Africa Defense Forum
ADF is a professional military magazine published quarterly by U.S. Africa Command to provide an international forum for African security professionals. ADF covers topics such as counter terrorism strategies, security and defense operations, transnational crime, and all other issues affecting peace, stability, and good governance on the African continent.

Tunisian AI Software Provides Early Warning of Harmful COVID-19 Variants

ADF STAFF

A Tunisian tech company’s artificial intelligence (AI) software is part of a new venture to identify potentially deadly COVID-19 variants before they spread.

InstaDeep was founded in Tunis in 2014 with just $2,000 to develop the next generation of AI software for electronics manufacturing, logistics and biotechnology. In early 2021, InstaDeep joined forces with Germany’s BioNTech to create an AI program to predict dangerous new variants. They describe the program as an early warning system.

“For the first time, high-risk variants could be detected on the spot, potentially saving months of precious time,” InstaDeep co-founder and CEO Karim Beguir said in a statement. “We are happy to make our research work publicly available and, most importantly, look forward to its continued real-world impact.”

After InstaDeep’s announcement in early January, the company raised $100 million from investors, including BioNTech, Google and German rail operator Deutsche Bahn. InstaDeep has researchers and engineers in Nigeria, South Africa and Tunisia along with Great Britain, France and the United Arab Emirates.

The AI software reviews COVID-19 samples uploaded to a global database that catalogs the genetic sequences of the virus as it moves around the world.

The software looks for important changes in spike proteins — the “keys” the virus uses to open cells in which it replicates. Spike proteins also are a vital part of the way the immune system identifies and neutralizes the virus.

Spike protein mutations can help the virus slip past the immune system’s gatekeepers.

The flood of COVID-19 samples coming into the global database is too much for human experts to analyze quickly, according to Beguir.

“More than 10,000 novel variant sequences are currently discovered every week, and human experts simply cannot cope with complex data at this scale,” Beguir said.

Variants evolve naturally as the virus replicates inside cells. Every infection is a new opportunity for the virus to create a variant that can escape the immune system. Most variants are harmless, but others, such as the delta variant, go on to become worldwide killers.

Recently, at least two variants discovered in South Africa — C.1.2 and omicron — showed the highest number of mutations ever recorded. Scientists say those mutations likely came about when an immune-compromised person struggled to fight off an infection.

As teams developed the software between September 2020 and November 2021, the InstaDeep AI flagged all but one of the world’s 13 major variants as potentially harmful weeks or months before the World Health Organization (WHO) raised the alarm about them. The AI system identified the alpha variant as a concern after reviewing 25 samples. The WHO made the same assessment after studying nearly 1,600.

The only variant to escape the InstaDeep’s early warning system was delta, which first was identified in India. India has limited capacity for sequencing virus genomes and restrictions on exporting biological data, so the global database lacked enough data on delta for the AI to do its work, Beguir noted in a paper published on the bioRxiv server.

According to BioNTech, given enough data, the software can pinpoint a potentially threatening variant within a day of examining a sample. Omicron was flagged the same day it was studied.

“Early flagging of potential high-risk variants could be an effective tool to alert researchers, vaccine developers, health authorities and policymakers, thereby providing more time to respond to new variants of concern,” said BioNTech co-founder and CEO Dr. Uğur Şahin.

You might also like

Comments are closed.