AI scientists Ching-Yung Lin and his startup Graphen, mapped out the evolution of the virus Source: Graphen AI
Interview and writer Chen Fangyu
At 11 o'clock on the night of March 9th, a LINE group gathered Taiwanese and American AI medical experts, someone uploaded an image-a little bit of green and red sparks cracked from the bottom right corner to the surrounding, pulling out a white light. It looks like a New Year's firework, and it looks like a starry trajectory.
"Up to March 5, the world's approximately 250 trees have been given gene sequence evolution of the virus," the speaker is Graphen CEO Lin Yong. As of March 11, the AI virus gene evolution tree , has shown nearly 400 virus strains from more than 30 countries.
Everyone in the artificial intelligence world knows Ching-Yung Lin —he had been with IBM for nearly 17 years, and was the Chief Scientist and Founder of the Supercomputer Department at Watson Network and Artificial Intelligence. He is also an Adjunct Professor at Columbia University. He was featured on the cover of BusinessWeek. In 2017, he founded AI startup Graphen in New York.
"Taiwan has two of the largest primitive types, two of which are prototypes of a large number of offspring variants on the left. This variant will spread, one of which will continue to Singapore and then continue to Australia; the other to the United Kingdom , And another one went to Guangdong, "Ching-Yung Lin explained the virus lineage to more than two hundred people in the group.
It turns out that a red dot represents a virus that has been genetically sequenced; a green dot represents a group of viruses with a common genetic sequence; and a white line is a virus association. Click on a green dot to show the path of infection of this group of viruses.
"It's very clear who transmitted who, and through which mutations," said Ching-Yung Lin. In this way, it is not necessary to infer the source of infection from contact history and travel history; even in community infections where the source cannot be found, the path of infection can be seen.
Gene map of the global pneumonia gene intelligently drawn by Graphen (as of March 14)
Europe and the United States each have a group of super infectious virus strains
At the end of February, Ching-Yung Lin found someone discussing the classification and source of the virus. After looking for reports and papers, he found that "the problem to be solved in the public health sector, we can solve it!" Then he found the genetic data of 250 virus strains in the GISAID database. After a week running out the relationship diagram with AI, and spending a few days editing the user interface, I now have the current look, and I can search according to variables such as country and gender.
Since Graphen has done medical treatment, money laundering, and social communication, it deals with millions of connected big data, and is good at analyzing the relationship and interaction between individuals and the whole; this time there are only hundreds of viruses, which is not technically difficult.
As soon as this picture was published, it immediately attracted the attention of major American TV media such as Flowserve, NBC, ABC, and Bloomberg. Foreign interview.
Spreading out this intricate map, we can see that two strains of highly infectious virus-one strain spread in China, Australia, and Asian countries one month later, and mutated into another virus on January 24 and passed into Washington State. ; The other was introduced into Germany on January 28, broke out in Italy, Finland, Switzerland, the Netherlands and other countries at the end of February, and finally spread throughout Europe.
Virus strains causing major infections in Washington State (arrowhead)
Ching-Yung Lin explained to the reporter of Future City with a sharing screen that the infection source of each state in the United States comes from a virus strain that is very close to the same prototype. After entering the United States, these patients may be quarantined. The virus stopped spreading; only the patient in Washington State caused a latent infection, and one month later, February 20, caused a localized infection in the local area, and finally there were 11 mutations; one of the mutant strains also caused Spread on the California Princess Cruises (Diamond Princess in Japan was caused by a prototype virus).
From the news, it is speculated that the cluster infection occurred in a nursing home in Seattle; San Francisco, California, and Seattle, Washington, are major technology towns in the western United States. Businesses are frequent, and the virus may spread to California and cause spread.
As of March 14, the most severe outbreaks were in Washington and New York, with 509 and 421 confirmed cases. New York State has not uploaded the virus strain and no way of knowing the transmission of the infection; however, US media reports that 50 patients in New York State can trace to the same source of infection-a 50-year-old lawyer working near Manhattan Central Station for participating in a large The event, which infected more than 100 people in less than 10 days, brought New Rochelle, a small town 25 miles from New York City, to an emergency closure.
"When encountering a highly infectious virus, you must take positive measures, such as closing the city." Ching-Yung Lin believes that visualizing the transmission route can highlight certain virus strains that spread particularly quickly and remind the epidemic prevention units to pay attention.
Viral strains circulating mainly in Europe (arrows)
Taiwanese women infected with European virus, or brought Asian virus to Europe?
Look at Europe again. It is possible that because of the borders of the country, the virus is like a firework, and there are three main virus groups and multiple small branches.
Three viruses reported from Italy, the Netherlands, and Wales in the UK were mutated and transmitted to Switzerland, and the other went north to Finland and Luxembourg; another virus originally came from Australia and spread throughout Europe, and finally Infected Taiwanese women traveling to Egypt. This also confirms that Egypt believes that it was the woman who transmitted the Taiwan virus to Egypt, but in fact the woman was infected with the European strain of virus locally.
In fact, the Asian epidemic has gradually subsided. Seven strains of viruses were uploaded in Taiwan, but it can be seen from the map that none of these viruses have continued to spread to the outside world. It may be because the isolation has been effective and has not caused community infection.
In contrast, in Europe, infections have been increasing recently. On March 13th, Germany uploaded 10 viruses collected at the end of February in one breath. It turned out that a brand-new virus (which has nothing to do with other pandemic disease strains in other European countries) has grown from its source, and has been widely transmitted in Germany.
As for the most controversial discussion recently, where did the virus come from? Ching-Yung Lin believes that because the virus mutation will increase with the number of infections, if the virus develops in the United States earlier than China but is mistaken for influenza, it should detect more complex and diverse mutations in China than in many strains . However, this phenomenon has not been found with 94 virus strains sequenced in the United States compared with 172 viruses in China; if more virus strains are sequenced in the future, it will be clearer in science.
Limited by the virus strains uploaded by governments of all countries (such as 114 strains uploaded by the Dutch government and laboratories, Italy only uploaded 8 viruses), this picture may not show the current status of the epidemic situation; For virus strains, the shape of the association graph also changes.
To be answered: On which host is the virus susceptible to mutation?
After completing the map of global virus strains, Ching-Yung Lin saw the news of "Beijing found to attack the central nervous system" and made him curious: Is Wuhan pneumonia virus more likely to mutate in unhealthy or elderly hosts?
Ching-Yung Lin explained that the old man's gene replication process is prone to mutation and cancer. The Washington state virus, which produces 11 mutations in a short period of time, and patients may be in nursing homes, making him associate.
What's more notable in this complex global infection network is the two sets of viruses in Europe and Washington State.
"These two viruses have a lot of mutations," explained Ching-Yung Lin. Most of the virus groups in Asia were "tightly held" and stopped after two or three mutations; but the European strains flew like dandelions. Go out and make a lot of variation.
Some European governments believe that Wuhan pneumonia is like flu. Sixty to seventy percent of people will be okay after infection. "This is a statement without public health knowledge!" Ching-Yung Lin explained that the virus will not only spread, but also in the host during the transmission process. The body mutates, "The more you pass, the more you mutate, the more you have the opportunity to accumulate new attack capabilities. This is where the virus is really terrible!" He emphasized in the video conference software Zoom, "Can't let it pass!"
The development of vaccines and drugs is faster and more flexible than virus mutation. Ching-Yung Lin hopes that this map can help biomedicine determine the appropriate virus strains for easy development of vaccines, and allow governments of all countries to adopt the most practical public health policy immediately.
He also called on governments and laboratories to actively provide virus decoding. In addition to better understanding the evolution and transmission path of the virus, if the patient's course and treatment data are combined with analysis of plans, there may even be an opportunity to find more virus mutation groups It is related to the course of disease and treatment, and has practical help for the treatment method.
Technology Toolbox: Taiwan AI Scientist Ching-Yung Lin
After graduating from the Graduate School of Electrical Engineering at the National Taiwan University in 1970, he obtained a Ph.D. in Electrical Engineering from Columbia University, specializing in Network Science—observing the relationship between individuals and the whole from a macro perspective to accurately predict individual behavior.
In early 2000, he first used this architecture to analyze Internet information as a counter-terrorism tool; later he used it to analyze the interpersonal network of IBM employees, and calculated that each additional friend of an employee can increase the company's income by $ 986 a year. Jump onto the cover of BusinessWeek. In 2011, he became the first Fellow in the field of network science of the International Institute of Electrical and Electronics (IEEE).
"Human behavior is not independent, and it is usually related to a certain model," explained Ching-Yung Lin. The difficulty of this learning lies in dealing with the interaction effects.
Therefore, he is not only dealing with big data, but also linked big data; this relationship is like human brain neurons and connected nerves. It is necessary for computers to perform complex memory like human brain, Thinking, observing, and judging require graph computing technology.
After more and more research, Ching-Yung Lin found that the biggest challenge was to program "interaction" into a model. When Graphen was established, he recruited colleagues who were supercomputers at IBM into the team and spent several years rebuilding the underlying architecture of the computer to speed up the calculation.
Source: Original story in Chinesehttps://futurecity.cw.com.tw/article/1319