PALO ALTO, Calif., Dec. 19, 2024 /PRNewswire/ -- Just as the transistor and microscope unlocked new waves of discovery, today, GenBio AI opens the doors to a new era of scientific exploration with Phase 1 of the world's first AI-Driven Digital Organism (AIDO).
In this first phase, scientists will have access to not one, but six groundbreaking models that can simulate, program, and predict complex biological outcomes for DNA, RNA, Protein, Protein Structure, Single Cell Expression, and Evolutionary Information.
Unlike earlier models that focus on specific aspects of a multi-cellular organism, GenBio AI takes a holistic approach where multiscale foundation models are built in modules and integrated into a single system. These State-Of-The-Art models are unprecedented in their capabilities:
They are the fastest and most efficient in the market - solving up to 300 diverse tasks simultaneously, compared to one to two tasks.
They are the only ones intentionally built to interact with each other - finally tackling biology and AI's research silo problem.
They allow biological data of all types and scales to be utilized to distill holistic and comprehensive knowledge of how living systems work - the first time this has ever been possible.
GenBio AI is set to fuel an unprecedented era of discovery across molecular, cellular, and organism levels. These "building block" Foundational Models are just the first phase on the company's road map to the world's first AIDO. More details are expected in the year ahead.
Professor Eric Xing, Co-Founder and Chief Scientist of GenBio AI, said: "GenBio will usher in a new era of medical and life science—through a paradigm shift powered by next-generation Generative AI technology beyond what has already brought us disruptive results such as ChatGPT. Our transformative technology allows biological data of all types and scales to be utilized to distill holistic and comprehensive knowledge of how living systems work. Therefore, multiscale biological complexities are no longer barriers but opportunities for breakthrough insights."
The company's world-class and internationally recognized team of scientists, scholars, and engineers are leading experts in AI, machine learning, high-performance computing, and computational and experimental biology. The team published all six technical papers detailing their novel approach to creating the world's first AIDO at the 38th Annual Conference on Neural Information Processing Systems (NeurIPS) – the top AI research conference.
By sharing this work, GenBio aims to help guide the community forward, enabling experts to understand the possibilities to observe, experiment with, and eventually program living systems with unprecedented precision across molecular, cellular, and systemic levels.
Breaking Down Biological Barriers - Detailing GenBio AI's Models
GenBio AI is the first company to release multifunctional models covering mainstream biological data - namely DNA, RNA, Protein, and Cell - and are poised to empower researchers to drive the next breakthroughs in medicine, drug discovery, and longevity research.
GenBio AI's AIDO-DNA Foundation Model features 7 billion parameters trained on data from 796 different species. This State-Of-The-Art model surpasses previous encoder-only architectures without requiring new data, indicating that new scaling rules may be essential for creating optimal DNA language models.
GenBio AI's AIDO-RNA Foundation Model, with 1.6 billion parameters, is the largest and most comprehensive model of its kind. RNA has emerged as a key technology for vaccine and drug design over the last several years and GenBio AI's model provides faster results for structure prediction, genetic regulation, and RNA sequence design.
GenBio AI's AIDO-Protein Foundation Model is one of the largest and most computationally efficient models of its kind - facilitating deeper exploration of these essential biological elements while maintaining scalability.
GenBio AI's AIDO-Single Cell Model includes a range of models with 3 million, 10 million, 100 million, and 650 million parameters, pre-trained on a dataset of 50 million human cells. This set of cell foundation models are scalable and the first capable of processing the entire human transcriptome as input without truncation or sampling techniques, allowing it to learn precise and comprehensive representations of the complete transcriptional landscape of human cells.
Dr. Le Song, Co-Founder and Chief Technology Officer at GenBio AI, said: "What we have built is revolutionary because our integrated system will use these state-of-the-art models to create interactive digital versions of biological systems that can be safely experimented on and precisely modified. This technology lets us program biology the way we program computers, opening up possibilities we've never had before in medicine and biotechnology."
Professor Xing continues: "Even before their integration into the AIDO system, each Foundational Model from the collection stands out already as a best-of-its-kind among their respective peers. GenBio AI does not just train best soloists, it is building an orchestra made of a herd of the best soloists."
Transforming Medical Research and Drug Development
GenBio AI's Digital Organism will revolutionize three critical areas:
Drug Discovery: Traditional drug development requires $1-2 billion and sees a 90% failure rate. AIDO will enable researchers to design and test millions of potential treatments in hours, enabling the prediction and analysis regarding how they affect entire biological systems before entering a lab.
Disease Understanding: With deaths from lung cancer, heart disease, and Alzheimer's rising across the G20, scientists need new tools to understand these conditions. AIDO will provide researchers with a foundational AI model to build a better understanding of disease mechanisms and potential interventions.
Personalized Medicine: With adverse drug reactions ranking among the top six causes of death worldwide, the need for personalized treatment is critical. AIDO will allow doctors to create digital patient twins, potentially accelerating the development of safe, effective treatment plans while minimizing adverse drug interactions.
Global Expertise, Global Impact
GenBio AI's achievement represents a collaboration of world-renowned scientists from leading institutions, including Carnegie Mellon University (CMU), Stanford, the Weizmann Institute of Science, and MBZUAI. Senior Scientific Fellows, including Eran Segal and Fabian Theis, also add to the company's robust team of internationally recognized experts.
Headquartered in Palo Alto with satellite labs in Paris and Abu Dhabi, GenBio AI takes a global approach to solving these fundamental challenges in biology. The intellectual foundation of this breakthrough was detailed in the six peer-reviewed papers presented at NeurIPS.
Professor Eran Segal, Department of Computer Science, Weizmann Institute of Science says: "GenBio AI's six multiscale foundation models are a leap forward in our ability to understand and predict biological phenomena. We now have the capacity to uncover systemic insights into how organisms function. This is transformative for genomics research, where the ability to simulate and program at multiple scales opens new avenues for precision medicine and disease intervention."
Professor Fabian Theis, Director of the Institute for Computational Biology at Helmholtz Munich and Professor at the Technical University of Munich says: "GenBio AI's achievement in creating scalable state-of-the-art models on multiple scales is a game-changer. This technology not only accelerates our ability to explore cellular dynamics but also bridges the gap between molecular and systems biology, unlocking unprecedented opportunities for disease modeling and therapeutic innovation."
A New Chapter in Biological Science
The launch of GenBio AI marks the beginning of a new era in biological science. This isn't just a new tool for research - it's a fundamental transformation in how we interact with and understand living systems.
Explore the research:
Accurate and General DNA Representations Emerge from Genome Foundation Models at Scale
Mixture of Experts Enable Efficient and Effective Protein Understanding and Design
Balancing Locality and Reconstruction in Protein Structure Tokenizer
Retrieval Augmented Protein Language Models for Protein Structure Prediction
A Large-Scale Foundation Model for RNA Function and Structure Prediction
Scaling Dense Representations for Single Cell with Transcriptome-Scale Context
About GenBio AI
GenBio.AI, Inc. (GenBio AI) is an innovative global startup dedicated to developing the world's first AI-driven Digital Organism, an integrated system of multiscale foundation models for predicting, simulating, and programming biology at all levels.
Our goal is to achieve comprehensive, actionable empirical understandings of the mechanisms underlying all organismal physiologies and diseases. This will pave the way for a new paradigm in drug design, bio-engineering, personalized medicine, and fundamental biomedical research, all powered by Generative Biology.
Our founding team consists of world-renowned scientists and researchers in AI and Biology from prestigious institutions such as CMU, MBZUAI, WIS, alongside prominent financial investors.
GenBio AI, a true global effort from day one, is establishing offices in Palo Alto, Paris, and Abu Dhabi.
For more information about GenBio AI and AIDO, visit GenBio.AI.
SOURCE GenBio AI
