MRSA is Increasingly Common in HealthCare Settings – Has AI Just Given a Tool to Fight Back?

The First Medical Revolution

Until the invention of penicillin, the first antibiotic, a simple lung infection or even a small cut, could turn out to be deadly. This was simply a fact of life people lived (and died) with.

For a long time, new antibiotic classes have been regularly discovered, ensuring that the threat of bacterial infection is kept at bay. Recently, the emergence of antibiotic-resistant bacteria has put into question if this can stay true for much longer.

Antimicrobial resistance (AMR) has caused 1.27 million deaths directly in 2019, with 1 in 5 of the victims a child under 5 years old. And 4.95 million people who died in 2019 suffered from drug-resistant infections, such as lower respiratory, bloodstream, and intra-abdominal infections.

This is especially problematic in hospitals, where the higher usage of antibiotics and weaker immune systems make antibiotic resistance a major cause of infections and deaths.

In the arms race between new antibiotics and antibiotic resistance, the bacteria are winning. The latest discovery of a new antibiotic class that has reached the market was back in 1987. Overall, the old-school approach of blindly screening tens of thousands of chemicals for antibiotic potential is not working anymore.

Source: ReAct Group

MRSA – The Looming Healthcare Disaster

Staphylococcus aureus is a bacteria living on the skin. Methicillin-resistant Staphylococcus aureus (MRSA) are bacteria resistant to common antibiotics like methicillin, which is often combined with resistance to other treatments/antibiotics.

Most MRSA infections happen in hospitals or healthcare settings, for example, dialysis centers or nursing homes. It can also spread from skin-to-skin contact.

Because of its presence on the skin, Staphylococcus aureus is a likely candidate to cause infections after surgery or open wounds.

MRSA infections kill more than 10,000 people yearly, just in the USA, with more than 80,000 infections. MRSA strains have also started to become resistant to vancomycin, one of the last effective antibiotics available against it.

This raises the spectrum of widespread unkillable MRSA and the return to the pre-antibiotics era of common deadly bacterial infection.

AI To The Rescue

If screening blindly thousands of candidates for new antibiotics does not work anymore, what about screening millions of candidates intelligently? This is the task MIT researchers led by researcher James Collins are working on.

The MIT researchers use a deep learning AI model to predict what chemical molecules could have antibacterial activity.

First, they taught the AI model what chemical patterns have been known to have antibacterial activity. And then, the AI can check on millions of potential molecules for similar potential.

For this, the researcher first tested the antibiotic activity of 39,000 molecules on MRSA and showed the result to the AI, which used it to learn what works.

They then developed a method called ‘Monte Carlo tree search' to make the AI results more understandable for the researchers by showing what subsection of the potential new antibiotic molecule is expected to be responsible for the medical result.

AI Teamwork

It is not sufficient to find what can kill the harmful bacteria. It is also important to check that the chemical compounds are safe for the patients.

Following the same principle as for antibiotic discovery, the researchers also trained 3 different deep-learning models to predict toxicity to 3 types of human cells.

The results of the findings would then be combined with the ones from the MRSA-focused AI in order to find compounds that are good candidates for both antibiotic activity and safety profile.

This illustrates well how AI-powered drug discovery relies not only on humans guiding the AI but increasingly on AI “talking” to each other and collaborating.

Impressive Results

MIT's researchers analyzed no less than 12 million chemical compounds. Importantly, all of these 12 million compounds were commercially available, making them easier to commercialize as new antibiotics, as the supply chain to manufacture them is already in place.

With the AI's help, a short list of candidates, 280 compounds, was established. They were tested against MRSA, and 2 compounds of the same class were shown to reduce MRSA by a factor of 10. The antibiotics work by disrupting the electrochemical gradient of the bacteria’s cell membrane.

This is a new success for the research team, which had previously found 2 new potential antibiotics (but not useful against MRSA), halicin and aubicin.

What It Means for Future Drug Discoveries

From Pharma To Biotech Back To Pharma?

Drug discovery has always been at the core of pharmaceutical companies' models. This is a vital part of the business, as patents expire and new exclusive drugs are needed to keep money flowing and to fund further R&D and medical progress.

Unfortunately, the approach of screening the natural world and artificial molecules for new active compounds has somewhat stalled.

This has been the driving force for the rise of biotech companies, which use biological molecules like antibodies instead of chemicals. And why “old school” pharmaceutical companies have been forced to acquire biotech startups, turning many into more “marketing machines” rather than the R&D-driven businesses they used to be.

If chemical drug discovery becomes more successful, we might see more pharmaceutical companies return to spending more on R&D and less on biotech acquisitions.

This could have far-reaching consequences for the industry, including:

• A return to growth for innovative pharmaceutical companies.
• The potential for lower bidding for biotech startups.
• The potential for discovering new “blockbuster” chemical drugs for poorly treated diseases.

The Merging Of Tech & Pharma

Another impact of successful AI-driven drug discovery is that the process is becoming less related to medicine and chemistry and more about data and neural networks.

This opens the field for deeper relations between tech companies and pharmaceutical companies. Potentially, drug discovery AI could become as transformative for the industry as the emergence of biotechnology in the 1980s.

We could also see the emergence of new joint ventures between pharmaceutical and AI giants. For example, a drug-discovery company co-owned by giants like Pfizer or Novartis on the pharma side and Google or Microsoft on the AI side.

If AI-drug discovery turns out to become one of the major profit centers of AI, we could also see tech companies acquiring pharmaceutical companies.

The Power Of Data

AIs are built thanks to massive treasure troves of data. The MIT research was no exception and relied mostly on the Mcule database. The field of digital data providers for drug discovery is still in its early stage and is dominated by non-profits and privately owned startups.

In the long run, we will likely see a consolidation of the industry and maybe the emergence of a few key providers that will also be IPOed.

Predicting Toxicity “In-silico”

One last effect of the MIT discovery is to rethink how to forecast chemical toxicity and clinical trial successes.

Until recently, human experience & intuition and animal & in vitro models were the go-to methods for assessing the risks associated with a new molecule.

AI deep learning could help not only find the new antibiotics class but also predict which candidates are likely to be non-toxic. This opens the possibility of using AI to reduce the number of drugs failing phase I of clinical trials, which is designed to test for toxicity in humans.

Around 35% of candidates' chemicals fail in phase I of clinical trials. Reducing this number would speed up drug development and reduce the average R&D costs.

Source: Duxin Sun et al

Top 3 Companies Benefiting From AI Drug Discovery

This MIT discovery was funded by foundations, non-profit, and public funding. However, investors can get exposure to AI-driven drug discovery through several publicly traded stocks.

In general, companies in digital and AI systems for drug discovery can be organized on two scales:

1. Computational versus integration of wet (biological) data.
2. Stage of development, partnerships, and computational/data resources.

Source: Recursion

(Companies on this list are going to use AI tools, like, for example, NVIDIA hardware. However, we are focusing here on purely drug-discovery companies and not AI in general).

1. Schrödinger, Inc.

The company specializes in physics-based models to find the best possible molecule for a given goal, balancing out conflicting metrics like potency, solubility, half-life, synthesizability, etc.

It also uses machine learning, but the addition of a physics-based model allows it to be tested in entirely novel fields for which no data set exists to “train” the AI. This allows Schrödinger to go from 1 billion potential molecules to just 8 solid candidates in a matter of days, exclusively through digital calculation.

Source: Schrodinger

Schrödinger signed with Bayer a 5-year collaboration agreement in 2020 for revenue of $10M. The idea of the agreement is to use Schrödinger technology together with Bayer in-silico prediction models.

Another recent partnership is with Lilly in 2022, with up to $425M in total milestone payments for successful discovery.

Past collaborations included Takeda, Sanofi Bristol Myers Squibb, and other smaller pharmaceutical companies.

Overall, Schrödinger is building a growing portfolio, including more and more proprietary and fully-owned molecules. It currently has 8 products in its proprietary pipeline, with 2 in phase I of clinical trials. And 23 products in partnered programs and collaborations, with 5 in phase I and 3 in phase II of clinical trials.

Source: Schrodinger

While not pre-revenue, the company is still not profitable, focusing on expansion and R&D spending to improve its technology. It should not be a serious concern in the short term, as the company has several years of operation worth of cash on its balance sheet.

It is also looking at expanding toward new segments beyond drug discovery, like complex biopharmaceuticals or even materials like chemicals, batteries, or polymers.

Source: Schrodinger

Investors will want to keep an eye on the new collaborations, as they will reflect the advances of Schrödinger's technology, as assessed by the leaders in the industry, as well as possible success in expanding the core technology to new markets.

2. Exscientia

The company is using AI to develop precision therapies.

It runs a “full stack” AI drug discovery technology with dedicated software at every stage of the drug discovery process.

Source: Exscientia

Instead of looking at existing molecules, Exscientia’s Precision Design AI designs custom molecules to match the target found by its Precision Target AI.

Exscientia's technology reduces 70% of the time required to go from a biological target to finding a corresponding drug and an 80% more capital-efficient process.

This resulted in 4 compounds in early clinical stages, focusing mainly on oncology (cancer) and inflammatory diseases. The company shows $4B in pre-commercial milestones potential.

Source: Exscientia

The company is only starting to register revenues but has a long cash runaway worth several years of spending on its balance sheet.

By sitting at the junction between AI-drug discovery and precision therapy, Exscientia is aiming for two of the most transformative fields of medical science. Judging from the established partnership with Merck, Sanofi, and BMI, other established pharmaceutical companies also consider that the platform has great potential.

3. Recursion Pharmaceuticals, Inc

The more AIs get involved in drug discovery and development, the more data will become precious for training the AIs.

Biology is an extremely complex field, with integrated and verified data sometimes in short supply. This is a serious problem when any error will create bias, limitations, and errors in the AI, which might then need to be retrained from scratch.

Another issue is the sheer amount of data needed, far beyond the ability of humans to generate and organize manually.

It is to answer these limitations that Recursion has built its platform, with the mission of “turning drug discovery into a search problem”. It combines dry lab (in silico) and wet lab (biological samples) with:

• A library of 1.7 million small molecules.
• Cell cultures, CRISPR gene editing, soluble factors, live viruses, etc.
• An automated laboratory robotics workflow that allows for up to 2.2 million experiments each week.
• High-throughput microscopes and sequencing systems.
• Continuous video feeds from cameras, recording holistic measurements of animal behaviors.
• Advanced computational resources, which have generated >21 petabytes of proprietary high-dimensional data.
• ADMET (absorption, distribution, metabolism, excretion, and toxicology) data.

This creates unique (and massive) datasets of proteomics (protein levels), transcriptomics (mRNA levels), phenomics (cellular morphology), ADMET and “in-vivonomics” (animal behaviors). The company is also looking to add metabolomics and genomics to its datasets in the future.

This wealth of data allows for a wider range of candidates at an early stage but a narrower funnel moving forward. It should also accelerate the development of high-potential drug candidates.

Source: Recursion

The potential of this approach is not lost on pharmaceutical giants, with advancing collaborations in Fibrosis (Bayer) and Neuroscience (Roche-Genentech).

The company has also received a $50M investment by NVIDIA in July 2023, the new giant of AI computing hardware. This was right after Recursion acquired in May 2023 the drug chemistry-focused preclinical startups, Cyclica and Valance, for a total of $87.5M.

In total, the company is looking at $13B in potential milestones across 50+ possible programs plus royalties.

The partnerships with tech (NVIDIA) and pharmaceutical (Bayer, Roche) companies, as well as the acquisitions of smaller AI-drug startups, could indicate that Recursion is turning into a keystone of the industry and a potential serial acquirer, with the target to become one of the first company to scale biological data to previously unmatched volume.