Introduction
Researchers at Washington State University have been busy with the creation of a machine learning based graphical UI software PARGT to identify drug-resistant antimicrobial genes in bacteria.
According to the researchers, they have achieved two things- i) they have successfully used protein characteristics as features to train the machine learning model to accurately identify antimicrobial (AMR) genes in Gram-Negative bacteria ii) they have created a software for this, named PARGT with graphical UI; a much welcome thing for users who are not technical to write ML code in R or Python.
What makes PARGT relevant in today’s times is its capability to make it easier to identify deadly antimicrobial-resistant bacteria that exist in the environment. These super-bugs are known to cause around 2.8 million critical and complicated cases of pneumonia or blood-related disorders. The U.S. alone accounts for 35,000 deaths associated with these bacteria.
Why AMR Bacteria are huge threat?
AMR stands for Antimicrobial Resistance which is a property developed by bacteria when they undergo genetic mutation to overcome the drugs used to treat them. They render the drugs ineffective and a whole new drug is needed to combat them.
We have seen cases of severe diseases such as tuberculosis, malaria, and pneumonia that were caused by strains of bacteria that have become drug-resistant. How are we expected to treat an infection if we are faced with a completely new bacterial organism that does not get affected by the usually working drugs for these diseases? The end result is often a life-threatening situation.
The treatments that we know for killing bacteria are limited. If the new strains are going to crop up every few years, the problem of in-treatable diseases is going to worsen in the decades to come. This problem is not just limited to those being affected. It may result in the skyrocketing of treatment costs and healthcare prices worldwide.
How PARGT Software works?
PARGT is the machine learning software to detect the AMR bacteria effectively, which can have great implications on public health and on the costs of healthcare.
The researchers put their focus on improvising the feature selection process for machine learning by using a Game Theory based feature selection method – “game theoretic dynamic weighting based feature evaluation”. This concept is based on the idea that a single feature might contribute little to the predictive value of the ML model but when used with other features it creates a strong coalition.
The ML model was created using the SVM algorithm and they were able to achieve a classification accuracy of 87%-90% on the bacterias resistant to bacitracin and vancomycin
According to the medical community, these are the most widely used drugs to treat infections such as Staph.
Most importantly they have created standalone software to enable non-technical people to perform their research.
Researchers have made the code available on GitHub under the common creative license. You may check their code here.
To quote the research team scientists, “Our software can be employed to analyze metagenomic data in greater depth than would be achieved by simple sequence matching algorithms,” Chowdhury said. “This can be an important tool for better understanding the emergence of new antimicrobial resistance genes that eventually become clinically important.”
“The virtue of this program is that we can actually detect AMR in newly sequenced genomes,” Broschat said. “It’s a way of identifying AMR genes and their prevalence that might not otherwise have been found. That’s really important.”
Augmentation of machine learning in the medical research may prove to be game-changer, especially during COVID-19 pandemic times.
To view the original research paper click here.
