While modern biotechnology possesses the capability to edit genes and design drugs, thousands of rare diseases still lack treatments. Executives from Insilico Medicine and GenEditBio point out that for years, the limiting factor has been a shortage of skilled researchers to advance the work. They argue artificial intelligence is now emerging as a critical multiplier, enabling scientists to address challenges the industry has historically overlooked.

Speaking at Web Summit Qatar, Insilico’s CEO and founder Alex Aliper detailed his company's objective to create what he terms "pharmaceutical superintelligence." The firm recently introduced its "MMAI Gym," an initiative designed to train generalist large language models, such as ChatGPT and Gemini, to match the performance of specialized models. The ultimate aim is to construct a multi-modal, multi-task model capable of solving numerous drug discovery tasks at once with what Aliper describes as superhuman accuracy. "So we need more intelligent systems to tackle that problem," he stated.

Insilico’s platform processes biological, chemical, and clinical data to formulate hypotheses about disease targets and potential drug molecules. The company asserts that by automating steps traditionally requiring large teams of chemists and biologists, it can explore vast design spaces, select high-quality therapeutic candidates, and repurpose existing drugs—all while significantly cutting costs and time. For instance, Insilico recently employed its AI models to assess whether approved drugs could be reused to treat ALS, a rare neurological condition.

However, the challenge extends beyond drug discovery. Even when AI identifies promising targets or therapies, many diseases necessitate intervention at a more fundamental biological level. GenEditBio represents part of a "second wave" in CRISPR gene editing, shifting the focus from editing cells outside the body to achieving precise delivery inside the body. The company's objective is to enable gene editing through a single injection directly into affected tissue. "We learn from nature and use AI machine learning methods to mine natural resources and find which kinds of viruses have an affinity to certain types of tissues," explained a company representative.

These 'natural resources' refer to GenEditBio’s extensive library containing thousands of unique, nonviral, nonlipid polymer nanoparticles—essentially delivery vehicles engineered to safely transport gene-editing tools into specific cells. The company's NanoGalaxy platform utilizes AI to analyze data and determine how chemical structures correlate with particular tissue targets, such as the eye, liver, or nervous system. The AI subsequently predicts which chemical adjustments to a delivery vehicle will help it carry its payload without provoking an immune response. GenEditBio tests its delivery vehicles in live lab settings, and the results are fed back into the AI to enhance predictive accuracy for future iterations.

According to the company, efficient, tissue-specific delivery is a fundamental requirement for in vivo gene editing. This approach is said to lower costs and standardize a process historically difficult to scale. "It’s like getting an off-the-shelf drug [that works] for multiple patients, which makes the drugs more affordable and accessible to patients globally," the representative noted. GenEditBio recently received FDA approval to commence trials of a CRISPR therapy for corneal dystrophy.

Progress in AI-driven biotech ultimately confronts a persistent data challenge. Modeling the complexities of human biology demands far more high-quality data than is currently available to researchers. "We still need more ground truth data coming from patients," Aliper emphasized. "The corpus of data is heavily biased over the western world, where it is generated. I think we need to have more efforts locally, to have a more balanced set of original data, or ground truth data, so that our models will also be more capable of dealing with it."

Aliper mentioned that Insilico’s automated labs generate multi-layered biological data from disease samples on a large scale without human involvement, which is then fed into its AI-driven discovery platform. From another perspective, the necessary data for AI already exists within the human body, refined by millennia of evolution. Only a small portion of DNA directly codes for proteins, while the remainder functions more like an instruction manual for gene behavior. This information has long been difficult for humans to interpret but is becoming increasingly accessible to AI models, including recent projects like Google DeepMind’s AlphaGenome.

In the lab, GenEditBio applies a comparable approach, testing thousands of delivery nanoparticles simultaneously rather than individually. The resulting datasets, deemed "gold for AI systems," are used to train its models and are increasingly shared to support collaborations with external partners. Looking ahead, Aliper identified one of the next major initiatives as the development of digital human twins to conduct virtual clinical trials, a process he describes as still in its early stages.

"We’re in a plateau of around 50 drugs approved by the FDA every year annually, and we need to see growth," Aliper remarked. "There is a rise in chronic disorders because we are aging as a global population […] My hope is in 10 to 20 years, we will have more therapeutic options for the personalized treatment of patients."