Profluent’s Use of Generative AI in Protein Design: A New Frontier in Biotechnology

Profluent, a biotechnology company based in Berkeley, California, is pioneering the application of generative AI to protein design. By leveraging large language models trained on vast datasets of protein sequences, Profluent aims to revolutionize the way proteins are designed and utilized in various applications, from therapeutics to gene editing.(Bakar Labs)

🧬 ProGen3: Scaling Protein Design with AI

At the heart of Profluent’s approach is ProGen3, a suite of billion-parameter language models trained on over 3.4 billion protein sequences. These models enable the generation of novel full-length proteins and the redesign of specific protein domains to enhance functionality. Notably, ProGen3 employs a sparse architecture, achieving a fourfold speedup without compromising performance .(profluent.bio)

🧪 OpenAntibodies and OpenCRISPR: AI-Designed Therapeutics

Profluent has developed OpenAntibodies, AI-designed antibodies targeting 20 drug targets associated with diseases affecting approximately 7 million patients. These targets have a combined historical sales value of $660 billion .(profluent.bio)

In the realm of gene editing, Profluent introduced OpenCRISPR-1, the world’s first open-source, AI-generated gene editor. This novel editor demonstrates activity and specificity comparable to SpCas9, a widely used gene-editing enzyme, despite having a highly dissimilar sequence .(BioSpace, profluent.bio)

🔬 ProseLM: Precision in Protein Design

To enhance the precision of protein design, Profluent developed proseLM (protein structure-encoded language model). This model incorporates structural and functional context into protein language models, enabling fine-tuned design of proteins for specific tasks. Experimental validations have shown that proseLM can improve gene editing activity and therapeutic antibody binding affinity, outperforming traditional methods in certain cases .(GEN)

📈 Implications for Biotechnology

Profluent’s integration of generative AI into protein design signifies a shift from traditional discovery-based methods to intentional, data-driven design. This approach has the potential to accelerate the development of new therapeutics, enhance the specificity and efficacy of gene editing tools, and open new avenues in synthetic biology.