Nature Sub-Journal: Artificial Intelligence Reveals "Important Genes" in Agriculture and Medicine

According to a new study published in Nature Communications, machine learning can pinpoint "important genes" and help crops grow under less fertilizer

Using genomic data to predict the outcomes of agriculture and medicine is both a hope and a challenge for systems biology

In the study of "Nature Communications", researchers from New York University and collaborators in the United States and Taiwan used machine learning (an artificial intelligence used to detect data patterns) to solve this challenge

Maize (corn) grows in the Rose Sunzege greenhouse on the roof of the Center for Genomics and Systems Biology of New York University

"Our research shows that focusing on genes that are evolutionarily conserved across species expression patterns can improve our ability to learn and predict the'important genes' of major crop growth performance and animal disease outcomes," New York University Department of Biology and Genomics and Systems Professor Carroll and Milton Petrie of the Center for Biology, who are also the senior authors of the paper, explained

"Our method takes advantage of the natural variation of whole-genome expression and related phenotypes within or across species," added Cheng Jiayi of the Center for Genomics and Systems Biology of New York University and National Taiwan University, who is the lead author of the study

As a proof of concept, the researchers demonstrated that two different plants (arabidopsis, a small flowering plant widely used as a model organism for plant biology, and maize, maize, and maize) respond to nitrogen.

Maize (corn) grows in the Rose Sunzege greenhouse on the roof of the Center for Genomics and Systems Biology of New York University

The researchers verified the importance of eight master transcription factors to nitrogen use efficiency through experiments

"Now we can more accurately predict which corn hybrids are better at using nitrogen fertilizer in the field, and we can quickly improve this trait

In addition, the researchers demonstrated that this evolution-based machine learning method can be applied to other traits and species by predicting other traits of plants, including the biomass and yield of Arabidopsis and corn

Coruzzi said: "Because we have shown that our evolutionary information pipeline can also be applied to animals, this highlights its potential and reveals important genes for any physiological or clinical characteristics of interest in biology, agriculture or medicine

"Many important agronomic or clinical features are genetically complex, so it is difficult to determine their control and inheritance

Original search:

“Evolutionarily informed machine learning enhances the power of predictive gene-to-phenotype relationships” 24 September 2021, Nature Communications .