AI As A Force Multiplier In Biotech
Executives from Insilico Medicine and GenEditBio said this week that artificial intelligence is helping the industry tackle problems that have gone untreated for years because of a shortage of skilled labor, even as modern biotech already has tools to edit genes and design drugs. Speaking at Web Summit Qatar, Insilico president Alex Aliper said the missing ingredient has long been finding enough capable people to continue the work, and that AI is becoming a way to scale scientific effort across thousands of rare diseases that still lack treatments.
Aliper said Insilico’s aim is to develop what he called “pharmaceutical superintelligence.” The company recently launched its “MMAI Gym,” which is designed to train generalist large language models such as ChatGPT and Gemini to perform at the level of specialist models. The stated goal is to build a multimodal, multitask system that can handle many drug discovery tasks at once with higher accuracy.
Insilico’s Drug Discovery Platform
In an interview with TechCrunch, Aliper said the technology is needed to raise productivity in the pharmaceutical industry and to address labor and talent shortages, pointing to the large number of diseases without cures or treatment options. Insilico’s platform ingests biological, chemical, and clinical data and generates hypotheses about disease targets and candidate molecules. By automating steps that once required large teams of chemists and biologists, the company says it can search large design spaces, nominate therapeutic candidates, and repurpose existing drugs while reducing cost and time.
The company recently used its models to examine whether existing drugs could be repurposed to treat ALS, a rare neurological disorder. Insilico says this type of work shows how automation can compress stages of discovery that previously took much longer to complete.
Gene Editing And Delivery Challenges
The labor constraint, executives said, does not stop at drug discovery. GenEditBio is working on what it describes as the second wave of CRISPR gene editing, shifting from editing cells outside the body to delivering gene-editing tools directly inside the body. The company’s stated goal is to turn gene editing into a single injection into affected tissue.
GenEditBio co-founder and chief executive Tian Zhu told TechCrunch that the company has developed an engineered protein delivery vehicle, or ePDV, built as a virus-like particle. She said the company uses AI and machine learning to analyze natural viruses and identify which ones show affinity for specific tissues, then applies that knowledge to design delivery systems.
Zhu said the company maintains a large library of thousands of unique, nonviral, nonlipid polymer nanoparticles that act as delivery vehicles for gene-editing tools. According to the company, its NanoGalaxy platform uses AI to study how chemical structures correlate with tissue targets such as the eye, liver, or nervous system, and to predict which chemical changes will allow a payload to be delivered without triggering an immune response.
Testing, Scale, And Regulatory Progress
GenEditBio tests its ePDVs in vivo and feeds the results back into its models to refine predictions in subsequent rounds. Zhu said efficient, tissue-specific delivery is required for in vivo gene editing and argued that the company’s approach lowers production costs and standardizes a process that has been difficult to scale. She compared the goal to producing an off-the-shelf drug that can work for multiple patients, which she said would make treatments more affordable and accessible. The company recently received approval from the U.S. Food and Drug Administration to begin trials of a CRISPR therapy for corneal dystrophy.
Data Limits And Model Training
Both executives said progress in AI-driven biotech runs into limits set by data. Aliper said modeling edge cases in human biology requires more high-quality data than researchers currently have and noted that much of the existing data is biased toward Western populations. He said Insilico’s automated labs generate multilayer biological data from disease samples at scale without human intervention and feed that data into its discovery platform.
Zhu said the data needed by AI already exists in the human body, shaped by long evolutionary processes, but has been hard for humans to interpret. Only a small fraction of DNA codes for proteins, while the rest functions as instructions for gene behavior, a structure she said is increasingly accessible to AI models, including recent efforts such as Google DeepMind’s AlphaGenome. GenEditBio applies a similar approach by testing thousands of nanoparticles in parallel rather than one at a time, creating datasets that it uses to train its systems and to support collaborations with partners.
Digital Twins And Approval Rates
Aliper said one of the next major efforts will be building digital twins of humans to run virtual clinical trials, a process he described as still in its early stages. He said the industry is seeing a plateau of about 50 drugs approved by the U.S. Food and Drug Administration each year and argued that growth is needed as chronic disorders increase with an aging global population. He said his hope is that within 10 to 20 years there will be more therapeutic options for personalized treatment.
Featured image credits: MDabstract
For more stories like it, click the +Follow button at the top of this page to follow us.