Biotech has never had more powerful tools. We can edit genes. We can design molecules in silico. We can simulate protein structures with impressive precision.

And yet, thousands of rare diseases still have no approved treatment.

According to leaders from Insilico Medicine and GenEditBio, the bottleneck is not ambition or technology. It is talent capacity. There simply are not enough highly specialized researchers to tackle the long tail of rare disorders.

AI is now emerging as the force multiplier that can change that equation.

From Drug Discovery to “Pharmaceutical Superintelligence”

At Web Summit Qatar, Insilico’s president, Alex Aliper, outlined the company’s ambition to build what he calls “pharmaceutical superintelligence.”

The idea is bold: train multimodal AI systems capable of handling multiple drug discovery tasks simultaneously, with superhuman accuracy.

Insilico recently introduced “MMAI Gym,” an initiative designed to train generalist large language models such as ChatGPT and Gemini to perform at specialist level across pharmaceutical workflows.

Instead of relying on separate narrow models, the goal is to build one integrated system that can:

Analyze biological, chemical, and clinical datasets
Generate hypotheses about disease targets
Propose candidate molecules
Evaluate repurposing opportunities
Prioritize high-quality therapeutic options

All in parallel.

This approach automates processes that traditionally required large teams of chemists and biologists. The result is dramatically faster candidate selection, reduced cost, and expanded exploration of therapeutic design space.

A recent example: Insilico used its AI platform to evaluate whether existing drugs could be repurposed for ALS, a rare neurodegenerative disorder. This kind of systematic screening would have been prohibitively labor-intensive in the past.

For an industry stuck at roughly 50 new FDA approvals per year, productivity gains are not incremental. They are existential.

The Second Wave of CRISPR: Precision In Vivo

Drug discovery is only part of the equation. Many rare diseases require intervention at the genetic level.

GenEditBio represents what many call the “second wave” of CRISPR innovation. Instead of editing cells outside the body (ex vivo), the company focuses on precise in vivo delivery, targeting affected tissues directly.

Its proprietary ePDV platform, engineered protein delivery vehicles, functions like virus-inspired particles designed to deliver gene-editing tools safely and accurately.

Using its NanoGalaxy platform, GenEditBio applies AI to:

Analyze how chemical structures correlate with tissue targeting
Predict modifications that improve delivery efficiency
Minimize immune response risks
Optimize nanoparticles for specific organs such as the liver, eye, or nervous system

The workflow is iterative. Wet lab experiments feed results back into the AI models, improving predictive accuracy with every cycle.

The outcome is not just scientific precision. It is standardization. Gene editing becomes closer to an “off-the-shelf” therapeutic model, potentially reducing cost of goods and increasing global accessibility.

The company recently received FDA approval to begin clinical trials for CRISPR therapy targeting corneal dystrophy, marking a significant milestone.

The Data Problem: AI’s Real Constraint

As with most AI-driven systems, the ultimate constraint is data.

Modeling edge cases in human biology requires high-quality, diverse ground truth datasets. Today, biomedical data remains heavily skewed toward Western populations.

Without more geographically balanced data generation, AI systems risk bias and reduced generalizability.

Insilico addresses part of this challenge through automated labs that generate multi-layer biological datasets at scale, feeding directly into its discovery models.

Meanwhile, advances like AlphaGenome demonstrate how AI is beginning to interpret the non-coding regions of DNA, historically difficult for humans to analyze. Only a small fraction of DNA encodes proteins; the rest regulates how genes behave. Unlocking this “instruction layer” may redefine therapeutic design.

GenEditBio’s high-throughput nanoparticle testing generates parallel datasets that it describes as “gold for AI systems.” These structured, experimental datasets are critical for training reliable predictive models.

What Comes Next: Digital Twins and Virtual Trials

Looking ahead, one of the most ambitious directions is the development of digital human twins for virtual clinical trials.

The concept remains early-stage, but the promise is compelling: simulate patient responses before entering expensive, time-consuming human trials.

For an aging global population and a growing burden of chronic disease, incremental gains will not be enough.

AI is not replacing scientists. It is amplifying them.

In rare disease treatment, where small patient populations make traditional R&D economically unattractive, AI could finally make the long tail viable.

The next 10 to 20 years may determine whether we stay on a plateau of limited annual drug approvals, or whether AI-enabled biotech unlocks a new era of personalized, scalable therapeutic innovation.

Source