AbSci and Other AI-Powered Biotechs Lead the Way in Antibody Discovery

by Andrii Buvailo        Biopharma insight

Disclaimer: All opinions expressed by Contributors are their own and do not represent those of their employers, or BiopharmaTrend.com.
Contributors are fully responsible for assuring they own any required copyright for any content they submit to BiopharmaTrend.com. This website and its owners shall not be liable for neither information and content submitted for publication by Contributors, nor its accuracy.

   702    Comments 0
Topics: Biotech Companies   

The monoclonal antibodies (mAbs) market has seen impressive growth in recent years, with the top-selling drugs increasingly being mAbs. The success of these therapeutics has driven the pharma industry to seek innovative methods for refining and improving their antibody pipelines. Artificial intelligence (AI) and deep learning, which have already revolutionized small molecule drug design, are now making significant strides in the development and optimization of antibodies.

Traditional mAb discovery involves experimental techniques like hybridoma screening or high throughput platforms such as yeast or phage display, followed by computationally guided mutagenesis or evolution of the antibody sequence. AI companies have made substantial progress in optimizing antibodies for binding, solubility, yield, and immunogenicity, among other properties. Some companies are even working on de novo design of antibody candidates, bypassing time-consuming and costly experimental methods.


Zero-shot moonshot

Absci has achieved a significant milestone in generative AI drug development: they are the first to design and validate de novo therapeutic antibodies using a zero-shot generative AI approach. Zero-shot methods involve designing antibodies that bind to specific targets without relying on any training data from known antibodies binding to those particular targets.

Zero-shot learning in the artificial intelligence field refers to a learning method where a model is able to recognize and classify new, previously unseen objects or data without having any prior examples or training data related to those specific objects. This is achieved by leveraging the model's existing knowledge and understanding of related objects or concepts and applying it to the new, unseen data.

In traditional machine learning approaches, a model is trained on a set of examples for each class or object it needs to recognize. In zero-shot learning, however, the model uses its existing knowledge base, often in the form of semantic relationships or shared attributes between known and unknown classes, to make predictions about the new, unseen data.

AbSci started its journey in 2011 in the city of Vancouver, Washington, US, and a decade later went public having raised about $200 million in an IPO in 2021. AbSci aims to change the biopharmaceutical industry with their protein production platform. Their advanced technology enables the fast and cost-effective production of complex proteins, including monoclonal antibodies, which helps create new treatments.

Zero-shot generaative AI

This approach matters because Absci's zero-shot model generates antibody designs that differ from those found in existing antibody databases, including de novo versions of all three heavy chain CDRs (HCDR123), the antibody regions most crucial for target binding. The effectiveness of this method has been demonstrated by validating against more than 100,000 antibodies, where the hit rate was found to be up to five to 30 times greater than the biological baselines examined.


A wave of companies applying artificial intelligence for antibody design

AbSci is not the only company applying artificial intelligence for designing monoclonal antibodies. In Vancouver, Canada, AbCellera was founded in 2012. The company achieved substantial business success, having conducted the largest biotech IPO in Canada in 2020, with $555.5 million. AbCellera focuses on antibody discovery, using their innovative platform to explore, decode, and analyze natural immune systems. By doing so, they find antibodies that can be developed into drugs to prevent and treat various diseases.

In Boston, Generate Biomedicine was founded in 2018 with a goal for advancing antibody discovery. With $420 million in funding, the company developed a unique approach that combines sequence and structure-based methods for designing proteins, including antibodies. Their innovative generative AI and automation platform accelerates the synthesis and testing of antibodies, making the discovery process more efficient.

On the other side of the United States, in California, BigHat Biosciences entered the scene in 2019, backed by a $100 million investment. Their integrated platform brings together high throughput synthesis and testing of antibodies with state-of-the-art AI-powered optimization. Utilizing a sequence-based approach and advanced machine learning models, they optimize antibody properties such as binding, solubility, immunogenicity, and yield.

Over in Israel, Biolojic Design has been making strides in the field since 2009. They employ a structure-based approach, training their model on millions of antibody-antigen pairs to identify template antibodies for specific targets. They then use a separate machine learning model to predict mutations and guide the evolution of the identified template, enhancing affinity and other biophysical properties. Biolojic Design has partnerships with Nektar Therapeutics and Eli Lilly for developing antibodies against autoimmune diseases and diabetes, respectively.

In 2022, Biolojic Design brought its computationally designed antibody, AU-007 into human clinical trials for the first time, targeting cancer patients. Aulos Bioscience, a Biolojic spinoff, is conducting the clinical trials in Australia and recruiting patients, with interim data expected later this year. The trial aimed to evaluate the monoclonal antibody, designed using Biolojic's proprietary AI platform, which harnesses the body's interleukin-2 (IL-2) to eradicate solid cancer tumors. 


Image credit: Aulos

In February 2023, Aulos Bioscience reported that the first patient in the United States received a dose of AU-007 in its Phase 1/2 clinical trial.


The Growing Interest in AI for Antibody Discovery:

Biopharma companies have taken note of the potential of AI platforms for antibody discovery and are exploring collaborations with AI companies or building internal capabilities. For example, Amgen partnered with Generate Bio in a deal worth up to $1.9 billion plus royalties, while Chugai published promising results from their internal AI platform, MALEXA-LI, for antibody discovery.

In addition, AstraZeneca, Merck, Pfizer, and Teva, along with Amazon Web Services (AWS) and the Israel Biotech Fund (IBF), have launched AION Labs, an incubator to accelerate AI-driven antibody discovery.

Mergers and acquisitions have become more common in this space, as seen with Evotec's acquisition of Just Biotherapeutics, AbSci's acquisitions of Denovium Inc. and Totient Bio, and Genentech's acquisition of Prescient Design.


In conclusion, AI-driven antibody discovery holds great promise for revolutionizing the development of novel therapeutics. Despite the challenges, such as limited training data, sequence and structural complexity, multi-objective optimization, interpretability, validation, generalizability, and integration with experimental workflows, innovative companies like AbSci, Generate Biomedicine, BigHat Biosciences, and Biolojic Design are finding ways to address these limitations. By continually refining their AI algorithms, incorporating new data sources, and developing robust experimental validation processes, these companies are paving the way for groundbreaking advancements in antibody design and drug discovery. As the field continues to evolve, we can expect to see even more impressive breakthroughs and practical applications of AI-driven antibody discovery in the near future.

Topics: Biotech Companies   

Share this:              


There are no comments yet. You can be the first.

Leave a Reply

Your email address will not be published. Required fields are marked *