AI Powers Biotech’s Push to Treat Rare Diseases & Drug Discovery

The search for treatments for rare diseases, which collectively affect an estimated 300 million people worldwide, is gaining momentum thanks to advances in artificial intelligence. While biotechnology has made significant strides in gene editing and drug design, a critical bottleneck has long been the availability of skilled researchers to translate these tools into effective therapies. Now, AI is emerging as a powerful force multiplier, accelerating discovery and potentially lowering costs in a field often hampered by economic and logistical challenges.

Speaking at Web Summit Qatar, Alex Aliper, president of Insilico Medicine, outlined his company’s ambition to develop “pharmaceutical superintelligence.” Insilico’s approach centers on the “MMAI Gym,” a platform designed to train large language models – similar to ChatGPT and Gemini – to perform drug discovery tasks with what Aliper describes as “superhuman accuracy.” The goal is to create a versatile AI capable of tackling multiple stages of the drug development process simultaneously.

“We really need this technology to increase the productivity of our pharmaceutical industry and tackle the shortage of labor and talent in that space, because there are still thousands of diseases without a cure, without any treatment options, and there are thousands of rare disorders which are neglected,” Aliper explained. Insilico’s platform works by analyzing vast amounts of biological, chemical, and clinical data to generate hypotheses about potential disease targets and promising drug candidates. By automating tasks traditionally performed by teams of chemists and biologists, the company aims to significantly reduce both the time and expense associated with bringing new therapies to market. Recently, Insilico utilized its AI models to explore the possibility of repurposing existing drugs for the treatment of Amyotrophic Lateral Sclerosis (ALS), a rare and devastating neurological disorder.

The challenges in rare disease treatment extend beyond initial drug discovery. Even when promising targets are identified, delivering effective therapies can be complex, often requiring interventions at a fundamental biological level. GenEditBio is addressing this challenge through advancements in CRISPR gene editing technology, specifically focusing on in vivo gene editing – delivering the editing tools directly into the patient’s body.

“We have developed a proprietary ePDV, or engineered protein delivery vehicle, and it’s a virus-like particle,” said Tian Zhu, co-founder and CEO of GenEditBio. The company leverages AI and machine learning to analyze natural viruses, identifying those with a natural affinity for specific tissues. This knowledge is then used to design non-viral nanoparticles – the ePDVs – capable of safely transporting gene-editing tools to the affected cells.

GenEditBio’s NanoGalaxy platform employs AI to correlate chemical structures with specific tissue targets, predicting how modifications to the delivery vehicle’s chemistry will optimize payload delivery while minimizing immune responses. The company then tests these ePDVs in vivo, feeding the results back into the AI to refine its predictive capabilities. This iterative process, Zhu argues, reduces the cost of goods and standardizes a process that has historically been 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,” she said.

GenEditBio recently received FDA approval to begin clinical trials of a CRISPR therapy for corneal dystrophy, demonstrating the potential of this approach to translate into tangible benefits for patients.

Combating the Persistent Data Problem

Despite the promise of AI, progress in biotech remains dependent on the availability of high-quality data. Modeling the complexities of human biology requires a comprehensive understanding of edge cases, which necessitates access to significantly more patient data than is currently available. Aliper emphasized the need for more “ground truth data coming from patients,” noting that existing datasets are often biased towards populations in the Western world. He highlighted Insilico’s efforts to generate multi-layer biological data from disease samples at scale, using automated labs to minimize human intervention and feed the resulting data into its AI-driven discovery platform.

Zhu suggests that much of the data needed already exists within the human genome itself. She points out that only a small fraction of DNA directly codes for proteins, while the remaining portion regulates gene behavior. This regulatory information, historically difficult for humans to interpret, is becoming increasingly accessible to AI models, as exemplified by initiatives like Google DeepMind’s AlphaGenome. GenEditBio also employs a high-throughput experimental approach, testing thousands of delivery nanoparticles in parallel to generate large datasets – “gold for AI systems” – that are used to train its models and facilitate collaborations with external partners.

Looking ahead, Aliper envisions a future where digital twins of humans are used to conduct virtual clinical trials, a process he acknowledges is still in its early stages. He expressed hope that within the next 10 to 20 years, there will be a significant increase in therapeutic options for personalized treatment, noting that the FDA currently approves around 50 new drugs annually, a number that needs to grow to address the rising prevalence of chronic diseases associated with an aging global population.