The US-based biotech business Atomic AI has unveiled ATOM-1, a proprietary platform component that functions as a large language model (LLM) that makes use of chemical mapping information to forecast ribonucleic acid (RNA) structure and function.
The approach aims to tackle the difficulties involved in creating RNA therapies by offering a mechanism that can maximize crucial aspects of RNA modalities, including translational efficiency, toxicity, and stability.
Using specialized wet-lab experiments, scientists at Atomic AI gathered vast amounts of chemical mapping data, including millions of RNA sequences and over a billion nucleotide-level measurements. This dataset is used to train ATOM-1.
“Through machine learning and generative AI, we now have a unique opportunity with ATOM-1 to predict RNA structure and function with high precision by tuning it with just a small amount of initial data points,” said Raphael Townshend, founder and CEO of Atomic AI, in a statement released in conjunction with the launch.
The announcement is made at a time when concerns are raised about generative AI platforms’ data accuracy. The most recent version of OpenAI’s LLM, GPT-4, has been shown in a recent research published in JAMA Ophthalmology to be able to create bogus datasets. When asked to find evidence supporting a specific conclusion, the AI can generate semi-random datasets based on a predetermined set of characteristics.
The healthcare sector has a major presence of artificial intelligence (AI). A report on GlobalData’s Medical Intelligence Center projects that by 2024, the market for AI platforms in the healthcare sector as a whole would grow to $4.3 billion. According to GlobalData, the value of the total AI market worldwide will reach $383.3 billion by 2030.
Google stated in April 2023 that Med-PaLM 2, its limited access LLM, would be available to the healthcare industry for use in providing precise and secure medical answers to queries. Clinically supported criteria, such as scientific consensus, medical rationale, prejudice, and potential damage, were used to assess the device.
