Machine learning to battle COVID-19 bacterial co-infection – UQ News

University of Queensland researchers have used machine learning to assist predict the chance of secondary bacterial infections in hospitalised COVID-19 sufferers.

The machine learning approach can assist detect whether or not antibiotic use is essential for sufferers with these infections.

Associate Professor Kirsty Short from the School of Chemistry and Molecular Biosciences mentioned secondary bacterial infections will be extraordinarily harmful for these hospitalised with COVID-19.

“Estimates of the incidents of secondary bacterial infections in COVID-19 sufferers are broad however in some research 100% of deadly circumstances have suffered a bacterial co-infection,” Dr Short mentioned.

“To cut back the chance of bacterial co-infections, it might be theoretically potential to simply deal with all COVID-19 sufferers with antibiotics.

“However, there’s a hazard that over-treating with antibiotics might doubtlessly lead to antibiotic resistance and the creation of bacterial superbugs.

“We’ve helped develop a strong predictive mannequin to decide the chance of bacterial infections in COVID-19 sufferers, facilitating a cautious use of antibiotics.”

The approach is called the ‘least absolute shrinkage and choice operator’ – or LASSO for brief.

Blood samples of COVID-19 sufferers from six nations had been analysed utilizing the LASSO approach.

The group discovered that the expression of seven genes in a COVID-19 affected person can predict their danger of creating a secondary respiratory bacterial an infection after 24 hours of hospital admission.

Dr Meagan Carney from the School of Mathematics and Physics mentioned these seven genes will now information clinicians to making a extra knowledgeable selection when it comes to antibiotic use.

“This information raises the thrilling risk that gene transcription and evaluation on the time of medical presentation at a hospital, along with machine learning, can change the sport for antibiotic prescription,” Dr Carney mentioned.

She additionally emphasised that LASSO is simplified in contrast to the complicated machine learning strategies surrounding synthetic intelligence which can be currenting being mentioned within the media.

“Research initiatives like this one, which use a lot much less complicated machine learning strategies, can assist bridge the hole between information scientists and scientists in different fields,” she mentioned.

“We ought to try in direction of making information science much less of a black field and encourage scientists the world over to higher perceive the way it can revolutionise the medical business.” 

The researchers would love to acknowledge the in depth worldwide collaboration of clinicians, virologists, bioinformaticians and plenty of different specialists that made this examine potential, together with the PREDICT Consortium, the Snow Foundation and the Nepean Hospital.

The analysis is printed in Lancet Microbe.

Media: Associate Professor Kirsty Short, [email protected], +61 452 374 811; Dr Meagan Carney, [email protected], +61 451 801 218; Faculty of Science Media, [email protected], +61 438 162 687.

https://www.uq.edu.au/news/article/2024/01/machine-learning-battle-covid-19-bacterial-co-infection

Recommended For You