Allen Institute’s AI-driven brain decoding
Mufti’s team at the Allen Institute is aiming to map the human brain at a cellular level. Their work has led to the creation of a genome-scale Allen Mouse Brain Atlas, a comprehensive, freely accessible online resource for neuroscience research, housing over 20,000 genes and more than 650,000 images.
The Allen Institute uses several AWS Cloud services to manage and analyze the vast amounts of data. They use Amazon’s data lake for analyzing large, diverse datasets, Amazon S3 (Simple Storage Service) for data archiving and backup, DynamoDB for handling large-scale databases and SageMaker for building, training and deploying machine learning models.
The Allen Institute’s strategy for data democratization involves enabling wide accessibility to complex data through natural language processing (NLP) and the use of Amazon Bedrock. Bedrock, an AWS service, supports the development and scaling of applications based on generative AI and foundation models.
Large-scale collaboration for future goals
The Allen Institute’s work is part of a broader collaboration with funding from the National Institutes of Health’s Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative and The BRAIN Initiative Cell Atlas Network, or BICAN. In addition to human brain mapping, the initiative also aims to create detailed atlases of macaque and marmoset brains.
The AWS Cloud platform facilitates collaboration across diverse communities. It serves not only as a repository for data but also as a hub for applications that can interact with this data, acting as an ‘App Store’ for data and applications.
Objectives and future plans
The short-term goal of the initiative revolves around enhancing natural language processing (NLP) capabilities, allowing users to interact with the system using natural language queries. In the medium-term, the focus shifts to machine learning pipelines for data processing, from labeling to segmentation, which can lead to significant time and effort savings for researchers. In the long run, the institute aims to develop a large language model capable of understanding seemingly unrelated data sets, potentially unlocking new discoveries within the extensive data generated by brain mapping studies.
Using both AI and the cloud, the Allen Institute aims to bring together diverse communities and leverage the broader ecosystem. “Services like the Amazon Marketplace are crucial in this regard,” Mufti noted. “We aim to create not just a data repository but also a platform for applications that can interact with this data. We’re building an ‘App Store’ where users can either bring their own apps or data.” This strategy underscores the Allen Institute’s commitment to innovation and community engagement in their AWS-assisted quest to decode the brain’s mysteries.
Filed Under: Brain Breakthroughs, Data science, machine learning and AI, Neurological Disease