Revolutionizing AIDS Research

May 19, 2025

Artificial intelligence—machines, systems, and apps that learn and improve from data to perform tasks that usually require human intelligence—already pervade many aspects of our daily lives. The mainstream commercial app ChatGPT, for example, boasts 400 million weekly active users worldwide. Needless to say, AI is revolutionizing the sciences, and biomedical research in particular.

Drowning in data
“Because the studies that we are all involved in produce such a huge amount of data—there’s immunology data, virology data, clinical data, host genetic data—all of these are resources that we can actually utilize for computer machine learning,” says virologist Dr. Guinevere Lee of Weill Cornell Medicine. “So instead of having a one-dimensional look at the data, we could have a machine that can actually analyze multiple dimensions of the data to see whether there are any correlations that we have missed.”

A computer scientist at the University of Illinois Urbana Champaign developed HINT (Hierarchical Interaction Network), a machine learning model that collates data to predict the success of a clinical trial based on the drug molecule, target diseases, and patient eligibility criteria at a greater rate than established methods.

Early diagnosis of HIV is essential for effective treatment and preventing transmission, but many people go undiagnosed until the disease has progressed. AI-powered diagnostic tools are helping to address this issue by improving the accuracy and accessibility of HIV testing.

AI algorithms can analyze medical images, blood tests, and patient data to identify subtle signs of HIV infection that might be missed by traditional methods. Additionally, AI-driven chatbots and mobile apps are providing people with risk assessments and encouraging them to seek testing, particularly in regions with limited access to healthcare.

In one study, deep learning algorithms were trained on a library of over 11,000 images to classify rapid HIV tests as positive or negative. Deployed as a mobile app in rural South Africa, the AI-based diagnostics approach outperformed visual interpretations by healthcare staff, reducing the number of false positives and false negatives.

Accelerating drug and vaccine development
One of the most promising applications of AI in HIV research lies in drug discovery and development. Traditional methods of identifying new drugs can take years and cost billions of dollars, but AI-driven techniques are speeding up this process. Machine learning algorithms can analyze vast libraries of chemical compounds and predict which ones are most likely to inhibit the HIV virus. This approach not only saves time but also reduces the cost of drug development.

AI is also being used to repurpose existing drugs. By analyzing data on how different drugs interact with the virus and the human body, AI can identify medications already approved for other diseases that could be effective in treating HIV. This method has the potential to fast-track the availability of new treatment options.

Using an AI algorithm, researchers at MIT identified a new drug candidate—halicin—an antibiotic capable of killing E. coli and many strains of bacteria that are resistant to treatment.

In addition, AI is accelerating the search for viable vaccine candidates. Machine learning models can analyze data from clinical trials, genetic studies, and viral behavior to identify the most promising vaccine targets. AI is also being used to simulate how the immune system responds to different vaccine formulations, allowing researchers to refine their approaches without the need for extensive animal or human testing.

Understanding elite controllers
“We have 40 years of data that potentially could be used for training AI,” says amfAR’s Director of Research Dr. Andrea Gramatica. “I’m particularly excited at the idea of trying to find a way to use all this data to understand elite controllers—the exceptionally rare individuals who, for reasons we don’t fully understand, are able to control the virus without ART. Perhaps AI can spot patterns in all that data to help us find the key to this phenomenon.”

“I think it’s a very exciting frontier,” says Dr. Brad Jones of Weill Cornell Medicine. “Initiatives to bring HIV researchers together with folks who work in AI and see what’s possible and what our data would need to look like going forward to make it as compatible as possible with AI absolutely needs to be happening. It’s clearly a direction that has a lot of promise.”

With the help of AI, a Harvard research team created an interactive 3D model of a small part of the human brain, creating a view through six layers of the cortex that is helping scientists understand better the behavior of neurons, how the brain functions, and potentially how to treat neurological disorders.

Potential pitfalls
While AI holds great potential, it also raises im­portant ethical questions. Data privacy is a prima­ry concern, as AI-driven tools often rely on sensi­tive health information. Ensuring that this data is collected, stored, and used responsibly is essen­tial. Bias in AI algorithms is another challenge. If the data used to train these models do not reflect diverse populations, the resulting insights and recommendations may not be applicable to ev­eryone. Researchers must prioritize inclusivity to avoid exacerbating health disparities.

Finally, the cost and accessibility of AI tech­nologies can limit their impact. Developing countries, where HIV prevalence is often high­est, may struggle to implement AI-driven solu­tions due to resource constraints. International collaboration and investment will be crucial to ensuring that AI’s benefits reach those who need them most. The challenge lies in harnessing this technology responsibly and ensuring its benefits are accessible to all.

Click Here to read more from the May 2025 issue of amfAR INNOVATIONS.

---