How AI helps solve the labor problem in treating rare diseases

Modern biotechnology has the tools to edit genes and design drugs, yet thousands of rare diseases remain untreated. According to executives at Insilico Medicine and GenEditBio, the missing ingredient for years is finding enough smart people to keep the business going. They say AI has become a force multiplier that allows scientists to tackle problems that the industry has long left unsolved.

Speaking this week at Web Summit Qatar, Insilico’s president, Alex Aliper, outlined his company’s goal of developing “pharmaceutical superintelligence.” Insilico recently launched “MMAI Gym” that aims to train large, general language models, such as ChatGPT and Gemini, to perform as well as specialized models.

The goal is to build a multi-modal, multi-tasking model that, Aliper says, can solve many different drug discovery tasks simultaneously and with superhuman accuracy.

“We really need this technology to increase the productivity of our pharmaceutical industry and address the labor and talent shortage in this field, because there are still thousands of diseases without treatment, without any treatment options, and there are thousands of rare disorders that are being neglected,” Aliper said in an interview with TechCrunch. “So we need smarter systems to address this problem.”

The Insilico platform ingests biological, chemical, and clinical data to generate hypotheses about disease targets and candidate molecules. By automating steps that previously required legions of chemists and biologists, Insilico says it can sift through vast design spaces, filter out high-quality therapeutic candidates, and even repurpose existing drugs — all at dramatically lower cost and time.

For example, the company recently used its AI models to determine whether existing drugs could be repurposed to treat amyotrophic lateral sclerosis (ALS), a rare neurological disorder.

But the bottleneck in the work does not end when the drug is discovered. Even when AI can identify promising targets or treatments, many diseases require interventions at a more fundamental biological level.

GenEditBio is part of the “second wave” of CRISPR gene editing, where the process moves away from editing cells outside the body (ex vivo) and toward precise delivery inside the body (in vivo). The company’s goal is to make gene editing a one-time injection directly into affected tissue.

“We developed ePDV, or engineered protein delivery vehicle, which are virus-like particles,” Tian Zhou, co-founder and CEO of GenEditBio, told TechCrunch. “We learn from nature and use AI machine learning methods to mine natural resources and find types of viruses that have an affinity for specific types of tissue.”

The “natural resource” Chu is referring to is GenEditBio’s vast library of thousands of unique, non-viral, non-lipidic polymer nanoparticles — essentially delivery vehicles designed to safely deliver gene-editing tools into specific cells.

The company says its NanoGalaxy platform uses artificial intelligence to analyze data and determine how chemical structures bind to specific tissue targets (such as the eye, liver, or nervous system). The AI ​​then predicts adjustments to the chemistry of the delivery vehicle that will help it carry a payload without triggering an immune response.

GenEditBio tests ePDVs in vivo in wet labs, and the results are fed back to the AI ​​to improve its predictive accuracy for the next round.

Efficient, tissue-specific delivery is a prerequisite for in vivo gene editing, says Zhou. She argues that her company’s approach reduces the cost of goods and standardizes a process that has historically been difficult to scale.

“It’s like having a medicine ready [that works] “This treatment is available to many patients, making the medicines affordable and accessible to patients globally,” Zhou said.

Her company recently received FDA approval to begin trials of a CRISPR treatment for corneal dystrophy.

Combat the persistent data problem

As with many systems that rely on artificial intelligence, progress in biotechnology eventually faces a data problem. Modeling cutting-edge cases in human biology requires much more high-quality data than researchers can currently obtain.

“We still need more real data coming from patients,” Aliper said. “The dataset is very biased towards the Western world, where it is generated. I think we need to do more locally, to have a more balanced set of original data, or ground truth data, so that our models are also better able to deal with it.”

Insilico’s automated labs generate multi-layered biological data from disease samples at scale, without human intervention, and then feed it into an AI-driven discovery platform, Aliper said.

The data AI needs already exists in the human body, shaped by thousands of years of evolution, says Zhou. Only a small portion of DNA directly “codes” for proteins, while the rest acts as an instruction manual for how genes should behave. This information has historically been difficult for humans to interpret, but it is increasingly available to artificial intelligence models, including recent efforts like Google DeepMind’s AlphaGenome.

GenEditBio applies a similar approach in the lab, testing thousands of nanoparticles delivered in parallel rather than one at a time. The resulting datasets, which Zhou calls “the gold for AI systems,” are used to train their models and, increasingly, to support collaboration with external partners.

One of the next big efforts, according to Aliper, will be building digital twins of humans for virtual clinical trials, a process he says is “still in its infancy.”

“We’re at a point where about 50 drugs are approved by the FDA every year, and we need to see growth,” Aliper said. “There is a rise in chronic disorders because we are aging as a global population… I hope that in 10 to 20 years we will have more therapeutic options for personalized treatment of patients.”