Baidu Open-Sources LinearFold Algorithm to Accelerate Coronavirus Research
Baidu releases an AI tool that cuts RNA secondary structure prediction time from 55 minutes to 27 seconds, offering crucial support to scientists fighting the 2019-nCoV outbreak.
Baidu open-sources its RNA prediction algorithm LinearFold to help researchers combat the 2019 Novel Coronavirus outbreak. This AI tool significantly accelerates the prediction of a virus's RNA secondary structure, giving frontline scientists a critical advantage in understanding the virus and developing targeted vaccines during the crisis.
Traditional algorithms face a major challenge because their runtimes scale in a cubic manner as RNA length increases, making them too slow for long viral sequences. LinearFold solves this by becoming the first approximate algorithm in RNA folding to achieve linear runtime and linear space, scanning sequences from left-to-right instead of using a bottom-up approach.
Baidu's AI scientists apply LinearFold to the 2019-nCoV, reducing the prediction time from 55 minutes to just 27 seconds. The algorithm proves highly efficient across various tests, processing the HIV genome in 8 seconds compared to 4 minutes for baseline systems, while maintaining high accuracy levels for all RNA families.