Under the partnership, Boehringer will leverage an AI model developed by IBM, initially pre-trained and subsequently fine-tuned using additional proprietary data from Boehringer
Boehringer Ingelheim partners with IBM to accelerate antibody discovery. (Credit: IBM)
Boehringer Ingelheim and IBM have entered into an agreement that allows Boehringer to leverage IBM’s foundational model technologies for the identification of innovative candidate antibodies in the pursuit of developing effective therapeutics.
Boehringer will leverage an AI model developed by IBM, initially pre-trained and subsequently fine-tuned using additional proprietary data from Boehringer.
Boehringer Ingelheim biotherapeutics discovery global head Andrew Nixon said: “We are very excited to collaborate with the research team at IBM, who share our vision of making in silico biologic drug discovery a reality.
“I am confident that by joining forces with IBM scientists we will develop an unprecedented platform for accelerated antibody discovery which will enable Boehringer to develop and deliver new treatments for patients with high unmet needs.”
IBM research accelerated discovery vice president Alessandro Curioni said: “IBM has been at the forefront of creating generative AI models that extend AI’s impact beyond the domain of language.
“We are thrilled to now bring IBM’s multimodal foundation model technologies to Boehringer, a leader in the development and manufacturing of antibody-based therapies, to help accelerate the pace at which Boehringer can create new therapeutics.”
The treatment of various diseases, such as cancer, autoimmune disorders, and infectious diseases, heavily relies on therapeutic antibodies. Despite significant technological advancements, the discovery and development of therapeutic antibodies across diverse epitopes remain intricate and time-consuming.
Boehringer Ingelheim and IBM researchers are collaborating to expedite the antibody discovery process using in-silico methods. Leveraging sequence, structure, and molecular profile information of disease-relevant targets, as well as key criteria for effective antibody molecules (such as affinity, specificity, and developability), they aim to generate new human antibody sequences through in-silico techniques.
This innovative approach is facilitated by IBM’s foundation model technologies, specifically designed to enhance the speed and efficiency of antibody discovery while improving the quality of predicted candidates. These technologies, successful in generating biologics and small molecules with relevant target affinities, are employed to design antibody candidates for defined targets. Subsequently, these candidates are screened using AI-enhanced simulations to select and refine the most effective binders for the target.
Boehringer Ingelheim will validate the selected candidates by producing them in mini-scales and conducting experimental assessments. The results from these laboratory experiments will then be utilised to enhance the in-silico methods through feedback loops.
As part of Boehringer’s commitment to advancing drug discovery, the collaboration extends to leading academic and industry partners, contributing to the establishment of a prominent digital ecosystem. This ecosystem aims to accelerate drug discovery and development, offering breakthrough opportunities to improve patients’ lives.
This collaborative effort aligns with IBM’s broader initiatives to leverage generative AI and foundation models for accelerating the discovery and creation of new biologics and small molecules. In a recent milestone, IBM’s generative AI model efficiently predicted the physico-chemical properties of drug-like small molecules earlier in the year.
IBM’s biomedical foundation model technologies draw upon a diverse range of publicly available datasets, including protein-protein interactions and drug-target interactions, to develop pre-trained models. These models are then fine-tuned using specific proprietary data from IBM’s partners, enabling the creation of newly designed proteins and small molecules with desired properties.