Will AI Crush the Bacterial Rebellion?

Antibiotics are drugs used to combat harmful bacteria. However, their magic seems to be fading with increasing bacterial resistance.

Remember going to the doctor with a sore throat? The next thing you know he is prescribing you a ten-day course of antibiotics. Considering the strong person that you are, you are relieved of the soreness in just a week. So, you discontinue your pills because why not?

But soon enough you see your symptoms reappear.

Antibiotics are drugs used to combat harmful bacteria. However, their magic seems to be fading with increasing bacterial resistance.

Antibiotics no longer scare bacteria.

Antibiotic resistance is more serious than you think. By 2050, it is expected to cause around 10 million deaths per year. So we need better antibiotics before the disaster, much faster than traditional drug discovery approaches.

We need an alternative to accelerate the process. Can Artificial Intelligence (AI) be the respite to the massive and rigorous research process?

In this article, we seek to cover:

The growing resistance of bacteria against antibiotics
The role of AI and ML in solving this problem

1) AI Accelerates Antibiotic Discovery

AI does have the potential to fast-track drug discovery. And there are studies that prove this. One of many is the study on E.Coli. The name might sound unfamiliar. But the food you eat and water you drink know it all too well.

MIT researchers James Collins and Jonathan Stokes were one of the first to turn to AI. They trained Machine Learning (ML) algorithms to identify chemical compounds that inhibit E.Coli. That made looking through more than 100 million compounds easy. From this data set, the algorithm discovered a compound called halicin that was also effective against numerous bacterial species, other than E.Coli.

That established the competence of AI in Antibiotic Discovery. But this barely solved the problem of antibiotic resistance.

2) Addressing the Real Problem

Antibiotic resistance does not arise from every other antibiotic. The sinners in this case are the broad-spectrum antibiotics.

As the name suggests, they target more than one type of bacterial species. And halicin is one such antibiotic.

Let us go back to your case. You need antibiotics that can fight the bacteria causing soreness in your throat. Instead, you take those that fight other bacteria as well (and also kill some good bacteria). With time, all the survivors evolve and become resistant to subsequent antibiotic exposure.

So, we don’t need just any type of antibiotic. We need specific antibiotics that target only the bacteria we want dead.

3) AI Discovers Specific Antibiotics Faster

Collins and Stokes proved this for us through a study published in Nature Chemical Biology. After tackling E.Coli to defeat, they took on Acinetobacter baumannii.

A.baumannii is known to take over hospitals and infect every other patient, staff and visitor. These infections lead to serious health complications including pneumonia and meningitis. That isn’t the worst. It offers resistance against multiple antibiotics. Hence, getting it under control wasn’t a cakewalk.

To mend this menace, Collins and Stokes focused on three main areas:

Training ML algorithms
Applying ML predictive models to drug discovery
Analyzing the mode of action of the discovered drug

One Click. Real Value.

Have the best of the world of technology and corporate at your fingertips.

3.1) AI Discovers Specific Antibiotics Faster

ML systems learn from data sets. They use this pre-fed data to make sense of other data not known to it. That is why they are often called ‘predictive models’.

Collins and Stokes needed this ready-made data to feed into the ML system. They exposed A.baumannii to about 7,500 different chemical compounds in the laboratory. And then entered the structure of these compounds into the system. They also told the model which structures could inhibit bacterial growth. That allowed the algorithm to learn chemical features associated with growth inhibition.

3.2) Applying ML Predictive Models to Drug Discovery

The hard part was over. Now it was all ML’s doing.

The researchers used this trained model to analyze another 6,680 compounds it had not seen before. The model took less than two hours to arrive at a few hundred close matches. Collins and Stokes then tested these in the lab for molecules different from existing antibiotics.

The tests yielded a highly specific drug that effectively killed A.baumannii but did not affect other bacterial species. They called this magic potion, abaucin.

3.3) Analyzing the Mode of Action of the Discovered Drug

Collins and Stokes discovered that abaucin acted by inhibiting an enzyme that interfered with a specific process. The process is called lipoprotein trafficking and the enzyme is LolE.

Much like A.baumannii, other gram-negative bacteria also express LolE. However, abaucin did them no harm. It was that specific of its target.

4) AI Can Solve the Larger Problem

Narrow-spectrum drugs are the alternative to their counterpart (broad-spectrum antibiotics) and the answer to the problem of antibiotic resistance. These are the drugs that shoot nothing else but their target, much like abaucin.

To find compounds with a highly specific mode of action is no easy task. That is where AI and its ML predictive models come into play. They make scanning large molecular data sets easy and in turn, accelerate drug discovery.

Combining AI predictive power with ML training can enable us to create effective solutions – whether it is for new drug discovery or for effective care delivery.

With the groundwork laid for AI-derived narrow-spectrum antibiotics, researchers are now looking to counter other lethal bacterial species as well. They are eyeing similar drugs to combat infections caused by Staphylococcus aureus and Pseudomonas aeruginosa.

These lead to serious health complications including bloodstream infections and septicemia. So clearly they need AI’s attention next.

Thanks to Collins and Stokes, we realized the potential of AI in fighting bacterial infections sooner than later.

Advanced Data ETL, Reporting and Gen AI

No-code data engineering

Automated data transformation

Enterprise-grade LLM

MODULES

Healthcare Data Management

Advanced Data ETL, Reporting and Gen AI

No-code data engineering

Automated data transformation

Enterprise-grade LLM

Solutions that ensure

Advanced Data ETL, Reporting and Gen AI

No-code data engineering

Automated data transformation

Enterprise-grade LLM

Explore more

Advanced Data ETL, Reporting and Gen AI

No-code data engineering

Automated data transformation

Enterprise-grade LLM