DESY News: Spreading of Covid-19 and the dynamics of the universe


News from the DESY research centre news_suche news_search eng 1 1 8 both 0 1 %Y/%m/%d Press-Release

Spreading of Covid-19 and the dynamics of the universe

How big data might help to predict immunity

What does the spreading of a pandemic and the evolution of the universe have in common? Nothing!, one is willing to say immediately. However, the understanding of the very disparate topics relies on statistics and immense masses of data. This is why DESY theorist Ayan Paul started to help fighting a novel virus causing a spreading pandemic. Experienced in using advanced statistical models capable of dealing with large datasets and multiparametric analyses, he knew that big data can open up insights that elude us otherwise.

Download [237KB, 1920 x 1017]
Population density and Covid-19 cases in Germany. Publicly available data like these are just some of many factors to calculate models of the pandemic spreading and the probabilities for immunity (illustration A. Paul, map: © GeoBasis-DE/BKG 2014).
“As the spreading SARS-CoV-2 virus sent us all home to muse over our lives and work, I was wondering how I could contribute with my scientific knowledge towards social good,” says Ayan. While rummaging through data and looking for avenues to express his skills, the young Indian came upon several opportunities amongst which one seemed particularly intriguing: He took part in a hackathon, the Covid-19 Challenge in early April, organized by the Massachusetts Institute of Technology (MIT). More than 1500 participants from all over the world with varied skills met virtually and formed teams to address different issues related to the pandemic. Paul formed a team with six other participants which went on to be one of the winning teams that weekend. They had proposed the development of an intelligent algorithm to track immunity development and reduce clinical burden.
Download [136KB, 999 x 1046]
Ayan Paul is a post-doc in DESY´s theory group (photo: A. Paul).
This victory marked the formal beginning of Ayan’s research related to Covid-19, and he wished to address several problems that demanded immediate attention. “One of the major problems with the spread of Covid-19 is the prevalence of asymptomatic carriers and pre-symptomatic spreading that was making the traditional methods of manual contact tracing completely useless in containing this pandemic in stark contrast with the previous ones due to Ebola, SARS and MERS”, he explains. The problems in manual tracing made several tech giants and governments propose the use of automated contact tracing through mobile devices using their capabilities of proximity tracking. Apple and Google have recently launched proximity tracing in iOS and Android, and the German government plans to use this in their contact tracing app.

With the knowledge of basic statistical analysis, it seemed to Paul that the advocates of automated contact tracing were being too optimistic. Together with Hyunju Kim from the Beyond Center for Fundamental Sciences at Arizona State University, Ayan realized that the disease characteristics must be well understood and connected with the transmission parameters to properly understand the efficacies of contact tracing. Between Kim and Paul, they had sufficient expertise in complex networks, statistical mechanics and information processing, but not in understanding the efficacy of automated contact tracing or mathematical epidemiology. Two rigorous weeks of studying medical literature based on clinical data, epidemiological modeling and virology cleared to them the path to understand the specific features of Covid-19 and how it spreads within the population.

Their work sends a simple message: Covid-19 has spreading characteristics that are very unique. In theory, automated contact tracing is a very effective means of containing Covid-19 and is possible with current technologies. However, as Kim and Paul’s work shows, this requires a rather large fraction of the population to not only participate in the tracing program but also be able to confirm that they have tested positive for Covid-19. According to their estimates, even in an optimistic scenario, 50% to 75% of the population need to enroll in this contact tracing program. “Considering the fact that less than 70% of the population have smartphones (which are necessary for this to work), this program in practice is impossible to implement,” Ayan sums up their paper that has recently been submitted to a peer reviewed journal. Similar conclusions were reached by a group at Princeton, but using a more involved statistical mechanics analysis.

Ayan Paul continues to work in several other directions to understand the spread of Covid-19 and its effects on highlighting disparities in society. One of his interests is to understand the development of immunity in asymptomatic and mildly symptomatic individuals, which might help to relax the burden on the medical system and bring immune individuals back to work or perform essential functions even where the risk of Covid-19 exposure is high. Kim and Paul want to formulate an algorithmic way of identifying immune individuals through a combination of mobility patterns and exposure to Covid-19 in high-risk locations. This work requires simulation of spreading of the infection in a population and subsequent training a machine-learning algorithm to find the optimal parameters for identifying immune individuals with reasonable confidence. This can then be used to prioritize immunity testing in critical settings like a hospital, or within the population itself.

With a different team from DESY and Seattle, Ayan is studying census data along with Covid-19 prevalence in the USA and Germany to study whether social disparities at the macroscopic level (at the level of counties in the US and the level of districts in Germany) show any correlation with the prevalence of the disease, enabling governments to bring changes to local policies that will focus on the regions in need during such a pandemic. The research is being done in a non-traditional manner using advanced machine learning methods to analyze big data, which is not very common in studies of economics.

Meanwhile, Ayan and his team went on to win the second advanced round of the MIT Covid-19 Challenge while working on a business model focusing on fighting Covid-19. Combining knowledge of contact tracing and immunity development in individuals, he and his teammates have founded “CoVis”. They will build a platform for assessing risk and immunity scores on mobile devices and are extending the platform for use by healthcare providers, corporate and government institutions to help them build business roadmaps and policies around the ongoing pandemic and help make better decisions in the future. Their platform is based on putting together knowledge from medical literature, large datasets on local disease prevalence, user information and participation using intelligent algorithms to give a probabilistic understanding of how much risk an individual faces of getting infected and determining how likely they are to have become immune from an unknown or unconfirmed Covid-19 infection.

All these works might seem very different from the career in physics that Ayan Paul has followed. However, he finds a conceptual unification amongst his current work on Covid-19 and particle physics: “Exploring the mysteries of the fundamental dynamics and symmetries of Nature continues to be a long and fascinating journey but, for once, life has put in our path a possibility of seeking knowledge that can directly benefit humankind. I wish to not let this opportunity pass by unexplored.”