Research

We are adapting the transformer architectures behind Large Language Models (LLMs) to understand the "language" of cells and genes. By training on vast datasets of single-cell profiles and biological interactions, we build "virtual cells" and multi-agent systems that can simulate biological responses and generate novel scientific hypotheses. We also develop mathematical models of disease progression and treatment response.

We use artificial intelligence to solve the complex combinatorial challenge of cancer treatment. Our algorithms predict how tumors will respond to drug combinations and sequences, allowing us to identify synergistic therapies and optimize dosing strategies computationally before validating them in the wet lab.

We bridge the gap between AI research and clinical practice by building intelligent assistants for oncologists. Our tools utilize Large Language Models and retrieval techniques to synthesize medical literature, interpret complex genomic data, and support tumor board decision-making.

We are reimagining disease screening by combining deep learning on Electronic Health Records (EHR) with novel biomarkers. Our two-tier screening approach uses AI to identify high-risk individuals from routine medical data, guiding the targeted use of advanced diagnostics including Multi-Cancer Early Detection (MCED) liquid biopsies and voice as a biomarker of health.

We map the tumor microenvironment in unprecedented detail to understand how cancer cells evade the immune system. Using cutting-edge spatial transcriptomics and imaging, we analyze the architecture of tissues to guide the development of immunotherapies and personalized cancer vaccines, with particular expertise in blood cancers including lymphomas and leukemias. Key tools include UTAG for unsupervised tissue architecture mapping.

We leverage comprehensive genomic profiling to uncover the specific mutations driving a patient's disease. By combining Whole Genome Sequencing (WGS) with AI interpretation tools, we connect genetic variants to clinical phenotypes to diagnose rare diseases and tailor oncology treatments.

We develop and share software tools that enable the broader research community to apply our methods to their own data. All tools available at github.com/ElementoLab.