Cover story │ Epidemic spreads globally, AI identifies super virus strain

Posted by World Journal on Mar 22, 2020

Alt Text Graph representation of the monitoring of the Coronavirus evolution. (Graphen official website)

Alt Text Ching-Yung Lin (right) took a group photo with staff. On the table was Adam, a research robot. (Xiong Chuanhui / Photography)

Since the outbreak of the Coronavirus in December 2019, confirmed cases have appeared globally. Governments of all countries have made every effort to curb the spread of the epidemic, maintain economic and social stability, and soothe the mind. Dr. Ching-Yung Lin, founder and CEO of Graphen, Inc., who specializes in artificial intelligence (AI) and Big Data Analysis, is thinking about: Variations and results of Coronavirus in the world to help track virus strains and develop vaccines?

Viral mutation during growth and replication

On March 11, Graphen released the "Monitoring New Corona Virus Evolution" platform (Monitoring COVID-19), announcing hundreds of variants of the global Coronavirus and how it spreads it globally, making it the first to provide an analysis and prevention platform.

Ching-Yung Lin has spent 20 years in the field of artificial intelligence and big data, and has accumulated many research and development achievements. He reviewed the process of developing the "Monitoring the Evolution of Coronavirus" platform, saying that this research was not supported by government departments or medical institutions. He just wanted to complete the tracking and mutation of virus strains and share it with more researchers. Virus research contributes. At the same time, this is an example of artificial intelligence supporting the development of precision medicine.

The plan to build a platform to monitor the evolution of the Coronavirus was launched on March 1. Ching-Yung Lin led the team to analyze the global Coronavirus sequencing data from December 2019 on the basis of the whole genome sequence path analysis and put it on one platform. The virus data was shared by laboratories around the world.

Ching-Yung Lin's team analyzed COVID-19 virus SARS-CoV-2 and more than 400 virus strains around the world for one week, and analyzed more than 50 known whole-genome sequencing variants to understand the mutations of each virus. And find out how these mutations change at nearly 30,000 Genetic Locations, can identify the evolutionary chain and clusters of the virus, and determine the process of mutation and spread. Analysis shows that the mutation of the virus is "mostly caused by growth and replication in the human body."

Find subtypes in only one week

At the beginning of March, there were 250 virus strain samples, and by 9 am on March 5, a total of 272 virus strain samples were accumulated. Ching-Yung Lin said that 10 of the 272 samples were genes of bats and pangolins, not human genes; after removing these 10, there were 262 virus strain samples.

Finding the answer to the mutation from analyzing the virus strains seems difficult to understand. Ching-Yung Lin explained that genes are biological mechanisms that have been replicated in human cells. The concept of gene mutation is that genes are located on chromosomes and are constantly replicated in the body. During this process, there may be mistakes in writing, and therefore mutations will occur.

The virus is the same. After entering the human body, it continues to grow through division and replication, and then spreads out, so the virus will also mutate. Moreover, the genome may be repeated or missing some fragments, so after aligning to get the 30,000 gene loci of these 262 virus strains, you can find out the normality of each gene locus. What is the value, when all these values are combined, a gene sequence can be obtained. This sequence does not necessarily correspond to an exact virus strain, but we can call it the genome of the prototype virus. Among these Coronaviruses, 90 virus strains are indeed prototype viruses, so these can also be called first-generation viruses.

Ching-Yung Lin said: "This is a more complicated process." He and the team studied day after day. The whole project was completed on March 10, only one week. On March 11th, Graphen released the world 's first artificial intelligence AI "monitoring the Coronavirus evolution" platform. Based on the analysis of the viral gene sequence path, each virus strain was compared with the prototype virus strain. It can be clearly seen that From which virus strain has become which virus strain, and how these hundreds of strains have evolved around the world, several subtypes of the virus with strong transmission have been analyzed at the same time.

Super communicators influence Europe

He used two possible super communicators as examples to illustrate the discovery of the "Monitoring the evolution of the Coronavirus" platform.

First, in the cases confirmed in Wuhan on January 5, the virus strains later appeared in Taiwan, Belgium, and Australia, and most of the initial cases in Washington state belonged to the same virus. Another virus, confirmed in Wuhan on January 24, spread to Germany, Italy and Switzerland, eventually affecting Europe as a whole.

Second, multiple generations of mutant virus strains appeared in Beijing and South Korea; people infected with Coronavirus on the Diamond Princess cruise ship tested positive, and most of the viruses in their bodies were prototype viruses of Coronavirus.

Ching-Yung Lin showed a virus evolution map. Each red dot represents a genetically sequenced virus strain, and each green dot is a collection of identical mutations. The arrows between the green dots point to evolution. The link between the red dot and the green dot represents which set this virus strain is. The analysis showed that after clearing the exact mutations of all virus strains, it was possible to find out which one of the previous generations of each virus was.

He said that Coronavirus is a new disease, but the virus will mutate, and there may be two consequences after the mutation, one is the change in transmission ability and the other is the change in attack ability. His monitoring research, with the help of scientific analysis of network technology, can intuitively see the global spread, mutation and spread of the virus, and understand the relationship between sequence changes and disease epidemics.

Now, laboratories have to figure out which types of Coronavirus are spreading, "but it is not enough to target only the prototype virus." Therefore, we must find the virus strain that will spread more and develop vaccines against the genes of the virus. Ching-Yung Lin said that the "Monitoring the Evolution of Coronavirus" platform can provide researchers with a reference for clinical treatment, drug and vaccine development, and also help public health units predict the speed of virus transmission.

Find descendants from source

Ching-Yung Lin is from Taiwan, and received a bachelor's and master's degree in electrical engineering from Taiwan University, and a PhD in electrical engineering from Columbia University in 2000. He later joined IBM and focused on graph computing and AI technology research and development. He established a network science and machine intelligence department at IBM's TJ Watson Labs, developed dozens of patents, and became the chief scientist. In 2011, he was promoted to the International Institute of Electrical and Electronics Engineers (IEEE) FELLOW, the first FELLOW in the field of IEEE network science. The IEEE has a great influence on the definition standards of the industry, with members from 175 countries and 420,000 people.

His research interests are mainly multi-modal signal analysis, network analysis, and the use of artificial intelligence in the social and cognitive sciences. In 2017, he left IBM to start a business, and set up Graphen in New York with a new venture, established the next-generation artificial intelligence (AI) platform, created industry solutions, provided AI solutions for financial services, healthcare and network security.

Alt Text Dr. Ching-Yung Lin, founder and CEO of Graphen. (Xiong Chuanhui / Photography)

Taking this development of the "Monitoring New Corona Virus Evolution" platform as an example, Ching-Yung Lin explained the contribution of AI and big data to the medical field. From these more than 400 strains, 90 virus strains were found to have exactly the same sequence, that is, the original virus. Among them, many of the largest strains of the same strain came from Hong Kong, Shenzhen, and Wuhan, and the next one was linked to Singapore, and the virus was completely seen. In which city did it mutate, and the relationship between the strain and other strains. It was further found that two cases in Italy were very similar to one in Singapore.

He described: "We can find the offspring of a virus strain from the source data."

Coronavirus has developed to date, and there is no ready-made toolkit to analyze the virus. The background of the members of Graphen is computer engineering. Ching-Yung Lin led them to obtain virus data from laboratories around the world. The team quickly wrote Code, analyzed big data, completed the analysis of the viral gene sequence path, put the information on the company's website and shared it with everyone. He said that with scientific basis, it can be done quickly.

Cases have been discussed in the medical community after the outbreak. As he plans, he compares his work with published papers. The 10 cases at the end of January and 80 cases at the end of February were roughly the same after comparison.

Future medical precision personalization

The development of a "Monitoring Coronavirus Evolution" platform was not supported by government departments or medical institutions. He said that science must be applicable and contribute to humankind; understanding the Coronavirus is a top priority, and he hopes that the information provided by this platform can be helpful to medical and public health.

He has worked with the Institute of Chemistry, Academia Sinica, Taiwan to study the impact of genetic evolution on lung cancer by detecting genetic proteins and predicting the direction of cancer. He said, "Future medicine is personalized precision medicine." Genes in the body will be mutated. Gene mutations are the cause of cancer. If you can know a person's genetic changes, it can help you prescribe the right medicine.

Soon after starting his business, Ching-Yung Lin bought a research robot from a French company, about 2 feet tall, and spent 10,000 yuan. This robot acts as a doctor and is named Adam. The team set up a medical database and a patient database for Adam. The user stated his symptoms to Adam. After receiving the voice information and reading it, Adam understood the medical history and past treatment methods and medications, and studied which direction the disease would take.

Next, Adam organizes the results and suggestions from the database, and then sends them back to the user with a voice message, tells you what disease you may have, and predicts the results and risks caused by different drugs. "This is the application of big data," he said.

Ching-Yung Lin said that the characteristic of big data is that all the data is analyzed very finely, not general analysis. For example, how a person 's behavior (such as consumption and health) really evolves. After analysis, he can recommend products to him, or he can prescribe the right medicine. These are all personalized directions.

Alt Text Graphen demonstrated the robot Adam. (Xiong Chuanhui / Photography)

Multiple applications of robotics research

He shared his big data analysis course at Columbia University online. From 2015 to 2017, if Baidu searched for the keyword "big data analysis", Ching-Yung Lin has always ranked first, and he can find five or six hundred projects. He uses open source to make it available to everyone.

Big data is the foundation for the development of AI. If you describe AI in one sentence, Ching-Yung Lin said that AI is a machine that can do things that humans can do, that is, displaying AI capabilities through robots. It is a kind of display and combination. Now there are many housework applications, such as sweeping robots.

However, the robot goes one step further. The robot has movements and interactions with people. If you speak, it will still face you. The robot pet dogs that have been developed now have a feeling of companionship; the application of robots in long-term care can also be expected, such as assisting with heavy housework.

From 2012 to 2015, while at IBM, Ching-Yung Lin led a team of 40 people and participated in the largest social media analysis project in the United States. He completed 26 tasks and published more than 100 papers. Some topics have never been studied. For example, how to judge fake news and who made rumors on the Internet; for example, put images on the Internet, predict how far these images can spread, and how they will feel to viewers.

Image recognition is currently all nouns, such as square, tall, and round. His research is "the feeling that the image brings to the viewer", whether it is excited, sad, calm, or scary. This is also the first team working on image adjectives. Descriptive image is to hope to establish a model of emotion (Emotion), which is very important for the spread of multimedia information.

Data analysis finds Unknowns

This research is also the first time that robots can have human senses and values. This has an important role in the development of robot functions.

Ching-Yung Lin's entrepreneurial goal is to build the next generation of artificial intelligence platform, which is to "imitate the function of the whole brain." He said: "A robot must have artificial intelligence, not only to speak, but also to be able to talk. This development direction can convince people." If the robot develops a personality, you can design it to do different things based on personality, then it can do One thing can certainly handle more than five times.

Alt Text Graphen gradually imitates the function of the whole brain to establish the next generation of artificial intelligence. (Graphen official website)

Graphen's achievements in the financial field include digital finance, asset security management, and money laundering prevention. In terms of digital finance, Graphen has developed a virtual trader platform. Based on the five personality trait theories of psychology, each individual trader is designed with a different personality, and the trader will automatically trade. Traditionally, most people don't know the personality of a trader, they just buy it on the phone. If you don't like it, you can change it; but if the trading behavior is based on personality, investors can choose a virtual trader as the It helps you make decisions. This virtual platform is being tested and is scheduled to go live this year.

In terms of money laundering prevention, Graphen built a system monitoring wall for a New York bank. Through database analysis, we can find out some "Unknown Unknowns" that have occurred. Ching-Yung Lin said that financial institutions do not know the existence of money laundering and the methods of money laundering, but based on a large amount of data, they can find some signs and clues that could not be found. ". Now he is discussing the deployment of this system with some financial institutions in Asia.

Deciphering the origin with AI is human

As a scientist, Ching-Yung Lin's caring and research direction has always been "people", especially combining information technology and social science to solve human challenges or using information technology to help people live healthier lives. He said that the human thinking process itself is the scientific development of human beings, "in fact, it is deciphering." He says. If the metaphysical things are represented by actual numbers, discussing what is going on is a manifestation of human intelligence. In decryption, AI can think and judge, "to a certain extent, machines can do better than humans."

Human development stems from different creations. Can people be happier, more convenient, safer, and healthier? That is how civilization develops. AI can become an important core of the development of human civilization. If robots can think and judge, they can decrypt the characteristics of philosophy, psychology, and linguistics one by one, and slowly solve more complex problems.

For a long time, Ching-Yung Lin used computer to solve problems, sometimes at the social level, and sometimes at the behavior of individuals. He started to enter the medical field two years ago, hoping to analyze the society with the human face of scientists, and from a verifiable perspective Come on. "What he is doing now is something everyone is paying attention to. Like this Coronavirus research, there is no institutional sponsorship and no actual revenue, but he hopes to decrypt and share it for reference in treatment and research and development. He said, "My origin is still human."

Alt Text (From the official website of Graphen)

Source: Original story in Chinese: