Antimicrobial Resistance

Stalling The More Permanent Future Pandemic We Still Can Prevent

Mukundh Murthy
10 min readFeb 16, 2021

You’re a tourist visiting India for the first time. You’ve had a great time so far, chatting with the locals and building a great relationship, trying new food, and visiting landmark tourist destinations that’ve been on your bucket list for quite some time. You get back to your cottage in the village you’re staying in, and suddenly you don’t feel right. Something doesn’t feel right.

As you start feeling sick, you ask for the locals to direct you to the nearest hospital and doctors who might be able to treat you.

It turns out you have an antibacterial infection. But you know how those are cured – through antibiotic prescriptions – so you’re not too worried. At the end of the day, maybe you can head back to your village and continue with your plan.

The doctors prescribe you an antibiotic. But it doesn’t work. You come back to the hospital, hoping it was a simple prescription snafú or medical error. Hopefully, that’s all it is. More antibiotics, and the same results. But with each antibiotic, your condition worsens. You find it harder to breathe, and you wonder if you’ll ever be able to come out of your condition.

You do have a bacterial infection. That’s true. But not a normal bacterial strain. A superbug – with no discovered cure.

By 2050, the WHO estimates that antimicrobial resistance will turn out to be one of the leading causes of death. As superbugs infect more and more humans, we’ll find ourselves more and more powerless to solve the problem.

Brief Tangent – how do bacteria even harm human cells in the first place?

Fortunately, we have a number of antibiotics that target various mechanisms of bacterial metabolism and life processes. Ampicillin interferes with the construction of the bacterial cell well, rifampin interferes with RNA polymerase, and streptomycin works by causing the misreading of tRNA by binding to ribosomal subunits.

Finding an antibiotic that targets a specific biological mechanism within the bacteria isn’t the issue. The problem, however, is that bacteria, over the course of multiple generations, can evolve mutations that lead to the same antibiotics having little to no effect on the bacteria.

Here’s a breakdown of what we’re going to cover in this article.

Part 1 — how do bacteria become resistant to antibiotics?

Part 2 — existing antimicrobial resistance solutions and their limitations

Part 3 — future directions for the field of antimicrobial resistance

How do bacteria become resistant in the first place?

The probability of a mutation in a specific position in a bacterial genome while replication occurs is 10^-10 per genome per replication. That seems minuscule – something that should hardly be a problem.

However, multiply this small number by the number of bacteria in a given generation, and the number of generations in a certain time period, we can see that within __ time period, there’s a high probability that there are is at least one bacteria in a large population with a mutation at a specified position.

When we design antimicrobials (either in the forms of peptides, small molecules, or biologics), the bacteria can mutate one of the multiple possible proteins in a given pathway. That increases the probability even further.

Let’s take the example of rifampin. Rifampin is a small molecule that targets two proteins in E. coli. It primarily targets RNA polymerase, one of the proteins involved in DNA replication in b bacteria. Secondly, it also targets mazE, a member of the mazEF toxin-antitoxin system in bacteria. A toxin-antitoxin system is in essence a measuring device that a cell uses to measure stress on the cell (lack of nutrients, mutagen, dangerous chemical elements, etc.) One promoter unit is responsible for the transcription of both genes (so when one is transcribed, the other is transcribed as well). Normally antitoxin is present to neutralize the toxin, but by binding to both RNA polymerase to prevent it from transcribing either the toxin or the anti-toxin and by binding to mazE, rifampin reduces the available amount of antitoxin (mazE) that is available to neutralize the already translated toxin (mazF). The bacteria could mutate either mazE or RNA polymerase to evade resistance. That takes the number that we previously found and multiplies it to make it an even larger factor if we’re trying to assess the probability that within a given number of generations, a bacteria will evolve resistance to a particular antibiotic.

Schematic of the mazEF toxin-antitoxin (TA) system

The numbers I’m using here are not entirely precise, but they’re accurate enough to convey that bacteria can almost always evolve resistance against a particular antibiotic (there seem to be exceptions though, which might lead to possible breakthrough solutions. We’ll talk more about this in a later section).

But there’s more… Bacteria–unlike humans–need not procreate or generate new offspring in order to spread mutations throughout the gene pool.

The Transposon

Transposons are self-automated pieces of DNA. I think of them as biological robots. They are pieces of DNA that excise (cut themselves) out of a particular bacterial sequence, form a short, circular piece of DNA called a plasmid, and then reinsert themselves into the genome of another bacteria.

When a bacteria acquires a new gene or a new allele within its own generation, that is called horizontal transfer (there are three main methods of horizontal transfer, including transmission via a virus, transmission from one living bacteria to another living one, and transmission between a dead bacteria and a living one – but I won’t go in-depth for each of these here).

Moreover, recent research shows that these transposons (unlike, say cas9 complexes in gene editing, which require a particular PAM sequence to facilitate sequence recognition, require only two nucleotide sequences in order to insert themselves into a new sequence of DNA.

Here I’ll give three case studies, showing some (but not even close to all) of the possible mechanisms bacteria can evolve against antibiotics.

The Notorious Efflux pump

An efflux pump is like the whole that you might find at the bottom of a plastic water bottle you’re trying to fill up.

Bacteria have evolved pumps that can take a good number of antibiotics and pump them outside of the cell. Right now, this seems to be the largest problem when it comes to working towards solving the problem of antibiotic resistance.

The topic of efflux pumps is a vast one and one that we cannot fully cover in this article. If you’re particularly interested in this mechanism of resistance, I’d encourage you to look at some papers here, here, and here.

Mutating the target

The last mechanism that I’d like to discuss is the most direct mechanism for resistance. Given a particular protein that is targeted by a small molecule, bacteria mutate that target, which results in a small molecule that has little to no affinity for the new mutant.

The nature of small-molecule — protein binding is such that even a small mutation to the active site will result in a huge resulting change in binding affinity. Most proteins have a catalytic triad, where only three amino acids are responsible for the majority of the enzyme activity. That’s why it is not surprising that mutating even one residue can result in a complete lack of affinity for the originally potent molecule. These single residue mutations that confer almost complete resistance to new antibiotics are so common that they’re given a special name – “escape” mutations.

Part 2 — Existing antimicrobial resistance solutions and limitations

The most obvious solution to the problem of antibiotic resistance is to keep developing them. That’s the approach the scientific community is, for the most part, taking right now. Once a target becomes irrelevant due to the transfer of an antibiotic resistance determining factor, you choose a new target and start the process all over again. Each cycle of developing a new antibiotic and then bacteria getting resistant to the antibiotic takes only a few years (In the lab, bacteria often take no more than a few hours to develop resistance to a drug, but obviously the selective pressure in the lab is much higher than that in nature. More on that here.)

There are solutions that attempt to discover new antibiotics faster and acquire a diverse array of antibiotics, but these approaches do not actually expand the cycle (i.e. they do not increase the amount of time that it would take for bacteria to get resistant to a given drug).

One example of recent work is Deep Learning for Antibiotic Discovery, work coming from a collaboration between the BROAD institute and CSAIL at MIT. The work utilizes a message parsing neural network architecture to train a classifier to predict whether a given molecule is antimicrobial or not. After training on a large set of molecules, the model was deployed on a set of already FDA approved repurposable drugs. After deploying the model on this set of repurposable drugs, they discovered halicin, a molecule that, when compared to other structurally, showed an ability to evade resistance that was much higher. They tested halicin in mice and murine mouse models to discover that it showed high levels of potency.

Again, the main limitations here are that the method extends new ways to discover antibiotics but doesn’t inherently elongate the development, resistance cycle.

Currently, many groups are also focusing on the development of antimicrobial peptides. Peptides are another modality of drugs (other than small molecules) that are made of small strings of amino acids. Peptides have been shown (we still don’t understand why) to have intrinsic antimicrobial activity and that’s why many groups seem to be capitalizing on them — for example, the de la Fuente lab at UPenn.

Part 3 — future directions for the field of antimicrobial resistance

A few days, in a call with an antimicrobial resistance author and research scholar from Princeton (Laura Kahn), I learned about the idea of one health.

The one health hypothesis explains that most organisms share many similar aspects of physiology, especially when it comes to pathology. More specifically, its the idea that solving antimicrobial resistance for humans might mean solving resistance in different species – for example, the pets that we live with or the flies that are capable of passing multidrug-resistant bacteria.

The idea of “one health” more generally leads to the understanding that antimicrobial resistance is a multifactorial problem. We can’t solve this problem (solve isn’t the right word here, since evolution will always catch up, maybe lighten) solely through innovating new mechanisms for molecular therapeutics. Rather, the only way to solve antimicrobial resistance is to, in addition to developing new medicines, propose new legislation and policy limiting the amount of off the counter antibiotics. We need to ensure that doctors do not over-prescribe antibiotics for viral infections or ear infections where they aren’t necessary. We need to work on global hygiene, especially placing a focus on open defecation, to lower the spread of multi-drug resistant bacteria.

Think about what the world has gone through during this one year of COVID-19. Schools shut down, social distancing, and all the mental health, isolation, and anxiety issues that have resulted from it. Now imagine another COVID-19 year in 2050. And then in 2051. And again in 2052.

Put quite simply, the world that we live in right now will become permanent if we don’t do something about antibiotic resistance. Right now – as grim as it sounds – resistance is on the trajectory to erase centuries worth of human discovery in medicinal chemistry and structure-based drug discovery.

The good news is –unlike with COVID-19 – we still have time to prevent types of catastrophes that could result from negligence.

Ultimately, we must ask ourselves this question — do we absolutely have to fight head-on with evolution when we design antimicrobials, or are there ways to harness the bacteria’s own circuitry?

Works Consulted

Munita, J. M., & Arias, C. A. (2016). Mechanisms of Antibiotic Resistance. Microbiology spectrum, 4(2), 10.1128/microbiolspec.VMBF-0016–2015.

M. A. Webber, L. J. V. Piddock, The importance of efflux pumps in bacterial antibiotic resistance, Journal of Antimicrobial Chemotherapy, Volume 51, Issue 1, January 2003, Pages 9–11,

Kalghatgi, S., Spina, C. S., Costello, J. C., Liesa, M., Morones-Ramirez, J. R., Slomovic, S., Molina, A., Shirihai, O. S., & Collins, J. J. (2013). Bactericidal antibiotics induce mitochondrial dysfunction and oxidative damage in Mammalian cells. Science translational medicine, 5(192), 192ra85.

Erik Gilberg, Swarit Jasial, Dagmar Stumpfe, Dilyana Dimova, and Jürgen Bajorath Journal of Medicinal Chemistry 2016 59 (22), 10285–10290, DOI: 10.1021/acs.jmedchem.6b01314

Shen Y, Radhakrishnan ML, Tidor B. Molecular mechanisms and design principles for promiscuous inhibitors to avoid drug resistance: lessons learned from HIV-1 protease inhibition. Proteins. 2015;83:351–372. doi: 10.1002/prot.24730.

Pines, G.; Fankhauser, R.G.; Eckert, C.A. Predicting drug resistance using deep mutational scanning. Molecules 2020, 25, 2265.

Stokes, J. M., Yang, K., Swanson, K., Jin, W., Cubillos-Ruiz, A., Donghia, N. M., MacNair, C. R., French, S., Carfrae, L. A., Bloom-Ackermann, Z., Tran, V. M., Chiappino-Pepe, A., Badran, A. H., Andrews, I. W., Chory, E. J., Church, G. M., Brown, E. D., Jaakkola, T. S., Barzilay, R., & Collins, J. J. (2020). A Deep Learning Approach to Antibiotic Discovery. Cell, 180(4), 688–702.e13.

Hey! I’m Mukundh Murthy, a 17 year old passionate about the intersection between machine learning and drug discovery. Thanks for reading this article! I hope you found it helpful :)

Feel free to check out my other articles on Medium and connect with me on LinkedIn!

If you’d like to discuss any of the topics above, I’d love to get in touch with you! (Send me an email at or message me on LinkedIn) Also, feel free to check out my website at

Please sign up for my monthly newsletter here if you’re interested in following my progress :)



Mukundh Murthy

Innovator passionate about the intersection between structural biology, machine learning, and chemiinformatics. Currently @ 99andbeyond.