TL;DR:
- Atomic AI introduces ATOM-1™, a groundbreaking Large Language Model (LLM) leveraging chemical mapping data.
- ATOM-1™ accurately predicts RNA structure and function, enhancing RNA drug discovery.
- RNA-based therapies have immense potential in treating various diseases, but challenges exist due to limited data.
- ATOM-1™ transforms RNA research by gathering vast datasets and outperforming existing methods.
- This innovation opens doors to efficient drug development and RNA therapy optimization.
Main AI News:
In the realm of biotechnology, where innovation converges with scientific excellence, Atomic AI is revolutionizing the landscape. By seamlessly integrating state-of-the-art structural biology with advanced machine learning capabilities, the company has taken a monumental step forward. Their recent revelation, the first Large Language Model (LLM) utilizing chemical mapping data, marks a significant milestone in the pursuit of unlocking the potential of RNA drug discovery.
In a groundbreaking preprint paper unveiled on bioRxiv, Atomic AI introduces the world to ATOM-1™, their proprietary platform component. This innovation promises to redefine the way we perceive RNA, as it possesses the unprecedented ability to predict RNA structure and function with unparalleled accuracy. The implications are profound, as ATOM-1™ is poised to revolutionize the development of RNA therapeutics across diverse medical domains.
The emergence of messenger RNA (mRNA)–based COVID-19 vaccines has undeniably illuminated the remarkable potential of RNA-based and RNA-targeting therapies. These therapies extend their reach beyond infectious diseases to encompass formidable adversaries like cancer and neurodegenerative conditions. However, a persistent challenge looms large—designing and discovering RNA therapeutics is hindered by the scarcity of data needed to predict RNA’s structure and function accurately.
The dearth of “ground-truth” data has hindered progress, primarily due to the inadequacies of existing approaches. Traditional methods, such as animal models and cryo‐electron microscopy (cryo-EM), are not only laborious but also time-consuming. Consequently, optimizing crucial RNA therapeutic attributes, including stability, toxicity, and translational efficiency, has proven to be a formidable task.
Dr. Manjunath “Manju” Ramarao, Chief Scientific Officer of Atomic AI, offers insight into ATOM-1’s transformative capabilities. “ATOM-1 enables the prediction of structural and functional aspects of RNA as well as key characteristics of RNA modalities, including small molecules, mRNA vaccines, siRNAs, and circular RNA, to aid in the efficient design of therapies,” he states. Atomic AI’s mission is clear—to streamline the drug discovery process, expedite their own pipeline, and collaborate with partners to accelerate the validation of RNA targets and tools. Ultimately, this collaborative effort aims to deliver much-needed therapeutics to patients swiftly and efficiently.
In the paper titled “ATOM-1: A Foundation Model for RNA Structure and Function Built on Chemical Mapping Data,” Atomic AI’s team of researchers reveals their groundbreaking platform component. Leveraging large-scale, in-house-collected chemical mapping data obtained through custom wet-lab assays, the scientists have meticulously gathered data on millions of RNA sequences, amounting to over a billion nucleotide-level measurements. ATOM-1, trained on this extensive dataset, has acquired an unparalleled understanding of RNA, thereby unlocking the potential to optimize the properties of diverse RNA modalities.
Dr. Stephan Eismann, Founding Scientist and Machine Learning Lead at Atomic AI, underscores the uniqueness of their achievement. “By building large datasets based on RNA nucleotide modifications and next-generation sequencing, the team at Atomic AI has created a first-of-its-kind RNA foundation model,” he remarks. This model holds the promise of transforming various facets of RNA research and enhancing the properties of RNA-based medicines. Whether it’s bolstering the stability and translation efficiency of mRNA vaccines or fine-tuning the activity and toxicity of siRNAs, ATOM-1 is poised to make an indelible mark.
Compared to previously published methods, ATOM-1 stands out for its superior ability to predict RNA secondary and tertiary structures. In a retrospective analysis, it outperformed all 1,600 other computational tools in predicting in-solution mRNA stability. These remarkable results highlight ATOM-1’s adaptability—it can be finely tuned with a limited amount of data to predict various properties of RNA. Beyond deciphering RNA structure, it can also forecast other pivotal features of RNA therapies.
Dr. Raphael Townshend, Founder and CEO of Atomic AI, reflects on the journey that led to this transformative achievement. “Over the last two and a half years, we’ve been purposefully designing and collecting data to train our foundation model,” he explains. With the power of machine learning and generative AI, ATOM-1 presents a unique opportunity to predict RNA structure and function with unparalleled precision, requiring only a minimal initial dataset.
Conclusion:
Atomic AI’s ATOM-1™ represents a game-changing advancement in the field of RNA therapeutics. Bridging the gap in RNA data availability and offering precise predictions has the potential to accelerate drug development, making RNA-based treatments more accessible and effective for a wide range of medical conditions. This development is poised to reshape the market, attracting increased interest and investment in RNA-related research and therapies.
