-Pratyush Mishra
With the development of artificial intelligence and biological data science techniques, it is now easy to carry out certain research which seemed to be cumbersome a few years ago. Artificial intelligence aims to develop the learning and reasoning ability of a computer program in order to mimic human intelligence. AI-based technologies are now widely used in the field of healthcare and other biosciences. One of the recent approaches of such AI-based tools is looking for potential antibiotics against bacteria and other pathogenic microbes. Antibiotics are assumed to be functional until and unless the targeted microbe develops resistance against it. Apart from that, there are many problematic bacteria strains that are resistant to all known antibiotics. The availability of chemicals to be used as potent antibiotics are not limited, however, their screening and sequential testing for determining the antibiotic/antibacterial actions is very difficult and time-consuming.
Here comes the use of AI-based algorithms in several in-silico researches to screen about a thousand of chemical compounds and then pick out the most potent antibiotic agent. New neural networks are developed which can map several molecules to their vector representations which are constructed with respect to the presence/absence of certain chemical groups. The use of deep learning algorithms has transformed the conventional in-silico approaches, making them more accurate and fast.
Researchers from MIT developed a similar model which was progressively trained using about 2500 molecules and several drugs to find out an antibiotic against E-coli. Once the model was trained to know the chemical features that make a molecule effective against E-coli, it was allowed to screen the number of compounds from drug databases. The model successfully figured out a molecule which was effective against the bacterium and even showed the least toxicity on human cells. The drug is known as Halicin, which kills bacteria by targeting their ability to maintain electrochemical gradient across the cell membranes. Interestingly this type of mode of drug action is always effective since it is difficult for a bacterium to develop resistance against it.
Sometimes it becomes difficult to target a specific microbe using general antibiotics. It’s very important to preserve those microbes that are really important and play a positive role in various microbiomes in our body. Using AI-based algorithms it is now even possible to produce species-specific drugs and antibiotics that can only target a single bacteria or microbe out of several millions. Those models are programmed to detect certain species-specific features, like a protein which is only produced on the cell membrane of the targeted microbe. By virtual screening of several available chemical compounds in online databases like the ZINC database which contains a suitable 3D representation of biologically active molecules, such optimization can be done easily by AI-based technology.
References:
- https://www.nature.com/articles/d41586-018-02174-z
- https://news.mit.edu/2020/artificial-intelligence-identifies-new-antibiotic-0220
- Durrant, J. D., & Amaro, R. E. (2015). Machine-learning techniques applied to antibacterial drug discovery. Chemical biology & drug design, 85(1), 14–21. https://doi.org/10.1111/cbdd.12423
- https://en.m.wikipedia.org/wiki/Halicin#:~:text=Halicin%20(SU%2D3327)%20is,to%20poor%20results%20in%20testing.
- https://en.wikipedia.org/wiki/ZINC_database
Badia Article heichi
ReplyDelete