AI Revolutionizes Treatment of Rare Diseases Amidst Labor Shortages
In the realm of biotechnology, the advent of gene editing and drug design technologies has opened new frontiers in medical science. Despite these advancements, a significant number of rare diseases remain without effective treatments. A primary obstacle has been the scarcity of skilled professionals dedicated to this specialized field. However, artificial intelligence (AI) is emerging as a transformative force, enabling scientists to address challenges that have long been neglected.
At the recent Web Summit Qatar, Alex Aliper, president of Insilico Medicine, discussed the company’s mission to develop pharmaceutical superintelligence. Insilico has introduced the MMAI Gym, a platform designed to train generalist large language models, such as ChatGPT and Gemini, to perform on par with specialized models. The objective is to create a multimodal, multitask model capable of simultaneously tackling various drug discovery tasks with superhuman precision.
Aliper emphasized the necessity of this technology to enhance the productivity of the pharmaceutical industry and to address the shortage of labor and talent in the sector. He highlighted the existence of thousands of diseases lacking cures or treatment options, many of which are rare disorders that have been overlooked. Intelligent systems are essential to confront this issue effectively.
Insilico’s platform integrates biological, chemical, and clinical data to generate hypotheses about disease targets and potential therapeutic molecules. By automating processes that traditionally required extensive human resources, the company can explore vast design spaces, identify high-quality therapeutic candidates, and even repurpose existing drugs, all while significantly reducing costs and time. For instance, Insilico recently utilized its AI models to assess whether existing drugs could be repurposed to treat amyotrophic lateral sclerosis (ALS), a rare neurological disorder.
The labor bottleneck extends beyond drug discovery. Even when AI identifies promising targets or therapies, many diseases necessitate interventions at a more fundamental biological level.
GenEditBio represents the second wave of CRISPR gene editing, transitioning from editing cells outside the body (ex vivo) to precise delivery inside the body (in vivo). The company’s goal is to develop gene editing as a one-time injection directly into the affected tissue.
Tian Zhu, co-founder and CEO of GenEditBio, explained that the company has developed a proprietary engineered protein delivery vehicle (ePDV), a virus-like particle. By leveraging AI and machine learning, GenEditBio analyzes natural resources to identify viruses with affinities for specific tissues.
The natural resources refer to GenEditBio’s extensive library of thousands of unique, nonviral, nonlipid polymer nanoparticles. These serve as delivery vehicles designed to safely transport gene-editing tools into specific cells. The company’s NanoGalaxy platform employs AI to analyze data and determine how chemical structures correlate with specific tissue targets, such as the eye, liver, or nervous system. The AI then predicts which modifications to a delivery vehicle’s chemistry will enable it to carry a payload without triggering an immune response.
GenEditBio conducts in vivo testing of its ePDVs in wet labs, feeding the results back into the AI to refine its predictive accuracy for subsequent iterations.
Efficient, tissue-specific delivery is a prerequisite for in vivo gene editing. Zhu argues that the company’s approach reduces production costs and standardizes a process that has historically been challenging to scale. This method aims to provide off-the-shelf drugs effective for multiple patients, making treatments more affordable and accessible globally. GenEditBio recently received FDA approval to commence trials of CRISPR therapy for corneal dystrophy.
Combating the Persistent Data Problem
As with many AI-driven systems, progress in biotechnology ultimately encounters a data challenge. Modeling the complexities of human biology requires more high-quality data than is currently available.
Aliper noted the need for more ground truth data from patients, emphasizing that existing data is heavily biased toward the Western world. He advocates for local efforts to create a more balanced dataset, enabling models to be more capable of addressing diverse populations. Insilico’s automated labs generate multilayer biological data from disease samples at scale, without human intervention, which is then integrated into its AI-driven discovery platform.
Zhu highlighted that the data AI requires already exists within the human body, shaped by thousands of years of evolution. While only a small fraction of DNA directly codes for proteins, the remainder acts as an instruction manual for gene behavior. This information has traditionally been difficult for humans to interpret but is increasingly accessible to AI models, including recent efforts like Google DeepMind’s AlphaGenome.
GenEditBio applies a similar approach in the lab, testing thousands of delivery nanoparticles in parallel rather than individually. The resulting datasets, referred to as gold for AI systems, are used to train its models and support collaborations with external partners.
Looking ahead, Aliper envisions building digital twins of humans to conduct virtual clinical trials, a process still in its early stages. He expressed hope that in 10 to 20 years, there will be more therapeutic options for personalized patient treatment, addressing the rise in chronic disorders associated with an aging global population.
