Chinese scientists create AI “Meta-Sorter” to label microbial samples
14 Sep 2023 --- Meta-Sorter, an AI-based program that leverages neural networks to improve biome labeling for thousands of microbiome samples, has been unveiled by Huazhong University of Science and Technology researchers.
The samples are captured in the MGnify database and ones with incomplete information are prioritized.
Meta-Sorter meticulously constructs a neural network model using 118,592 microbial samples from 134 biomes and their respective biome ontology. The model accurately classifies samples with detailed biome information and is regarded as a strong foundation for further analyses.
To address the challenge of newly introduced samples with different characteristics, researchers incorporated transfer learning with 34,209 recently added samples from 35 biomes, including eight novel ones. The transfer neural network model successfully predicts biome information for newly introduced samples annotated as “mixed biome.”
Predictive technology
Meta-Sorter facilitates intelligent and automatic knowledge discovery from microbial communities and achieves an overall accuracy rate of 96.7% in classifying samples among the 16,507 lacking detailed biome annotations.
A study about the program’s capabilities and related research findings was published in Environmental Science and Ecotechnology, an international peer-reviewed journal Elsevier.
Another feature of Meta-Sorter is it prevents the cascading accumulation of inaccurate labels in environmental scientific research.
With the ongoing development of standardized protocols for data submission and incorporating additional meta-data information, Meta-Sorter sets a new standard in how researchers analyze and interpret microbial community samples. It will lead to more accurate and insightful discoveries in microbiome research.
Meta-Sorter has a near-perfect accuracy rate for classifying annotated samples of the biome.
In 2022, the gut microbiome was pegged as the “black box” of nutrition research as diet-microbiome interactions were anticipated to contribute to the foundation of dietary physiological effects. According to researchers from the University of Alberta, Canada and the University College Cork, Ireland, dietary guidelines could be improved, modified and innovated based on data on diet-microbiome-host connections.
Annotating samples
Meta-Sorter also refines the biome annotation for under-annotated and mis-annotated samples. According to the researchers, its intelligent and automatic assignment of precise classifications to ambiguous samples provides valuable insights beyond the literature.
Moreover, they note that the differentiation of samples into specific environmental categories enhances the reliability and validity of research conclusions.
The research has been funded by the National Natural Science Foundation of China and the China Ministry of Science and Technology’s National Key R&D Program. The study is endorsed by the Chinese Society for Environmental Sciences.
Meanwhile, research into the gut microbiome’s role in health continues to advance. US-based scientists at Harvard Medical School found a link between suppressing emotions and gut health among women but found no association with diets. Meanwhile, a China-based study found a link between Western diets and Crohn’s disease – a chronic inflammatory bowel disease in the gastrointestinal tract.
According to nutrition giant ADM, the microbiome and personalized nutrition are set to take center stage across the industry in the next four years, as consumer interest in immunity is slated to continue.
By Inga de Jong
