In the realm of modern biotechnology, while tools for gene editing and drug design are advancing, numerous rare diseases still lack effective treatments. Executives from Insilico Medicine and GenEditBio highlight that a significant challenge has been the shortage of skilled professionals to drive this innovation forward. They assert that artificial intelligence (AI) is emerging as a crucial ally, enabling scientists to tackle previously neglected challenges in the field.
During a recent presentation at Web Summit Qatar, Alex Aliper, the CEO and founder of Insilico, shared insights into his company's vision for creating "pharmaceutical superintelligence." Insilico has introduced its "MMAI Gym," which aims to enhance generalist large language models, such as ChatGPT and Gemini, to perform at levels comparable to specialized models.
The ambition is to develop a multi-modal, multi-task model that, according to Aliper, can efficiently address various drug discovery tasks with exceptional accuracy.
"This technology is essential for boosting productivity in our pharmaceutical sector and addressing the workforce shortage, as there are still thousands of diseases without any treatment options," Aliper stated in an interview. "We require more advanced systems to confront this issue."
Insilico's platform processes biological, chemical, and clinical data to formulate hypotheses regarding disease targets and potential therapeutic molecules. By automating processes that once necessitated extensive human resources, Insilico claims it can navigate vast design spaces, identify high-quality therapeutic candidates, and even repurpose existing medications--all while significantly reducing time and costs.
For instance, the company recently utilized its AI models to explore the repurposing of existing drugs for treating ALS, a rare neurological condition.
However, the challenge of labor shortages extends beyond drug discovery. Even when AI identifies promising therapies, many diseases require interventions at a fundamental biological level.
GenEditBio is at the forefront of the "second wave" of CRISPR gene editing, transitioning from ex vivo techniques to precise in vivo applications. The company aims to enable gene editing through a single injection directly into affected tissues.
"We have created a proprietary engineered protein delivery vehicle (ePDV), resembling a virus-like particle," explained Tian Zhu, co-founder and CEO of GenEditBio. "We draw inspiration from nature and apply AI machine learning to discover which viruses are best suited for specific tissues."
The "natural resources" Zhu refers to include GenEditBio's extensive library of unique, non-viral, non-lipid polymer nanoparticles designed as delivery vehicles for gene-editing tools.
The company's NanoGalaxy platform employs AI to analyze data and uncover correlations between chemical structures and targeted tissues, such as the eye or liver. The AI predicts modifications to enhance the delivery vehicle's efficiency without provoking an immune response.
GenEditBio conducts in vivo testing of its ePDVs in wet labs, feeding the results back into the AI to enhance predictive accuracy for subsequent trials.
According to Zhu, efficient, tissue-specific delivery is crucial for successful in vivo gene editing. She argues that this approach reduces costs and standardizes a process that has historically been challenging to scale.
"It's akin to producing an off-the-shelf drug that is effective for multiple patients, making treatments more affordable and accessible globally," Zhu noted.
Her company has recently received FDA approval to initiate trials for a CRISPR therapy targeting corneal dystrophy.
Addressing the Ongoing Data Challenge
Like many AI-driven initiatives, advancements in biotechnology face a significant data challenge. Accurately modeling the complexities of human biology requires a wealth of high-quality data that is currently lacking.
"We still need more reliable data from patients," Aliper remarked. "Most existing data is heavily biased towards the Western world. It's essential to gather more localized efforts to create a balanced dataset, which will enhance our models' capabilities."
Aliper mentioned that Insilico's automated laboratories generate multi-layer biological data from disease samples at scale, without human intervention, which is then integrated into its AI-driven discovery platform.
Zhu believes that the necessary data already resides within the human body, shaped by millennia of evolution. Only a small portion of DNA directly codes for proteins, while the remainder serves as an instructional guide for gene behavior. This information has traditionally been challenging for humans to interpret, but AI models are increasingly adept at accessing it, as seen with projects like Google DeepMind's AlphaGenome.
GenEditBio employs a similar methodology in the lab, testing thousands of delivery nanoparticles simultaneously, rather than sequentially. The resulting datasets, which Zhu describes as "gold for AI systems," are utilized to train its models and foster collaborations with external partners.
Looking ahead, one of the significant initiatives, according to Aliper, is the development of digital twins of humans to conduct virtual clinical trials, a process he describes as still in its infancy.
"Currently, we are seeing around 50 new drugs approved by the FDA each year, and we need to see that number grow," Aliper expressed. "As the global population ages, chronic disorders are on the rise. My hope is that in the next 10 to 20 years, we will have expanded therapeutic options for personalized patient treatments."
