The world’s largest bioresource of eye images offers window into human health and serves as basis of a new AI foundation model to boost drug discovery.
insitro, a machine learning-enabled drug discovery and development company, and the INSIGHT Health Data Research Hub at Moorfields Eye Hospital, London, today announced a new collaboration to advance drug discovery. The collaboration will develop a novel AI foundation model to aid in genetic discovery of new ocular biomarkers for precision patient segmentation and therapeutic targets for neurodegenerative and related conditions.
INSIGHT is the world's largest ophthalmic imaging bioresource, containing 35 million eye images, including an extensive repository of optical coherence tomography (OCT) images linked to several decades of clinical information. These data include diagnostic information for various chronic disorders of ageing, such as neurodegenerative, ophthalmic, and metabolic diseases. Recent peer-reviewed studies enabled by INSIGHT have shown how AI foundation models built on OCT data can predict neurological and other diseases years before symptoms appear.
insitro CEO and Founder, Daphne Koller.
“OCT data at this scale contains a treasure trove of information about human health, providing a new window into diseases of the brain, the eye, and more,” said Daphne Koller, CEO and founder of insitro. “Building a world-class foundation model with INSIGHT at Moorfields will help us unravel disease biology and identify novel targets, including for our programs in neurodegenerative diseases. We are delighted to broaden our network in the UK health research ecosystem, allowing us to advance our mission of finding new and better medicines while ensuring the safe use of patient data.”
The model will support insitro’s neuroscience programs by identifying OCT-based signatures linked to dementia risk and disease progression. As part of the NHS, Moorfields Eye Hospital collects large volumes of OCTs before dementia diagnosis, enabling the development of a model that can differentiate OCTs of those who develop dementia from those who don't. The vast scale of OCT images curated by INSIGHT, linked to dementia diagnosis information, is globally unique and will enable the generation of a high-quality model for insitro to use for drug discovery.
“Working with insitro to develop a novel foundation model serves INSIGHT’s mission to benefit patients, the NHS and wider society through the development of new tools for disease detection, diagnosis, and treatment,” said Professor Pearse Keane of INSIGHT at Moorfields. “This will be the first collaboration of its type for INSIGHT at Moorfields and represents a mutual commitment with insitro to unlocking new discoveries to improve human health using clinical data and AI. We have designed the collaboration to ensure the safety and privacy of patient data, with only INSIGHT at Moorfields researchers accessing the data, in a secure environment.”
Professor Pearse Keane.
The model will support insitro’s genetic discovery efforts to identify differentiated targets by being applied to research cohorts that have genetic data but contain substantially smaller numbers of OCT cases with dementia diagnosis. By learning representations of OCT data from large-scale datasets and fine-tuning on disease-specific cohorts, the model will accelerate insitro's ability to discover new biological targets supported by human genetics.
The joint work will utilise eye research data infrastructure developed by INSIGHT at Moorfields which has enabled previous pioneering work in Oculomics. This rapidly growing research field uses ocular biomarkers to detect systemic disease, underpinned by high-resolution eye imaging such as OCT and the power of data science.
