Israel invests 70 million shekels in a common language for biological data so AI models can learn from hospital and laboratory data
Israel invests 70 million shekels in a common language for biological data so AI models can learn from hospital and laboratory data
The Israel Innovation Authority is forming the IBFI consortium with NVIDIA, Teva, Sheba Medical Center, biotechnology companies, and universities. Its purpose is to establish a shared framework for representing different types of biological data and to provide participants with an environment for developing and testing models.
On 12 July, the Israel Innovation Authority announced IBFI. Data on cells, proteins, tissue images, and medical histories share an inconvenient feature: they describe the same organism but are stored in separate systems. A laboratory measures gene activity, a hospital maintains clinical records, and a pharmaceutical company observes responses to a drug. A model can identify relationships among these pieces only if the data can be matched.
IBFI aims to build a common infrastructure for this purpose. Its first component, Bio Tokens, is intended to provide a standard way to represent molecular, cellular, time-varying, and clinical data. The second, Factory Model, is conceived as an environment in which participants can build models, connect them to one another, run them, and validate their results.
The announced consortium brings together participants that usually work separately: the technology company NVIDIA, the pharmaceutical company Teva, a hospital, biotechnology startups, and research groups. The government and its partners are allocating about 70 million shekels to this work. The initial projects concern tumor responses to treatment, the immune system, sepsis, transplant rejection, autoimmune diseases, and drug discovery.
A model does not acquire an understanding of biology merely by having a large number of parameters. It sees data in whatever form researchers have managed to collect and connect. The future value of IBFI therefore depends on more practical questions: whether participants can agree on compatible formats, provide secure access to the necessary data, and verify that a model works with patients and in laboratories that it has not encountered before.
Aging research depends particularly heavily on this type of integration because researchers need to relate the state of cells and tissues to a person's ability to maintain muscle strength, memory, immunity, and resistance to disease over many years. IBFI does not yet have a dedicated aging program. The consortium is creating a general research platform. Its quality will be determined by the data, models, and validation procedures that become available after its launch.