you a newspaper, fetching an umbrella, and watering the flowers, playing chess with you, singing and doing homework with you. It can voice chat and has other emotional interactions. In the future, we will make robots that have a more diversified role in the entire home. Walker will play the role to help the family, accompany you as you grow up and accompany children and the elderly. It can be a teacher of the family as well as a helper who can do a variety of housework in the family.

AI is also helping drive medical progress. Morphogo is a bone marrow cytomorphology analysis system. It can make a diagnosis within 1 day compared to the usual 5 to 10 days with traditional hospital methods.

Yang Chuhu, ZhiWei Information Technology Co. Ltd.: "This kind of high-efficiency detection is crucial for accurate diagnosis and timely treatment which cannot be achieved only by manual work. Taking China for example, China has vast remote areas and it is impossible for people living in remote villages to go to a big hospital for a rapid morphological examination of bone marrow. Without such a system, it can be said that the life and health of patients with blood diseases in remote areas cannot be effectively guaranteed. This is not a problem for China alone. This is a global problem."

China has set its sights on building a domestic AI industry worth nearly $150 billion in the next few years and of becoming the leading AI power by 2030.

2021.07.07 U.K. Deploys Fastest Supercomputer to Fight Dementia, MS

Adeola Eribake. Bloomberg. July 7, 2021, 12:01 AM GMT+1. https://www.bloomberg.com/news/articles/2021-07-06/nvidia-boots-up-u-k-supercomputer-to-boost-drug-research-nhs

Abstract: Developed in partnership with AstraZeneca, GlaxoSmithKline. Nvidia waiting for approval for its $40 billion takeover of Arm.

Nvidia Corp. has built a U.K. supercomputer for scientists and drug makers that will allow researchers to model diseases, discover new medicines and advance knowledge of the human genome.

The Cambridge-1 will be the country's most powerful and one of the top 50 supercomputers worldwide. It was developed in partnership with AstraZeneca Plc and GlaxoSmithKline Plc, Nvidia said in a statement on Wednesday.

Other founding members include Guy's Hospital and St Thomas' Hospital trust, part of the U.K.'s National Health Service, as well as King's College London.

"Through this partnership, we will be able to use a scale of computational power that is unprecedented in health-care research," said Sebastien Ourselin, head of the School of Biomedical Engineering & Imaging Sciences at King's College London. "It will be truly transformational for the health and treatment of patients."

Oxford Nanopore, a DNA sequencing startup that's preparing to debut on the London Stock Exchange will also get access to Cambridge-1, allowing it to improve its artificial intelligence algorithms in hours instead of days, the startup said in the statement.

Rather than a single unit, the Cambridge-1 is made up of several modules working together. This parallel processing lets the computer carry out tasks simultaneously, allowing for greater speeds. It was also faster to build, coming together in about a quarter of the time it takes to construct similar machines.

The supercomputer has a computing capability of more than 400 petaflops. That's close to the world's fastest, the Fugaku machine in Japan. Top500, which publishes a twice-yearly list of the fastest supercomputers, included the Cambridge-1 at No. 41 on its most recent ranking, published at the end of June.

The speed and capabilities that supercomputers offer scientists are crucial as they try to make sense of unprecedented amounts of data. "We accumulated more data in one quarter in 2020 than in the last 300 years," said Kim Branson, GSK's senior vice president and global head of AI-ML.

The machine, which cost $100 million to build, was announced in October, a month after Nvidia launched a $40 billion bid for semiconductor designer Arm Ltd. Arm's technology is ubiquitous in smartphones and other consumer electronics, making the deal controversial among Nvidia's competitors, who rely on the designs. The U.K. has also expressed concerns that selling Arm, which is currently owned by Softbank Group Corp., has national security implications and the country's competition authority is due to submit a report at the end of the month.

Nvidia said it's already developing another supercomputer, which is to be made with Arm technology.

The Cambridge-1 will be kept at a site run by Kao Data Campus just outside London and about 40 miles south of its namesake city.

King's College and London hospital group Guy's and St Thomas' will use the new AI capability to learn from MRI scans and then generate synthetic brain images, creating models to help scientists better understand diseases such as dementia, stroke, brain cancer and multiple sclerosis. That could allow for earlier diagnosis and treatment, the university and teaching hospital said.

Nvidia said it would work with its partners to share the insights gained with the wider scientific community. For example, AstraZeneca will open up one of its projects, an AI-based model for chemical structures to help with drug discovery, to other scientists.

"Cambridge-1 will empower world-leading researchers in business and academia with the ability to perform their life's work on the U.K.'s most powerful supercomputer, unlocking clues to disease and treatments at a scale and speed previously impossible in the U.K.," said Jensen Huang, founder and chief executive officer of Nvidia, in a statement.

2021.07.05 China beats Google to claim the world's most powerful quantum computer

Matthew Sparkes. NewScientist. 5 July 2021. https://www.newscientist.com/article/2282961-china-beats-google-to-claim-the-worlds-most-powerful-quantum-computer/

A team in China has demonstrated that it has the world's most powerful quantum computer, leapfrogging the previous record holder, Google.

Jian-Wei Pan at the University of Science and Technology of China in Hefei and his colleagues say their quantum computer has solved a problem in just over an hour that would take the world's most powerful classical supercomputer eight years to crack, and may yet be capable of exponentially higher performance.

2021.04.30 Goldman Sachs is betting on Artificial Intelligence to drive growth

Disha Sinha. Analytics Insight. April 30, 2021. https://www.analyticsinsight.net/goldman-sachs-is-betting-on-artificial-intelligence-to-dive-growth/

Artificial Intelligence (AI) has taken a major role in acting as the main driver of upcoming hi-tech future in the world. It is shifting the information age to a completely new digital domain where upgraded machines help to solve critical decisions and assist in diverse sectors of a country. The banking and financial sector is very keen in investing in AI to protect their customers against the competitors in this market. Thus depending on the current scenario, Goldman Sachs, a leading American multinational investment bank, is betting on AI to drive growth in the economy. It has a fund of $72.5 million exclusively for the investment in Artificial Intelligence algorithms and data analytics.

There is a sudden surge in the cases against cyber-security, cyber-threats, phishing, spamming, hacking and many more unethical behaviours from dark web. It is due to the digital transformation in online banking through websites or mobile apps. Goldman Sachs deals with investment management, securities, asset management, prime brokerage and securities underwriting. This means banks need the best security possible against the online threats. Here comes AI to assist Goldman Sachs in the best possible way.

Why Goldman Sachs is betting on AI to drive growth?

  1. The constant fear of cyber-attacks has been reduced with the help of filtering application of AI. Yes, there are possibilities of receiving fraud applications with unethical motives. AI cybersecurity detects and blocks these applications while collecting real-time data from the users on a large scale.
  2. AI-enabled Investment Trust has already become a successful test for Goldman Sachs in enabling the best investment options for the customers. This has improved customer engagement while running through various kinds of analyst reports and news reports. The investment trust runs on NLP (Natural Language Processing) to assist asset managers in identifying undervalued shares and probable opportunities for profits.
  3. Partnership with an AI-based startup, H2O.ai, helps to focus on deep machine learning transparency and model interpretability to predict a better future. It assists in decision-making process for finance department and equity trading floor such as market making, providing liquidity to the bank and many more.
  4. This is the best opportunity to increase growth in hyper-personalised banking through conversational AI. AI will help in 24*7 two-way communication with innumerable personalised responses and feedbacks to the users.

Goldman Sachs prefers to focus on the net profit over the investment cost of the AI. The bank is eager to maintain its brand image through efficient customer experience, upgraded security and service enhancement.