By Logan Roberts and Susanna Camp
We’re based in San Francisco, the artificial intelligence capital of the world, and the bleeding edge of the AI industry. With decades of collective experience in industrial biotech, Hawkwood believes that the sector is primed to reap the rewards of computational advancements as much as any industry. Protein prediction, strain engineering, and bioprocessing have all emerged as hotspots for cutting-edge research and implementation of AI. We’re bullish about the developments but also recognize that implementation still requires careful resource management and scientific subject matter expertise. Here’s our take on the state of AI in industrial biotech.
One of the biggest bottlenecks in biotechnology is the ability to predict how proteins will fold and behave. Artificial intelligence has played a huge role in solving this problem. Just a few weeks ago, the Nobel Prize in Chemistry was awarded to three scientists who pioneered applications of AI in protein prediction with AlphaFold and RoseTTAFold. These platforms have slashed the time required to understand the structure and properties of proteins from several years to a matter of minutes. Big tech companies have also followed suit in training pioneer AI/ML models to understand, predict, and even design new protein structures. Meanwhile, biotechnology companies are rolling out their own market-specific models; for example, Novozymes, a world leader in enzyme engineering, has developed a tool to predict how their detergent proteins will behave during laundry cycles.
Artificial intelligence has dramatically improved our ability to screen and optimize microbial strains to make bioproducts via fermentation. Traditionally, biologists make microbes better producers over time by altering their environment and forcing them to adapt – a process called directed evolution. Advancements in genetic engineering technologies like CRISPR, coupled with machine learning, have shown the ability to accelerate the Design-Build-Test-Learn cycle by 25-35%. Companies have released biology-specific computer-aided design platforms to engineer strains as a service, while product companies have invested in entire departments dedicated to creating and improving strains.
Machine learning has also begun to play a role in analyzing vast amounts of bioprocessing data. AI platforms are now able to automate process optimization, maintenance, quality control, and data integration. Manufacturing, and especially industrial-scale bio-manufacturing, is extremely expensive. Reducing failed or unoptimized runs can cut hundreds of thousands or millions of dollars off companies’ bottom lines – critical when capital is precious, which is the case for much of the industry.
Hawkwood has witnessed the arrival of the “next big thing” in industrial biotech many times over. When it comes to artificial intelligence, we’re seeing that the race to develop the best model, trained on the most data, can drain millions of dollars – both in units of staffing and kilowatt hours. In our industry, companies have repeatedly been unable to recoup exorbitant investment into AI infrastructure. What separates successes and failures is often a dedicated focus to a product that the market will want, championed by a lean, cross-functional team. In the case of AI, this means establishing collaborative teams of computer scientists and biologists, with each party iteratively learning from the other.
While AI has quickly emerged as a tool that can benefit almost any industry, it is critical to remember that its utility right now is to increase human efficiency and optimize resources. The best way to achieve these goals is to let subject matter expertise guide the technology. We think it’s significant that the Nobel Prize winners were cross-trained in many scientific disciplines. Demis Hassabis, best known for his pioneering influence in artificial intelligence, also has a PhD in cognitive neuroscience. John Jumper has both computer science and chemistry degrees. David Baker is a biochemist and computational biologist. Domain knowledge remains indispensable for the next generation of AI in biotech.
contact@hawkwoodbiotech.com
©Hawkwood Biotech Partners 2023, All Rights Reserved