Software that is revolutionary has been created by researchers from the National University of Singapore and the Agency for Science, Technology, and Research. This program has a high degree of accuracy in predicting chemical changes to RNA molecules. The team’s approach, known as m6Anet, was made public in the esteemed academic journal Nature Methods.
RNA molecules contain many chemical compounds that govern how they work. However, common methods employed by scientists to read RNA typically fail to detect these RNA modifications. The most common RNA modification is N6-Methyladenosine(m6A). Finding RNA modifications has historically taken a long time and proven difficult because they are related to human diseases like cancer.
By utilizing direct Nanopore RNA sequencing, the researchers were able to get beyond these constraints. This cutting-edge method sequences unmodified RNA molecules along with their changes. They produced m6Anet. By leveraging a Multiple-Instance Learning (MIL) technique and direct Nanopore RNA sequencing data, the software trains deep neural networks to detect m6A accurately.
Each example in traditional machine learning is assigned one label. However, finding m6A calls for an enormous volume of data with unclear labels. To resolve this issue, the team applied the MIL method. The MIL issue entails having a sizable photo album with a cat picture buried among millions of other images. Then, without any labels to use as a guide, try to identify that specific image.
The scientists showed that m6Anet can forecast the presence of m6A from a single sample across species with high precision at a single-molecule resolution. The ability to recognize RNA alterations in various biological samples can be utilized to comprehend their significance in a variety of applications, such as cancer research or plant genomics. This is according to Dr. Jonathan Goke, Group Leader of the Laboratory of Computational Transcriptomics at ASTAR’s GIS.
The AI model has only come across data from a human sample. Even samples from species the model has never encountered before can be used to precisely identify RNA modifications. “The MIL method provides a sophisticated answer to this difficult issue. A reward for our work is seeing the program get adopted by the scientific community so quickly!” affirmed study co-leader Associate Professor Alexandre Thiery, Department of Statistics and Data Science, NUS Faculty of Science.
The scientific community can now access and utilize the study’s software and findings. The long-standing problem of precisely and effectively identifying RNA alterations is addressed by m6Anet, according to Professor Patrick Tan, Executive Director of ASTAR’s GIS. Researchers in several sectors can advance their work with this ground-breaking technology. They will also be able to comprehend the function of RNA modifications in plant genomics and human diseases like cancer.