How Is Machine Learning Being Used To Fight Antibiotic Resistance

Antibiotic Resistance

Bacterial infections contracted from the hospital environment are highly dangerous. This is because they are primarily resistant to antibiotics and are usually lethal. The World Health Organization has predicted that antibiotic resistance, which is currently very common, can be the cause of 10 million deaths every year till 2050.

To combat and overcome the deadly aspect of antibiotic resistance, scientists are now employing Machine Learning in healthcare.

Artificial Intelligence advantages in healthcare have indeed been tremendously leveraged, thanks to the availability of databases containing genomes from various strains of pathogenic bacteria including the ones that are antibiotic-resistant too.

So, how has artificial intelligence and machine learning fared in battling antibiotic resistance? Let us find out what two of the most recent projects have shown.

Drugs and their dosage

Scientists have so far relied on identified antibacterial resistance genes, to match bacterial strain with drugs that can kill them. But AI can bring much better results in this. AI scientists and computational biologists have developed a Machine Learning approach to differentiate the resistant strains from the susceptible ones, to understand the drug profile of new bacterial strains.

This Machine Learning algorithm is not biased by a list of known resistance genes and protein-coding genes. It can thus be used to identify unknown genetic factors that play a part in developing bacterial resistance across the entire genome.

Gene prospecting

Researchers who have been studying bacterial resistance have found gene elements that interact with a specific drug directly. But other factors can also influence the susceptibility to antimicrobials like those genes that impact the permeability of the bacterial cell wall.

For instance, Mycobacterium tuberculosis infects about 10 million people every year all over the world and about 500,000 of these infections are known to be resistant to over-the-counter antibiotics. That is why bioengineering researchers have begun to look for new resistance genes in the genome of this bacterium. They have employed AI to create algorithms like Support Vector Machine (SVM) and L1-norm to distinguish between susceptible and resistant strains.

Usually, AI models that are used for this research, employ standard reference genome, which may or may not include the most common strains. But keeping in mind the possibly opposite outcomes than are exceptions to the standard, the team has used a "pangenome" method to cover all possibilities and eliminate bias.

Apart from these couple of instances, there are many more ways in which Machine learning in neuroscience and medicine is being used to combat this global problem of antibiotic resistance.