Despite possessing advanced tools for gene editing and drug development, contemporary biotechnology faces a significant challenge: thousands of rare diseases still lack effective treatments. Leaders from Insilico Medicine and GenEditBio contend that the primary obstacle has historically been a scarcity of skilled individuals. They assert that artificial intelligence is now serving as a crucial force multiplier, enabling scientists to tackle complex issues previously neglected by the industry.
During Web Summit Qatar this week, Alex Aliper, president of Insilico, detailed his firm’s ambition to cultivate “pharmaceutical superintelligence.” Insilico recently introduced its “MMAI Gym” initiative, designed to elevate the performance of general-purpose large language models, such as ChatGPT and Gemini, to match that of specialized models.
Aliper articulated that the objective is to construct a multimodal, multitask model capable of concurrently resolving numerous drug discovery challenges with unparalleled accuracy.
In an interview with TechCrunch, Aliper emphasized, “This technology is essential for enhancing the productivity of our pharmaceutical sector and addressing the talent and labor deficits within it, given that countless diseases remain uncured and without treatment, alongside numerous neglected rare disorders.” He concluded, “Therefore, we require more intelligent systems to confront this challenge.”
Insilico’s platform processes biological, chemical, and clinical data to formulate hypotheses concerning disease targets and potential drug molecules. By automating processes that previously demanded extensive human labor from chemists and biologists, Insilico claims it can efficiently explore expansive design landscapes, identify superior therapeutic candidates, and even find new uses for current medications—achieving all this with significantly lower costs and faster timelines.
As an illustration, the company recently deployed its AI models to ascertain if existing pharmaceuticals could be adapted for the treatment of ALS, a rare neurological condition.
However, the scarcity of human resources extends beyond drug discovery. Even with AI pinpointing promising targets or treatments, numerous ailments necessitate interventions at a foundational biological scale.
GenEditBio is actively participating in the “second wave” of CRISPR gene editing, shifting the methodology from manipulating cells externally (ex vivo) to accurate internal delivery (in vivo). The firm’s ambition is to streamline gene editing into a single injection administered directly to the compromised tissue.
Tian Zhu, co-founder and CEO of GenEditBio, informed TechCrunch, “We have engineered a unique proprietary ePDV, or engineered protein delivery vehicle, which functions as a virus-like particle.” She added, “Our approach involves learning from natural processes and employing AI machine learning techniques to explore natural resources, identifying viruses that exhibit a specific tropism for particular tissue types.”
Zhu’s reference to “natural resources” pertains to GenEditBio’s extensive collection comprising thousands of distinct, nonviral, nonlipid polymer nanoparticles—which serve as sophisticated delivery systems engineered for the secure transportation of gene-editing instruments into designated cells.
The company states that its NanoGalaxy platform leverages AI to scrutinize data, discerning correlations between chemical structures and particular tissue targets (such as the eye, liver, or nervous system). Subsequently, the AI anticipates chemical modifications to a delivery vehicle that would enable it to transport a payload without eliciting an immune reaction.
GenEditBio conducts in vivo testing of its ePDVs within wet labs, with the outcomes then integrated back into the AI system to enhance its predictive precision for subsequent iterations.
Zhu emphasizes that effective, tissue-specific delivery is fundamental for in vivo gene editing. She asserts that her company’s methodology decreases production costs and streamlines a procedure that has traditionally posed scalability challenges.
Zhu stated, “It’s akin to acquiring a readily available drug effective for numerous patients, thereby rendering these medications more economical and universally accessible.”
Recently, her firm secured FDA approval to commence clinical trials for CRISPR therapy targeting corneal dystrophy.
Addressing the Ongoing Data Challenge
Similar to numerous AI-powered systems, advancements in biotechnology inevitably confront a data obstacle. Accurately modeling the complex and unusual aspects of human biology demands a substantially greater volume of high-quality data than is presently available to researchers.
Aliper articulated, “Additional ‘ground truth’ data derived from patients remains crucial.” He noted, “The existing data pool exhibits a strong bias towards the Western world, where it originates. I believe we must intensify local efforts to acquire a more evenly distributed collection of original, or ground truth, data, which will in turn enhance our models’ capabilities in processing it.”
Aliper mentioned that Insilico’s automated laboratories produce multi-layered biological data from disease samples on a large scale, entirely without human involvement, subsequently inputting this data into its AI-powered discovery platform.
Zhu posits that the essential data for AI is inherent within the human body, having been refined over millennia of evolution. A mere segment of DNA directly encodes proteins; the remainder functions more as an interpretive guide for gene expression. This information, historically challenging for humans to decipher, is becoming progressively approachable for AI models, exemplified by contemporary projects such as Google DeepMind’s AlphaGenome.
GenEditBio employs a comparable laboratory methodology, simultaneously evaluating thousands of delivery nanoparticles instead of sequentially. The resultant data sets, which Zhu labels as “gold for AI systems,” are then utilized to train their models and progressively foster collaborations with external entities.
Aliper indicated that a significant forthcoming endeavor involves constructing digital human twins to conduct virtual clinical trials, a methodology he describes as “still in its nascent stage.”
Aliper observed, “We currently face a stagnation, with approximately 50 drugs approved annually by the FDA, and expansion is necessary.” He further noted, “Chronic disorders are increasing as the global populace ages […] My aspiration is that within the next decade or two, we will possess an expanded array of therapeutic choices for individualized patient care.”