Enough about the pandemic.

We are plenty of regular problems- like antibiotic-resistant microbes.

Believe it or not, there are still some scientists left NOT working on vaccines or therapies for SARS-CoV-2 who are able to address this recalcitrant problem.

The conventional method is to add compounds to growth media and see if the microbial population becomes attenuated.  That requires lots of lab space, research time, and, of course, dollars.

Enter Dr. James Collins.  (Medical Engineering & Science and Biological Engineering at MIT).   He developed a computer model (a machine-learning algorithm, also known as AI or artificial intelligence) that can screen millions of chemicals in very short order (using what is termed ‘in silico’ testing [under glass]), identifying those most likely to deliver the knockout punch to the recalcitrant microbes. (His co-principal investigator is Regina Barzilay, EE and Computer Science, MIT. Their post-doc, Jonathan Stokes, got credit as the lead author for the publication in Cell, with 17 more co-authors.)  And, this new process is head and shoulders above previous screening systems.

Collins’ group used the algorithm to identify several promising compounds.  Now, they have to do the testing on these newly isolated materials.

The initial work sought as a target invasions from Escherichia coli.  And, the AI examined some 2500 molecules, including 1700 FDA approved treatments and 800 natural compounds (with vastly different bioactivity). Once that hurdle was completed, they examined  an additional roster, a proprietary list of compounds (Broad Institute [of MIT] Drug Repurposing Hub, comprised of some 6000 compounds).

The algorithm highlighted one molecule with a similar structure to existing antibiotics.  It was predicted to manifest strong antibacterial activity.  The compound acts by disrupting bacterial capability to maintain an electrochemical gradient across the bacterial membrane.

(These gradients  involve ionic transport across the membrane; it is multicomponent.  The first part is the chemical gradient- the difference in the solute concentration on both sides of the membrane.  The other part is the electrical gradient, the resultant differential of charge due to the different ionic concentrations across the membrane.)

Of course, this group would a fan of the movie 2001: A Space Odyssey.  So, they named the target compound halicin, after HAL, the bot of the movie.  Thankfully, despite the cuteness of its name, it was effective in two different mouse models.

The MIT-based group didn’t stop with identifying halicin.  Using the ZINC15 database, they examined some 100 million additional compounds- over the course of 3 days.  (This would take months, if not years, without the developed algorithm.)  Out of the 100 million, they identified 23 other potential candidates.  (Collins’ group used a different model to discern that the desired compound would not manifest toxicity to our normal cells.)

And, now, the real testing will begin.  On a much smaller roster of candidates.

Share this: