DeepMind is building a single system that aims to read regulatory DNA as a unified code. IBM’s approach focuses on decomposing biological questions into well-defined tasks using models that are optimized for the mathematical and biological structure of each domain.
“Our work with biomedically based models (BMFM) is taking a more hands-on, modular approach,” Michal Rosen-Zvi, director of AI for healthcare and life sciences at IBM Research, said in an interview. IBM thinks. “We break down complex biological problems into well-defined components and identify the mathematical and algorithmic innovations needed for the specific task at hand.”
Based on this analysis, IBM is developing specialized models for different domains, including RNA transcriptomics, DNA sequence analysis, and small molecule and protein expression, Rosen-Zvi said. “Each model is designed to best capture the modality most relevant to that domain, such as primary sequence, two-dimensional structure, three-dimensional conformation, or, in the case of our RNA model, a mathematical representation that more faithfully captures whole-genome expression at the cellular level,” she said.
Rosen-Zvi said IBM’s DNA research seeks to avoid treating the genome as a single “standard” sequence. “Importantly, our DNA models explicitly incorporate population-level variation and are trained not only on reference sequences but also on SNPs and other sites of variation,” she said. This design allows the model to learn evolutionary and functional signals that cannot be captured by a single reference genome, signals that may require training on thousands of whole genomes to approximate, Rosen-Zvi explained.
Rosen-Zvi framed the biomedical basic model as a powerful and practically actionable tool. “Overall, the BMFM approach emphasizes efficient training and inference, and is particularly suited to problems where the underlying biology spans multiple layers of information, abstraction, and observation,” she said. In her view, that is precisely the territory scientists must cross when trying to explain diseases, identify drug targets, propose mechanisms of action, generate candidate compounds, and predict which compounds are worth pursuing.
IBM has focused its recent modeling efforts on two areas of drug development that tend to be time-consuming and expensive: biologics and small molecules. She pointed to IBM’s MAMMAL, which is designed to predict the binding strength of antibodies and antigens. He also highlighted the excellent performance of IBM’s MMELON in predicting the therapeutic properties of small molecule candidates, saying it can provide early information to help research teams decide what is worth pursuing before they begin research.
A new IBM paper co-authored with the Cleveland Clinic explains more clearly how MMELON works. It describes a “multi-view” method for representing molecules, published in a paper by IBM Research as an example of domain-specific foundational models in biomedicine. This project stems from the Discovery Accelerator partnership between IBM and Cleveland Clinic. The two organizations announced a collaboration that will use AI and quantum computing to accelerate biomedical discoveries.
IBM Research is also involved in larger data-building efforts. The company recently joined LIGAND-AI, a consortium announced in January 2026 that aims to generate open, high-quality datasets of protein-ligand interactions. The consortium, led by Pfizer and the Structural Genomics Consortium, includes 18 partners from nine countries, according to a project announcement.
Organizers say the initiative has a budget of more than 60 million euros and will investigate thousands of proteins related to both existing treatments and major unmet needs such as rare diseases, neurological diseases and cancer. The Structural Genomics Consortium said the project plans to use complementary screening techniques to generate billions of data points, creating a resource that researchers around the world can use to train and benchmark AI systems that predict molecular interactions.
The market for AI in biotechnology is rapidly expanding. Precedence Research projects continue to experience double-digit growth globally, with the market estimated to exceed USD 25 billion by the mid-2030s, according to a January 2026 analysis by Ardigen. The US market alone will be worth approximately USD 2.1 billion in 2025, with growth primarily driven by adoption in drug discovery, genomics, and precision medicine, according to the analysis.
