At Boehringer, AI develops new antibody-based medicines


technologies

In terms of artificial intelligence, Boehringer uses a pre-trained model from IBM, enriched with its own proprietary data. (Photo: Boehringer Ingelheim)


In its search for new antibody-based treatments to fight serious diseases, the pharmaceutical laboratory Boehringer Ingelheim is counting on AI. Goal: identify new drug candidates faster.

AdvertisingTo accelerate the development of new medicines, the family company Boehringer Ingelheim, founded in 1885 and which today is the 15th laboratory worldwide, is betting on artificial intelligence, analysis and data science. To do this, the company also created its own Global Computational Biology and Digital Sciences (gCBDS) division to use large amounts of data to research new active substances.

This research is particularly aimed at isolating new therapeutic antibodies. Antibody treatments are an important weapon in the fight against serious diseases such as cancer, autoimmune diseases and infectious diseases. In addition, they appear to have fewer side effects than traditional medicine, raising hopes for new treatments.

However, the development of new antibodies requires a lot of time and work. It requires not only meticulous precision but also extensive series of laboratory tests.

AI rather than lab tests

Hence Boehringer’s desire to accelerate this process with “in silicon” methods. In other words, the aim is to speed up the search for new antibodies using computer simulations. According to Andrew Nixon, director of biotherapeutic research at Boehringer Ingelheim, the goal was clear: “In-silicon development should enable Boehringer to develop and offer new treatments to patients with significant unmet needs.”

To achieve this, the company relies on collaboration with IBM’s research team. Information on the genetic sequence, structure and molecular profile of disease-relevant targets, as well as success criteria for antibody molecules of therapeutic interest, such as affinity, specificity and scalability, should serve as basic criteria for the generation of new human antibody sequences by simulation.

At the heart of Boehringer’s project is a pre-trained AI model developed by IBM, which will be further refined with additional proprietary data from the lab, which employs 52,000 people worldwide (including nearly 3,000 in France). To do this, the company is leveraging IBM’s latest foundation model technologies, which are expected to make antibody development faster and more efficient and improve the quality of antibody candidates suggested by the algorithm.

Identify the best candidates

These basic models, which have been shown to be effective in generating biologics and small molecules with target-relevant affinities, are used to design candidate antibodies to the defined target molecule, protein or gene to be tested for, for example, inactivation. The results are then tested using AI-based simulation so that the best binders can be selected and refined for the particular goal.

AdvertisingIn a validation phase, Boehringer then plans to produce the candidate antibodies on a laboratory scale and evaluate them experimentally. Subsequently, the results of the laboratory tests will be used to improve computer simulation methods through feedback loops.

Proprietary data beyond public databases

IBM’s biomedical modeling technologies used for this purpose are based on a large amount of publicly available data. In particular, databases of protein-protein interactions and drug-target interactions on which models are trained. These models can then be refined by partners, such as Boehringer, using specific proprietary data to produce engineered proteins and small molecules with desired properties.

The collaboration with IBM is just one of many AI projects carried out by the laboratory in Ingelheim am Rhein. The company is currently establishing a digital ecosystem with partners from academia and business to discover and develop medicines faster and create services that can improve patients’ lives. For example, the lab is working with Zeiss medical technology to develop predictive analytics that enable early detection of eye diseases and prescription of personalized treatments to prevent blindness in people suffering from serious eye diseases.

To do this, the partners rely on AI-assisted analysis of large sets of image data. The goal is to create a basis for clinical studies to develop personalized and more precise treatments in the early stages of chronic retinal diseases, to improve the prognosis of vision preservation in affected patients. Via the same approach, AI must also lead to earlier detection of eye diseases.

Share this article



Source link

Leave a Comment