A more deadly strain of mpox is spreading across Africa, with more than 17,000 suspected cases and a death toll already surpassing last year’s total. With the more severe Clade I variant driving the uptick, mpox continues to pose a significant threat to global health with the World Health Organization (WHO) classifying it as a public health emergency on August 14.

With the COVID-19 pandemic catalyzing the maturation of data and algorithmic approaches in epidemiology, a growing number of AI-based tools are emerging that can help health authorities respond to this evolving threat.

On digital staining and other tools

Paul Pallath, Ph.D., vice president of applied AI at Searce, a technology consulting firm, notes that a variety of AI-powered tools are assisting in rapid diagnostic testing and predicting outbreak patterns. Use cases for AI in mpox range from helping identify or repurpose drugs to identify patients for clinical trials. This latter application, as Pallath explains, offers tangible efficiency gains: “The other use case is about identifying the cohort of people that are the near match for the kind of drug they’re trying to test efficacy on once it gets to clinical trials,” he said. “Identifying subjects for clinical trials is a highly complicated and long process. AI is able to look at all possible combinations to figure out the subjects that should be part of the clinical trial.”

One hot spot is digital pathology. For instance, AI algorithms are being used to automatically analyze digital slides for specific histopathological features of mpox. For instance, some of Searce’s customers are using AI-powered digital staining techniques, which involve converting high-resolution tissue images and employing convolutional neural networks (CNNs) to virtually apply the desired stain colors. “This allows increasing or decreasing the staining without using chemicals,” Pallath said.

This digital staining process represents a significant advance. “Typically, in drug discovery, researchers take tissue cells from various animals and different body parts, stain it with chemicals, and look at the impacted area,” Pallath said. This process, as Pallath explains, was not only time-consuming and expensive but also posed significant health hazards. “And if the staining is not done right, they have to redo the entire process,” he added. Digital staining can help eliminate the need for these chemicals, promising a safer, faster, and more cost-effective alternative. By converting high-resolution tissue images and applying virtual stains using CNNs, researchers can now analyze drug impact with greater precision and efficiency. This technique holds promise for accelerating the development of effective treatments for mpox and other diseases.

Scrutinizing infected tissues

Beyond digital staining, CNNs also can help analyze infected tissues. Pallath explained that these AI-powered tools can differentiate between normal and infected tissue, ultimately classifying the type of infection as malignant or benign. This capability is particularly valuable for pharmaceutical companies researching the effects of potential mpox treatments. By precisely identifying and classifying infected areas, researchers can gain a deeper understanding of how the virus interacts with host tissues and how different drugs impact the disease progression. “That’s one use case we’re currently helping a large pharma with,” Pallath said.

AI tools also can help drug companies research the effects of potential mpox treatments. By precisely identifying and classifying infected areas, researchers can gain a deeper understanding of how the virus interacts with host tissues and how different drugs impact the disease progression. “That’s one use case we’re currently helping a large pharma with,” Pallath said.

Pallath notes that AlphaFold 3, DeepMind’s AI platform that can predict structures and interactions of an array of biomolecules, including proteins, DNA, RNA, and small molecules, can help scientists understand the interplay of proteins involved in the mpox infection process. “There are multiple layers to the mpox infection,” Pallath said. “One is at the surface level — the lesions and challenges it brings to people.” But there are also internal biological changes involving gene structures and proteins within the body

Platforms like AlphaFold are empowering researchers to explore into the molecular mechanisms of mpox infection by analyzing protein structures. In finding proteins that exhibit resilience to infection, scientists can gain identify potential targets for therapeutic interventions. “Understanding these 3D protein structures allows creating vaccines faster by narrowing down the proteins that are impactful or the reason for the infection and spread,” Pallath said. “It allows drug companies to focus on those protein structures.”

New tools for gauging potential mpox treatments

AI tools also can help drug companies research the effects of potential mpox treatments. By precisely identifying and classifying infected areas, researchers can gain a deeper understanding of how the virus interacts with host tissues and how different drugs impact the disease progression. “That’s one use case we’re currently helping a large pharma with,” Pallath said.

Additionally, AI holds promise in predicting the efficacy of potential drugs. “Another use case we had initial discussions around but did not implement is the efficacy of the drug itself,” Pallath said. This involves developing AI models that can predict factors such as solubility, excretion rate, duration in the body, and potential toxic effects.

Beyond drug discovery, AI is also proving valuable in diagnostics where CNNs can analyze high-resolution images of skin lesions.

Finally, AI is poised to play a role in the early detection of new variants. “Detecting mutations is a highly compute-intensive process,” Pallath said. “Having AI identify changes in the structure of the viruses themselves is an AI problem.”.

Digital groundwork

One silver lining of the COVID-19 pandemic was the digital maturation it drove, especially in fields such as epidemiology and data science. COVID was “unprecedented,” Pallath said, in both its global reach and its diverse impact on human health. “When dealing with such a scale of impact, AI as a technology simplifies a lot,” Pallath explained. For instance, it helped with clinical trial design by helping narrow down and segment patient cohorts. Algorithms can offer insights into which patient populations are most likely to benefit from specific treatments or vaccines. By analyzing vast datasets and identifying patterns, AI tools can help researchers understand the varying impacts of a disease, predict its trajectory, and develop targeted interventions.

The emergence of large language models in the wake of the pandemic has both pros and cons. “Making data available [for research] is important, but you need to ensure it’s factual data, because now with LLMs and generative AI, there could be a lot of data generated with just prompts,” Pallath said. “Medically certified data, certified by hospitals, ensuring there is some sort of digital watermark or verification that it’s factual data you can work with, versus data generated just for experimental purposes by LLMs or others, is going to be key as well.”

“Structure of Mpox virus poxin (2024) https://doi.org/10.2210/pdb8c9k/pdb“