How AI is revolutionizing protein design for biologists and drug discovery

Artificial intelligence is reshaping drug development, but its full potential hinges on making advanced tools available to scientists without machine-learning expertise. OpenProtein.AI, founded by MIT alumni Tristan Bepler and Tim Lu, is addressing this gap by offering a no-code platform that provides biologists with access to powerful AI models for protein engineering, structure prediction, and functional analysis.

OpenProtein’s platform empowers researchers across pharmaceutical, biotech, and academic sectors to leverage foundation models like PoET (Protein Evolutionary Transformer) without requiring coding skills. The company’s tools are already in use by teams at institutions like Boehringer Ingelheim, where they accelerate protein design for therapies targeting cancer, autoimmune diseases, and inflammatory conditions.

Democratizing AI for biological research

Most biologists lack the time or training to implement machine-learning workflows, yet the need for faster, more precise protein design has never been greater. OpenProtein.AI’s platform bridges this divide by offering an intuitive web interface where researchers can upload data, train models, and analyze protein sequences—all without writing code. The platform’s open-source foundation models, including PoET, are trained on evolutionary protein data to predict functional properties and generate novel sequences.

"We built this platform because biologists want to focus on discovery, not infrastructure," says Bepler, OpenProtein’s CEO. "Our tools let them experiment with AI-driven design in real time, reducing the trial-and-error cycle in protein engineering." The PoET model, for example, learns from protein families to generate related sequences, enabling researchers to explore sequence-function relationships without retraining the model for each experiment.

From sequence to function: The AI-powered workflow

OpenProtein’s platform streamlines a workflow that traditionally involves months of manual experimentation. Researchers begin by inputting protein sequences or structural data, then use predictive models to generate variants with desired traits. These in silico designs are evaluated virtually before lab testing, significantly narrowing the pool of candidates for wet-lab validation.

The platform’s no-code interface supports several key functions:

• Protein structure prediction: Models analyze amino acid sequences to forecast 3D structures, even for proteins without known homologs.
• Functional design: AI generates protein variants optimized for stability, binding affinity, or enzymatic activity.
• Data integration: Users can fine-tune models with their own experimental results, improving accuracy over time.
• Cross-modality exploration: Beyond proteins, the tools are being adapted to model other biological molecules, expanding their utility.

"We’re not just building models for proteins—we’re creating a framework to describe biological systems more broadly," Bepler explains. "This could unlock new approaches in synthetic biology, metabolic engineering, and beyond."

Collaboration and real-world impact in drug discovery

OpenProtein.AI’s partnership with Boehringer Ingelheim exemplifies how AI-driven tools are transitioning from research labs to clinical applications. The pharmaceutical giant integrated OpenProtein’s models into its protein engineering pipeline in early 2025, with a recent expansion of the collaboration to include PoET-2, the platform’s updated protein language model. Preliminary benchmarks suggest PoET-2 outperforms earlier versions in accuracy and efficiency, further reducing the time required to identify viable drug candidates.

Academic institutions also benefit from OpenProtein’s free access tier, enabling students and researchers to experiment with AI models that were once the domain of specialized computational teams. "We’re removing barriers to entry," says Lu, OpenProtein’s co-founder and former MIT associate professor. "A biologist with a hypothesis can now test it in hours instead of weeks."

The future of AI in biology

As AI models grow more sophisticated, platforms like OpenProtein.AI are poised to play a central role in the next wave of biological innovation. The company is expanding its toolkit to support RNA design, enzyme engineering, and even multi-protein complex modeling. With advancements in generative AI and foundation models, the gap between computational prediction and experimental validation continues to shrink.

"The next frontier isn’t just making proteins—it’s understanding how they interact within larger biological networks," Bepler notes. "Our goal is to give scientists the tools to explore those questions faster and more creatively than ever before."