Could a computer accurately predict the next pandemic?
Is it possible to prevent 6,597,200 COVID-related fatalities so far reported from happening next?
Predicting the next pandemic sounds like Philip Dick sci-fi but scientists continue putting in the work to make the world a safer space for us all.
It’s not easy
HIV/AIDS and the flu influenza torment humanity without a certain cure. Therefore, when one looks at the overall structure of pandemics, it’s imperative to appreciate the dynamics of highly infectious diseases. In a study published by Nature, observations showed “335 emerging infectious disease events occurred between 1940 and 2004. Of those, 60·3% were zoonoses, and most originated in wildlife, including the virus that causes severe acute respiratory syndrome (SARS) and the Ebola virus” and the threat of infectious diseases to humanity keeps rising. Well, the COVID pandemic cleared any skepticism on the severity.
International movements ease the travel and spread of disease-causing pathogens. For instance, there’s a global consensus that COVID originated in Wuhan, China yet it reached every corner of the world. In short, a pandemic in any part of the world threatens us all.
Human and animal interactions are becoming intimate though conflicting due to deforestation, population increase, and surging demand for animal products. While the truth remains elusive, some sections argue COVID originated from bats. HIV is also believed to have developed from baboons. The bigger question is: how did patient zero come across the virus?
When could the next pandemic occur?
Pandemics are random events hence getting an accurate timeframe on incidence is virtually impossible. However, a common pattern for most pandemics is the correlation between proximity to animals and human interactions. The infectious pathogens find their way to humans, and the cycle begins.
Three things happen whenever a pathogen comes infects a human being: causes a wider outbreak such as the Ebola outbreak in Sierra Leone and the Democratic Republic of Congo (DRC), affects a single person such as rabies or explodes to a full-blown pandemic such as HIV and COVID with the possibility of becoming endemic.
Interestingly, risk factors push the incidence to higher levels. In 1998, the El Nino weather caused flooding across the East African region especially in Kenya, forcing livestock to live closely with human beings in the dry lands. A vaccine shortage against the Rift Valley Virus, a common infection among ruminant animals, led to cross-species spread. Thus, from the flooding came the deadly virus.
Climate change and other seemingly unrelated events may become catastrophic to the human race if unaddressed.
Where will the next pandemic start?
Southern China has historically been seen as the origin of most influenza strains. The H1N1 flu of 2009 caused a detour since it was traced to Mexico and Southern US rather than China. It shows the unpredictability of the phenomenon.
From the 1940s to the 2000s, the science community actively tried to develop gene sequencing and expand a database for disease pathogens. Professor Nathan Wolfe, an epidemiologist at the University of California blames the ‘wait and respond’ approach toward HIV/AIDS emergence led to its spread. Proactive policies and surveillance perhaps could bring better outcomes.
Keeping an updated database of disease pathogens cannot comprehensively cover the millions, if not billions, of pathogens out in the wild. However, a pandemic constant is its correlation with populations who directly and closely interact with animals. A pathogen must successfully move from the source to human, then human to human in the locality, and then overcome regional limitations to attain its status by reaching various the globe. That’s what constitutes a pandemic.
Predicting the Next Pandemic
Machine Learning (FluLeap)
When a group of Russian poultry farmers got infected with the H5N8 virus, which was otherwise not known to infect human beings, scientists quickly updated the genomic sequencing for the flu on the genetic data repository GISAID.
At Georgetown University in Washington DC, scientists uploaded the genetic sequencing on FluLeap. It’s a machine learning algorithm trained on a huge number of influenzas, including the H5N8 to differentiate those risky to human beings. The H5N8 posed an outlier since it’s historically known to infect birds, not people.
The puzzle came when FluLeap identified H5N8 as 99.7% compatible with human infection. This became a breakthrough since the compatibility signature led to swift action that saved human lives. Researchers state only “around 1% of the mammalian viruses on the planet have been identified” and most of the pathogens remain unidentified.
Attempts to expand human knowledge in virome led scientists to sample wildlife. Zoonotic risk prediction integrates both machine learning and statistical models to target survey areas at high risk of being pandemic epicenters. The end goal is vaccine development to protect human beings against devastating impacts such as the ones witnessed during the COVID pandemic.
Here are some of the best pandemic prediction models.
The United States Agency for International Development (USAID) funded a $200 million prediction model known as PREDICT. By the time it was wound up in 2020, PREDICT had collected 949 new viruses from humans, livestock, and wild animals. Some of its findings seem prescient.
PREDICT found “there are thousands of undiscovered coronaviruses in bats (widely thought to be the source of the virus SARS-CoV-2), and predicted that Southeast Asia would be home to the greatest number of viruses in the family to which SARS-CoV-2 belongs” and the unregulated wildlife markets would most likely lead to a pandemic outbreak. Statistical modelling allowed scientists to generalise animal groups and regions with the largest number of undiscovered viruses, and bats featured prominently.
Realising the enormity of the task, scientists working for the PREDICT model created the Globe Virome Project (GVP) in 2016 to collect information on the estimated 1.3 million viruses still unknown to humankind. However, the project never took off since governments and NGOs gave it a wide berth. According to critics, the task is simply a non-starter. Gene sequencing shows viruses mutate rapidly. In this regard, a database becomes more of an archive as viruses are known to spill over creating data bias for predictions.
Additionally, a virus genome doesn’t indicate whether it can infect human beings and lead to a pandemic. Validating all these aspects takes time and costs a fortune, something which most stakeholders seem unwilling to commit.
Luckily, machine learning offers some reprieve since it can be modeled to detect high-priority targets. The future of pandemic predictions requires downstream triaging to help consolidate efforts for maximum output.
Predict and Prepare
If predictions want to provide utility to humanity, they must create publicly accessible tools that ignite actionable plans with relevant information. Experimental work must also be integrated into pathogen predictions since this leads to the validation of any assertions. Mistrust, overselling of findings and a lack of collaborative communication characterise current predictions in the health sector. The vice versa needs to be normalised.
Not so a fun Fact
The Influenza pandemic of 1918 claimed more lives than the First World War. Estimates show that the influenza pandemic caused the deaths of 50 million people compared to the latter’s 16 million deaths.
Ultimately, judgment will be attached to the ability to spur improved virus surveillance, targeted vaccine development, and build resilient healthcare systems to cope with any future pandemics.